#qi-hardware IRC log for Saturday, 2011-02-26

wpwrakrjeffries: (quote) that shop seems hopeless. that's then for an even simpler variant than what tuxbrain is getting01:58
wpwrakrjeffries: in about a week we should know how things went in spain. i think it's neither inaproppriate nor impolite for you to mention in further conversation with paula that you know of a considerably less expensive quote that a company in spain is using. this does not imply that you weren't serious or anything. after all, if you had been able to obtain a lower price, maybe you would have tried to compete in the future (with a few ifs02:01
wpwrak and whens, obviously)02:01
wpwrakrjeffries: also, i believe you can get a much lower price also in the US. unless there's some magic feature that makes producing this board expensive and that the others have overlooked so far, tuxbrain's price point would seem feasible also in the us, e.g., at 4pcb.com02:03
wpwrakrjeffries: whether there _is_ a magic problem feature or not, we'll see in a week :)02:03
rjeffrieswpwrak I have assumed teh cheapie fab houses can not do the thin fiberglass UBB requires02:11
rjeffriesbut after I shut down this snipe hunt I can try02:11
wpwrakrjeffries: (cheapie) 4pcb.com seem pretty advanced :)02:13
wpwrakrjeffries: but you're right, the rock bottom ones, batchpcb and such, wouldn't know what to do with a thin board02:14
rjeffries4pcb will be worth a try yes02:15
rjeffrieshave you used their web interface02:16
wpwrakyes, i ran a test quote for ubb there. alas, they don't seem to have fractional inch sizes, so the quoted board would have been twice the needed size02:17
wpwraki didn't ask for a "manual" quote02:17
wolfspraulrjeffries: nice ROS link - thank you! never heard of that before...02:48
wolfsprauland it seems even the freedom requirements are satisfied :-) the wiki is cc-by licensed, the source code bsd licensed02:49
wpwrakwolfspraul: btw, i think i have a plan for the shipments: send to roh with regular mail and adam with fedex in parallel, then wait a few days and send another with regular mail to richard. so unless they're really good at crossing that data (which they usually aren't, although IT usage is improving), they won't find the match between roh's and adam's. for richard's, we'll have to hope for the best.03:11
wpwrakgrmbl. i wish i could give xchat a completion preference, so that the usual wo<tab> completes to wolfspraul, not woakas :-(03:14
wolfspraulfedex to Adam will be very expensive03:15
wpwrakabout USD 60, it seems03:16
wolfspraulI doubt they will accept it as 'document', they are quite strict in limiting that to actual PAPER :-)03:16
wolfspraulalthough not sure, see electronic birthday card etc. I don't know.03:16
wpwrakoh, that's okay. i think they'll be okay with the "pro forma invoice"03:16
wolfspraul60 usd is not too bad, I would have thought it's more03:17
wpwrakremember, i can send once per month pretty much whatever i want ;-)03:17
rjeffrieswplfspraul thanks (re ROS it does look cool)03:17
wolfspraulyour plan is good because you try multiple ways and gain some experience.03:17
wpwrak(price) i've gotten a range of prices between USD 56 and 100+. i just hope the price i get at the counter is what they advertize online :)03:18
wolfspraulrjeffries: do you have any idea about resource requirements?03:18
wpwrak(plan) thanks :) i've had minor hickups with outbound shipments, but not from customs but the courier company asking for more details.03:18
wolfspraulI couldn't figure out after first glancing over it.03:18
rjeffrieswpwrak in any case these protos will never need to be shipped in volume because wolfspraul plans to manufacture i assume03:19
wpwrakand i think i can solve all their questions with the schedule b harmonized code03:19
wolfspraulmaybe it's very desktop centric, so resource requirements are high because nobody cared to optimize, and the software only makes sense if you have tons of stuff connected via a USB hub or so.03:19
rjeffrieswolfspraul no idea re resouces ysage of ROS I'll poke around03:19
wolfspraulrjeffries: remember, this is copyleft hardware, we very much try to create income opportunities everywhere, equally. So referring to a 'we make all this in China anyway' is not what most of us are after, actually.03:20
wpwrakrjeffries: yeah. the problem are multiple shipments. there seems to be a silly restriction that you can't send more than one shipment per month and person03:20
wolfspraulI doubt I will be in China for that long, every year I feel slowly I've seen and had enough, need to move on.03:20
wolfspraulok let's say 'a few more years' first. then africa :-)03:20
wpwrakrjeffries: so if they truly enforce it, it would be bad if two idential-looking parcels would show up in front of the same customs official within minutes of each other03:21
wolfspraulthe most important thing that happened on the copyleft side in recent weeks was that kristian paul made a milkymist one case with roh's files locally in Bogota, for 15 USD! (including shipping)03:21
rjeffriesI understand real estate in Libia is VERY reasonable now.03:21
wolfspraulthe business opportunities need to emerge not only at the 'center' (whereever that may be), but at all nodes equally. and they will.03:22
wpwrakrjeffries: (enforcement) sometimes they just shrug simple things through, even if they're not 100% according to the rules. but you never know.03:22
rjeffrieswpwrak I guess we will not select Argentina as the next sharism.cc world HQ03:23
wpwrakrjeffries: (libya) you should see their southern neighbours ;-)03:23
rjeffriesneed to check google maps;)03:23
wolfspraulwhy not?03:24
wpwrakrjeffries: (argentina) oh, if you do all the stuff needed to become a commercial importer/exporter, then you enjoy a lot more flexibility03:24
wolfspraulI am in China to dig into the supply chain, extract some of the mysteries and secrets connected to it. Not to entrench myself.03:25
wpwrakrjeffries: also, the government is unlikely to mess with you, because their focus is on milking agricultural exports. so you're not on their radar.03:25
wolfspraulthe worst that could happen is if copyleft hardware is connected to some untracable chinese street vendors, with unknown environment standards, zero documentation, unknown social standards, etc. it won't happen.03:25
wolfspraulso forget China, China is just a country like many others. Moving the sharism hq to Buenos Aires sounds like a very promising proposition to me :-)03:26
wolfspraulmy work in China is not finished though, not yet03:26
wolfspraulman if I could move to Buenos Aires, that would be something :-)03:28
wolfspraulyou make me thinking here, Ron03:28
wpwrak;-))03:28
wolfspraulstill want to work with a few more vendors first, LCM fabs, foundries. they will be hard to come by in Argentina...03:29
wpwrakwolfspraul: i must warn you, though: rafa is moving away from buenos aires, so we'll have barbecues a bit less often.03:29
rjeffrieswolfspraul selecting location based on quality of wine and appearance and plentitude of beautiful women seems questionable yo me03:29
wpwrakrjeffries: you have to pick your priorities :)03:30
wolfspraulrjeffries: when can I welcome you as my new neighbor in Beijing?03:30
wolfspraulI'll prepare a few survival essentials for you already...03:30
wpwrak"when in a restaurant, always order two portions"  :)03:31
wolfspraulfor me this is all simple. I worked in the US for 10+ years, on smartphones. and I slowly started to become more curious about where these things that would be sent to us by Motorola, Nokia, Palm, Sony Ericsson etc. would come from, how they were made.03:33
wolfspraulsince I was always digging around low level software, I guess I wanted to look a bit deeper still.03:33
rjeffriesI am looking at Chad and Niger, hard call choosing between them. both are big opportunities03:33
wolfspraulI knew nothing back then, zero. just a black box of 'hardware', falling from the sky (delivered by a courier).03:34
wpwrakwolfspraul: you still have to crack displays and ASIC-making03:34
wolfspraulwpwrak: exactly, see above :-) LCM and foundry03:34
rjeffriesand designing tools for injection modled parts03:34
rjeffriesmolded03:34
wolfspraulso compared to back then (say early 2007), man I've learnt a lot! cannot believe!03:34
wolfspraulrjeffries: nah, we are not stupid. it's not the flintstones here.03:34
wpwrakwolfspraul: (above) ah yes ;-)03:35
rjeffriesthink how cheap I could live in Chad. I wonder if they have good broadband? that's my only requirement03:35
rjeffrieshow is their airport?03:36
wpwrakrjeffries: broadband may be great. i would worry more about "old" infrastructure. water, electricity, ...03:36
wpwrakrjeffries: you mean  air traffic control ? :)03:37
rjeffrieswater? i  will take tablets. ;)03:37
wpwrakrjeffries: do you have tablets that produce humidity when you chew them dry ? :)03:37
rjeffriesI once flew in a small plane from Cancun to Cozomel. in flight I noticed the door was tied partially open.03:38
rjeffrieswe landed safely03:38
wpwrakeven in buenos aires we have some troubles with water. not on the overall supply side but a lot of the infrastructure in buildings seems to be on the marginal side. poorly maintained on top of it.03:38
rjeffrieswhat about bottled water? that could work, no? ;)03:38
wolfspraulwpwrak: do you know small smt/reflow shops in Buenos Aires?03:38
rjeffrieswpwrak but you have such beautiful woemen, who cares about the plumbing?03:39
wpwrakwolfspraul: i think there are some places, but i never really investigated or made contacts03:39
wpwrakrjeffries: they don't like smelly men :)03:39
rjeffrieswolfspraul: I highly recommend wpwrak ;s apartment as a reflow vendor03:39
wpwrakrjeffries: hmm, i have a toaster oven that done some (really bad) reflow ...03:40
rjeffriesyou need to hack a Ben as a controller03:40
wpwrakyeah, i should get a better model and give it some real process control system03:41
wpwrakbut then i'd also have to make stencils. haven't looked into that yet03:41
rjeffrieswolfspraul may I ask how long you have lived in China? I assume since OpenMoko took a dive?03:44
wpwrakwolfspraul: for things you can't get done in argentina, there's always brazil. also with somewhat lowered trade barriers, but i don't know if this just means less fees and taxes or also less paperwork03:44
rjeffriesI vote Brazil. they have  acool guy running the place03:45
wolfspraulrjeffries: when I saw China I mean a mix of Taiwan, Hong Kong, mainland China. I moved here before Openmoko, in early 2007.03:45
wolfspraulthen I looked for a job, and found Openmoko03:46
rjeffriescool beans03:46
wolfspraulOpenmoko dragged me to Taiwan for 90% of about 1.5 years, which was a great and unexpected experience.03:46
rjeffriesso was openmoko team geographically distributed I assume03:46
wolfspraulotherwise I would have never understood Taiwan (as opposed to mainland China) as well as I do now.03:46
wolfspraulbut I was in mainland (primarily) before I went to OM, and after I left OM.03:46
wolfspraulyes openmoko was very global. in taipei, mainland china, and engineers in the rest of the world.03:47
rjeffriesI have been to Taiwan once or twice on business, also Hong Kong, but have not seen the mainland. I want to one day.03:47
wolfspraulso it's a total of almost 4 years in greater China now, phew03:48
rjeffriesI also have "visited" Vietnam, but not as a tourist if you get my drift03:48
wolfspraullike I said (and Werner guessed my 'black spots' very well) - another few years that should be it03:48
rjeffriesare there many ex=pats where you live now?03:49
wolfspraulI never understood this term. don't know. To me China is a country like many others.03:50
rjeffriespeople from europe or US who live in another country03:50
wolfspraulfree software scene is small here, but among those few it's a very close and supportive circle.03:50
wolfspraulwpwrak: I'm curious. Have you read about the Ubuntu/Banshee/GNOME donation/money dispute?03:55
wolfspraulI'm thinking about it but cannot come up with a clear judgment.03:55
wolfspraulhttp://www.networkworld.com/community/canonical-alters-banshee-agreement03:56
wolfspraulI still don't fully get it.03:56
rjeffrieswolfspraul I think Banshee is much ado about not very much. all's well that ends well03:56
wolfspraulI guess Ubuntu has some default settings so that when people purchase music at amazon, some percentage goes back to whoever. and now Canonical/Ubuntu wants more of that for itself than for GNOME.03:57
rjeffriesthat is old news they caved in today03:57
wolfspraulfine but why? who should 'cave in', and to what?03:58
rjeffriesthey==ubuntu03:58
wolfspraulUbuntu is less needy or worthy of money than GNOME?03:58
wolfspraulwho gets to decide that?03:58
rjeffriesubunto was on the wrong side of this in my opinion. a bit greedy I'd say03:58
wolfspraulis it a public perception game? whoever has more people to flood blog comments wins?03:58
wolfspraulwhy? because Ubuntu is less 'deserving' than GNOME?03:59
rjeffriesi don't know. tomorrow we will have a different story. that one is old news l(03:59
wpwrak(banshee) haven't heard of that yet. reading ...04:00
wpwraksounds like a negotiation meta-problem. it's always psychologically better to not give what you want for yourself than to first give and then take it away again. well, unless your objective is to establish a lousy reputation.04:04
wolfspraulwpwrak: you will like this one http://ndftz.com/nickelanddime.png04:05
wolfspraulubuntu decides what they install by default04:05
wolfspraulmozilla gets a lot of google money (not for the small Linux desktop share of course). what if Ubuntu makes a new firefox build that uses bing and canonical then pockets some money from Microsoft?04:06
wolfspraulmozilla would only get a little less money from google (if at all), and canonical might get a little from ms04:07
wpwraknice diagram ;-)04:07
wolfspraulis that good or bad now? for whom?04:07
wolfsprauldoes that make ubuntu 'greedy'? why?04:07
wolfspraulisn't mozilla even greedier for not wanting to share?04:07
wpwrakthe article also has the banshee revenue. about USD 2k. over what time frame ? a week ? a month ? a year ?04:08
wolfspraulwhat stops someone from making a ubuntu-derived distro that funnels money elsewhere?04:09
wpwrakah, july to december. so something like USD 400/month04:09
wpwraksounds like peanuts in any case ...04:10
wolfspraulmaybe like werner said, if ubuntu would just say 'we take all money from the power of our distribution/default settings', and stick to it, there would be no 'negotiation meta problem'04:10
wolfspraulif you don't like that, build your own distro (binaries), change all money feeds to yourself, and distribute it04:10
wpwraka derived distro would have less exposure, hence even smaller peanuts :)04:10
wolfspraulthen the different causes that are in need of money compete with each other openly in front of the users04:10
wolfspraulyes but there are many derived distros still04:11
wpwraki mean if they grow the 400 USD/month, then that may actually feed one person in the whole community full-time04:11
wolfspraulubuntu has an obligation to not take money? why?04:11
wpwrakyeha, i don't quite see the problem either04:11
wolfspraulso it's mostly the badly communicated/executed long-term position of ubuntu/canonical that leads to this04:12
wpwraksure, if they make tons of revenue and keep it all to themselves, that would create unrest04:12
wolfspraulthat's the case with mozilla for years and nobody cares04:12
wpwrakbut for petty cash ?04:12
wolfspraulthe money is piling up there, hundreds of millions04:12
wpwrakyeah, mozilla. entirely different story :)04:12
wolfspraulwell, first you need to have some idea of how it should be, irrelevant of the amoutn of money.04:13
wolfspraulmore money is good, right? it feeds people. the end user should have a say in whom he wants to support.04:13
wolfspraulif those clicks or default settings are so valuable, that's a great opportunity for the end user to support whatever cause he wants to support.04:13
wpwraki'd file this under "story of low relevance". manwhile, how's the hunt for mozilla's treasures going ? :)04:15
wolfspraulneed to ask Jon04:15
wolfspraulyes banshee is little money, I was just surprised about the confusion.04:16
wpwrak(banshee) i guess people hear "amazon" and then think entire cities made of gold :)04:17
wolfspraulfor hardware, I'm leaning towards staying out of any subsidization game. If I could, I would let people upload and pick the default settings/install they want when they order a product.04:17
wolfspraulso someone could make a nice distro, and put in some feeds back to him. then upload the whole image like a 'skin'.04:17
wpwrakwolfspraul: (subsidize) i guess it would depend on the situation. if the price is right and it's easy to turn off/remove, why not04:17
wolfspraulif someone buys a NanoNote and chooses that image/skin, it comes delivered like that, and if the end user uses the software whoever uploaded the image makes money.04:18
wolfspraulnot that I can do this technically today, but I think that might be interesting.04:18
wolfspraulyes, I'm totally not against money. creating value and monetizing is good, very needed to make things grow. I'm wondering about who decides about distribution of the money.04:19
wpwrak(situation) well, and nothing overly repulsive in other ways04:19
wolfspraulon the hardware side, we don't have this problem today04:19
wolfspraulour volumes are too low, and I'm not aware of any connection to any sort of 'pay feed' that anyone is proposing or wants to preinstall04:20
wolfspraulbut sooner or later (if we are successful), it will come up...04:20
wpwrak(decides) sharism would as a first step, so you :) if tuxbrain, pulster, etc., throw it away and do something completely different, that would be their choice04:20
wolfspraulyes04:20
rohthe pay-feeds like the ubuntu/gnome conflict showes are bad in my opinion (for opensource in general)04:21
wolfspraulroh: so how should it be?04:21
wpwrakmaybe pulster could make a deal for preinstalling a copy of guttenberg's thesis ;-)04:21
wolfspraulwhat is Debian's policy about this actually?04:21
rohopensource lives from pluralism and stuff like the 'whats the default search engine' and 'who is the default webstore' only hurt choice (which is the whole reason to use opensource)04:22
wolfspraulmy debian iceweasel has a Google box in the top right04:22
wolfspraulI don't think I installed it.04:22
wpwrakroh: defaults are good. as long as they fit user's needs and are reasonably easy to change04:22
wolfspraulso it's a default? Debian's or Mozilla's? who gets money for this default?04:22
rohwolfspraul: add all shops/search engines which are technical feasible/make sense from a technical pov. (means only bad/broken code is a reason to remove one)04:22
wpwrakroh: (fit user's needs) so if your default search engine is itunes, that probably wouldn't receive much applause04:24
rohgive the user information about all his choices, and how to select preferences. listen to the users about preferences, and dont do 'dicatory selections' (like ubuntu does)04:24
wolfspraulwhere is my Google box coming from? maybe Debian's policy is to not touch what upstream is doing? I remember some apps whose only purpose was to shop at amazon were rejected at some point.04:24
wpwrakroh: most people are probably happy with google04:24
rohwpwrak: ack. i use it too. i still have other ones installed and use them when it makes sense (like segor (local electronics dealer) or wikipedia, etc)04:25
wpwrakroh: what would suck is a hard-coded google search. worse yet, if it takes away a lot of screen real estate.04:25
wpwrakroh: of course, you could then always fork or "cleanse" :)04:25
rohto be fair.. i dont use mozilla browsers where i dont need to anymore. much too slow04:25
wpwrakroh: i still love my konqueror04:25
wpwrakalas, it's also getting slow04:26
wolfspraulI will try netsurf in ernest once it supports javascript :-)04:26
rohwpwrak: the whole reason to use ubuntu instead of debian for me is 'fresher versions, more testing, less crappy default configs'04:26
wpwrakroh: yeah. ubuntu does many things right. what i don't like it how persistently it tries to give you a "desktop"04:27
roheven when the last one is really getting worse the last few years. ubuntu needs to remove all the bullshit-politics from decisionmaking and get back to 'what the community wants' .. or they _will_ loose customers (me included)04:27
wpwrakroh: also things like plymouth, console-kit, .... argh04:27
wolfspraulseems we don't have a very clear idea about who/how to decide about distribution of money from whatever feed/source04:27
rohwpwrak: desktop is ok in some kind of way. there is no way around it for a workstation.04:27
wolfspraulhopefully the issue will grow (more money), then we can come up with something good :-)04:27
wpwrakwolfspraul: you can just decide and see what happens. as long as not too many envelopes with explosives slip through the chinese postal system, you're fine :)04:28
wolfspraulwe don't have this problem on the table now04:28
rohwhat i will not do is pay for my distro (ever) or use one of their music or videostores as long as i dont get proper quality and can buy without revealing lots of personal data.04:29
wolfspraulbut I think there is no clearly and fully accepted policy, or at least I'm not aware of one04:29
wpwrakroh: well, my desktop is centered around fvwm. other ideas are welcome, as long as they don't change that my desktop is centered around fvwm ;-)04:29
wolfspraulroh: ok, but that sounds like you are not too far away from actually considering to pay for content at some point ;-)04:29
wolfspraulit's not that you go as far as saying "100% of software and content must be free" and "ever paying even 1 cent for any digital good, click or service is a bad thing that I will never do"04:30
rohwolfspraul: well.. i am not a foe to licensing or so. i license my works in gpl as you know. (and help others to defend their gpl rights)04:30
rohwhat am i a foe of is paying the wrong guys money which they have not earned. (means most parts of the video/audio distribution chain)04:31
wolfspraulyes that makes sense04:31
wolfspraulso it's about making the flow of money transparent, and making sure people have control over where exactly their money goes04:31
rohwhat i am completely against is giving up any personal data to get content. that never was part of the deal and they have done more than enough to proove they are not worthy or able to care for such data. so in the end it will be sold/lost.04:32
wolfspraulok, I fully agree but anonymous payment systems is a separate problem (aside from me trying to understand about the money distribution side)04:33
wolfspraulmoney distribution I think I have some sort of idea now :-) thanks!04:33
rohgive me working anonymous micropayment and we have a completely different market situation ;) that would mean i would also 'spend' more money on 'makers' of art and culture more easily if i can just transfer a few euro if i like the goods04:33
rohwolfspraul: there are technical solutions for secure, anonymous micropayment. the problem is: lots of banking lawmakers prohibit it.04:34
wolfspraulcan you drop some names/hints?04:34
rohthey fear moneylaundering etc.04:34
wolfspraulthey should fear their own greed the most :-)04:35
wpwrakwolfspraul: it's probably also good not to reveal unnecessary financial details. otherwise, you just get endless arguments over trivial amounts.04:36
wpwrakwolfspraul: if you want total transparency, then a way of hiding things would be just by having so much data present that it's hard to find anything in there :)04:36
wolfspraulwell I'm always in favor of practical solutions. 'total transparency' sounds like the wrong priority.04:37
rohwolfspraul: i think some parts of it were patented, but these patents ran out some time ago. i dont have any names but when googling the name 'david chaum' pops up.. also it was something using 'blind signatures'04:37
wolfspraulnot because one wouldn't want it, but because these things need to be meaningful. if nobody uses data, there is no point in collecting it.04:37
wolfspraultransparency will create an entry point for bad priorities (=endless arguments over trivial amounts), so that requires some education for people to ignore such endless arguments in favor of more worthy things to do.04:38
wolfspraulI'm fine with that challenge.04:38
wpwrakwolfspraul: yeah. and opens you to annoyance problems04:38
rohhttp://osaka.law.miami.edu/~froomkin/articles/tcmay.htm04:38
wolfspraulroh: thanks, david chaum and blind signatures is a starting point, always curious to learn...04:39
wpwrakyup04:39
rohhttp://en.wikipedia.org/wiki/Anonymous_internet_banking04:39
rohbasically the technology is there for ~15-20 years.. i still hope i will see it implemented and used in real world in my lifetime ;)04:40
rjeffriessomething similar to this hack might(??) work between Ben and ASUS Wl 520GU04:41
rjeffrieshttp://todbot.com/blog/2010/12/16/wifi-for-arduino-with-asus-wl-520gu/04:41
wpwrakroh: 15-20 years is good. no patents ;-)04:42
rohrjeffries: dunno about 'that' hack. but basically 'every computer with linux, a wifi and a usb host interface' will work04:42
wolfspraulroh: I liked what you said the other day about you not wanting to fight for resources. made me think. (thinking not finished yet :-))04:42
rohwolfspraul: i think its a question if one 'gets' that ressources on this planet are limited or not ;)04:43
wolfspraulnah no problem, it's obvious that you wouldn't want a system where the strong abuse the weak until they are fed up.04:44
wpwrakas long as i get those i want i'm happy ... :)04:44
wolfspraulthat is only more clear to me every day living in China with a brutal culture of exploitation everywhere.04:44
wolfspraulso we gotta protect our happier way of life :-)04:45
rohand the social question 'in what kind of society' one wants to live. i dont like having massive differences in wealth in society.. they generate unrest (for real reasons, like basic human needs). from my pov capitalism failed (and is currently in denial about it) .. communism or socialism never got implemented (and better arent in any pure forms)04:45
wolfspraulok I will read about this anonymous payment stuff a little. good task for a Saturday...04:46
rohi think we as a society still need to find a form to organize better than we do now. and i dont mean that in a revolutionary context. thinking about 'better solutions' is the easy part. finding any way to get something as complex as society there.. the migration.. thats hard.04:47
wolfsprauland I sort of have an idea for the problem of distributing money feeds in free software/content04:47
rohi think sw devels need to be paid from the same pots as hw devels. nobody needs to get fsckingly rich. its on every one himself to find a level of 'enough'04:47
wolfspraulI don't care whether we are only talking about a few cents or dollars at the beginning, but the system must be well explained, understood and accepted, otherwise it can never grow.04:48
rohmy concept of open hw and sw is 'sell hw, pay devels and production from it'04:48
rohdo the 'better' product. make it last longger, be sturdier etc. i want back to a world without built in obsoletion.04:49
rohhttp://www.youtube.com/watch?feature=player_embedded&v=tI798T2tRrQ04:50
roha very good video about it. currently we live in a world which throws everything away very fast. all the time. i want stuff to be reused, recylced, resold?, upgraded? etc.04:51
wolfspraulyes, hardware can be much more upgradable04:53
wolfspraulone of my uber-priorities04:53
wolfspraulwow that video is 1.15h04:53
wolfspraul:-)04:53
wolfspraulin the next few years we will see a reduction of product cycles, especially on the phone/smartphone side04:54
wolfspraulpeople will throw away their phone every few months, maybe even weeks04:54
wolfspraulif the phone costs 20 USD there comes the point when you stop thinking about it, because every parking fee, dinner, not to mention other services in your life like medical stuff, are more04:55
rohand we CAN do it. just check how long a airplane lasts. >30 years. one changes spareparts which have wear. one replaces avionics every generation of technology. the german army has hercules tranport planes (lockheed C130) still in use. the first of that type were built in the 50s (of last century!)04:55
rohwolfspraul: from my pov we can't get out of the 'faster and faster' garbage cycle without billing the 'real cost' for such products.04:56
wolfspraulyes but if the phone hardware costs 20 USD it is very easily subsidized, i.e. you throw away the phone because a new 'free one' has just arrived in your mailbox :-)04:56
wolfspraulmuch like the AOL CDs...04:56
rohmeans we need laws to make products _really expensive_ when they are throwaways.04:56
rohin the end it would be ok to buy pricy products if the last multiple times as long. in the end its a question of reason and mathematics. we just dont have the ressources on this planet to continue as before.04:57
wolfspraulI don't see such laws emerging. phones will get cheaper, and be thrown away faster.04:58
wolfspraulmaybe they will even arrive unasked for in your mail, really like the 'get online' CDs 15 years ago04:58
rohwolfspraul: heh. not yet. but more and more people get that argument.04:58
rohthink of the 'stoffbeutel'-bewegung which counteracted the plastic bag culture. that battle is still on. yet some small islands already completely oneway banned plastic bags due to their amount in the garbage04:59
wolfspraulI'm not sure. the phone is 20 USD because the chips are really simple and are not / should not be expensive.05:00
rohislands somewhere in the caribic i think (where they still have no space left for bigger garbage piles)05:00
wolfspraulit's not because the environment is ruined or kids are used as work slaves.05:00
wpwrakwolfspraul: (financing) i think you need to adapt as you go. at the beginning, it will have to be unfair to be useful. the more money you have to throw around, the more fairness you can afford.05:00
wolfspraulthe reason the prices are coming down is because of inredible advances on the IC side.05:00
wolfspraulreal advances, no downside05:00
rohwolfspraul: nah. i think prices only go down due to the 'not invented here'. everybody is just copying05:01
wolfspraulthe Samsung fabs in South Korea are working by very high environment and worker salary standards.05:01
rohand copies are cheaper than innovation ;)05:01
wpwrakwolfspraul: (phone life cycle) don't forget the investment users make in terms of learning, customizing, etc.05:01
wpwrakwolfspraul: you can even increase the lifespan by making them want to invest more of their resources (monetary and other)05:02
wolfspraulyes! totally. that's how I think. the continuity is on the knowledge side.05:02
wolfspraul(and the value)05:02
rohi think some stuff just 'wears down' .. phones also. but thats after 3-4 years. not after 18month.05:03
wolfspraulbut the price of the hardware and subsidization possibilities, together with reduction in life span, are a reality that we need to make use of, not fight against it.05:03
rohatleast then you need major service05:03
rohwolfspraul: i dont and will never try to attack the 'low end' segment of stuff.05:03
wpwrakroh: in argentina, we have really thin plastic bags. not the "built to last" monstrosities you get in europe ;-) and they double as garbage bags. nice solution.05:03
wolfspraulroh: yes, I agree. but if a whole phone (the hardware) costs 20 USD, that's a significant thing/development, we cannot stop it (wouldn't even know why), we need to think about it and make the best of it.05:03
rohwpwrak: nah. you get the thin ones here also. every night-shop has them. the cool ones give you a plastic woven fabric one which you can reuse a looong time.05:04
wolfspraulof course electronics should last 10, 20 or more years. I'm talking about something else. What does it mean, and what will happen, if the phone hardware costs 20 USD.05:04
wolfspraulbecause that's what's going to happen for sure.05:04
rohwpwrak: similar to a textile bag. some kind of plastic based textile. surely not easy to recycle. but atleast it lives as long as a textile bag would (mine is 2 years old now)05:05
rohwolfspraul: phones already cost only 20E05:05
wolfspraulyes, and more will come down. right now many people still think it's a 'few hundred usd' something.05:06
rohwolfspraul: you get nokia, motorola, samsung and more for 20E .. usually with a prepaid sim.05:06
wolfspraulbut what if they realize it's more like. I don't know. like a one-time parking fee in the city?05:06
wolfspraulwhat will happen?05:06
rohwolfspraul: proper phones still cost a few hundred. and that one will not get down. it didnt in the last 10 years.05:06
wolfspraulyes because the industry works hard to keep prices up.05:07
rohthey just get more features and thus more cost while production costs go down. a smartphone with 300-700E basically opened up the former upper limit (before there were smartphones)05:07
rohi remember paying nearly 300 or 400DM for my siemens S25 (when it was new and fresh.)05:08
wolfspraulI think when the price of hardware comes down, what kind of software you are familiar with becomes more important, and equally the question of money feeds becomes more important (aka 'subsidization')05:09
wpwrakyes, software will become the common denominator people will try to find. also combined with services.05:09
rohsubsidization only works on stupid people bad at math ;)05:10
wpwrakandroid is on the right path there05:10
rohwpwrak: right path for market domination yes.05:10
rohright path not to clutter it themselves.. no. not at all.05:10
wpwrakroh: (stupid people) who said the smart ones are given a choice ? :)05:10
wolfspraulroh: no subsidization work everywhere. you just said even yourself wouldn't mind paying for video, if you could do so anonymously, and if you knew the money goes to the right people.05:10
rohin the end you are quite fscked with an android 1.x or 2.1 phone.. new versions coming out fast and on most devices there is no upgrade path at all.05:11
wolfspraulsubsidization just means you don't pay for the physical hardware goods, but for some ease-of-use service someone put onto that hardware for you, to make it work in ways that you find helpful in your life.05:11
rohwolfspraul: subsidization on videos? nope.05:11
rohthats paying money if i like it. like giving it to street musicians when they are good.05:11
wpwrakroh: android still needs to stabilize. but give them time ...05:11
rohor buying a cd if you are at a concert and the band is good.05:12
wolfspraulok when I say 'subsidization' I mean that you don't pay for the physical goods in front of you. instead the money is made back because that piece of hardware connects you to some service or value YOU LIKE. like the video example you gave.05:12
rohwpwrak: thats not their issue. its their broken old-world development model combined with stupid management. google isnt a  really intelligent company at all levels ;)05:12
rohwolfspraul: nah. subsidization is 'hiding the real cost of things' from my pov05:13
rohatleast thats what carriers in the phone market try. and the dsl market (selling expensive routers or asking for 2 euro a month)05:14
wolfspraulmaybe because the types of subsidization on the market don't match causes/things you like.05:14
wolfspraulif subsidization makes the money back on something you would want to do anyway, you will very much like the whole package (the free/subsidized beginning, plus the paid service)05:15
rohwolfspraul: the point is: its fraud on the customer. he wouldnt have bought it for the same money if told before plainly.05:16
wolfspraulnot sure about that. it's a credit scheme.05:16
wolfspraulpeople buy furniture on credit.05:16
rohthats what i think is cause for a lot of 'badly working business cases' even when the product isnt that bad05:16
wolfspraulI'm not trying to talk you into subsidized anything :-)05:17
rohselling too aggressively is a sure thing to make me NOT buy it for example.05:17
wolfspraulI just say if phone (or other mobile electronics hardware) are coming down even more in price, those questions will become bigger.05:17
rohstupid people believe in credit. yes. i see that as one of the great wrongs which are told people in school.05:17
rohcredit is what keeps people poor in the end. its a good way to controll them when you are rich and want to keep your earning distance05:18
rohi think 'credit is good' is one of the biggest lies of the capitalism. its the only things which makes it work (for some) at all.05:19
wpwrakroh: do your living arrangements involve renting ? (or do you just live at ccc anyway :)05:20
wpwrakroh: credit can be useful. but it's dangerous.05:20
wolfspraulif Mozilla can give away a free phone, with 100% free software on it, because it can make back the money from Google, why is that a bad thing?05:20
rohwithout it a lot of 'people not helping society get further' wouldnt have jobs at all.05:20
rohwolfspraul: because money doesnt get made from nothing. there is ALWAYS a price. and hiding it doesnt make it go away.05:20
rohgoogle is a big octopus which lives from peoples personal data. its NOT my friend.05:21
wolfspraulyou can reflash that phone with another distro05:21
rohwolfspraul: still doesnt make the intention to fraud people out of their personal data go away. i dont like such businesscases and will continue to fight them.05:22
rohe.g. i bought a xbox (the first one back then) knowing its subsidarized by games and no games. never used it for playing.05:23
rohwouldnt do it again.. nvidia hw sucks ;)05:23
wolfspraulso you would not touch that free Mozilla phone. you would throw it away and educate others to do the same?05:24
rohused it some time to watch 'ip tv' ;) .. back when sdtv was enough (and 700mhz celeron)05:24
wolfspraulactually maybe you would contribute even more to 'throw away' then :-)05:24
rohwolfspraul: i wouldn only get one when being able to do so anonymously and repurpose the hardware.05:24
wolfsprauljust saying... you didn't ask for it, but when those things get really cheap (and connected to valuable services), they will pop up EVERYWHERE05:25
wolfspraulok05:25
wolfspraulthat makes sense05:25
rohand would help others to do the same. would make the thing for mozilla mostly a costly experience.05:25
wolfspraulI doubt that.05:25
wolfspraulsubsidization schemes don't work if they are truly against the interests of the end user05:25
wolfspraulas werner said above - he doesn't actually 'mind' the google links05:25
rohmy point is: personal data is NOT anything which may be traded. its unethical and (sometimes) illegal. its never in the interrest of the customer.05:25
wolfspraulyes sure, but other well educated people may not agree with you.05:26
wolfspraulthey may actually like those services05:26
rohthus: if you want to be a credible as a company. dont do it or people like me will kick your ass and nag you all day long ;)05:26
wolfsprauland that means, together with a reduced cost of hardware, that subsidized (free) hardware will show up05:26
rohwolfspraul: i dont see people as well educated if they entrust such companies with their data ;) atleast not in information technologies05:27
wolfspraulthe only thing the supplier of such subsidized hardware needs to do is to avoid gross waste, like someone trying to get 10,000 free Mozilla phones to grind them down to recycled park benches05:27
rohmy point is: naming something subsidized as free is fraud.05:27
rohits plain and clean advertising fraud.05:28
wolfspraulso if he can direct the subsidized goods with some real-life mechanisms towards actual users, it could work05:28
wolfspraulbecause (let's say, in this example) Google sees enough traffic from the phones to give Mozilla enough money to make more of them05:28
rohi dont know where that idea comes from.. but i dont believe mozilla is able to do phones anytime soon.05:29
rohand i wouldnt work for them. wrong mentality there. technical and ethical.05:29
wolfspraulI just created the idea for fun, to think about.05:30
rohmozilla seems to be on a good way to destruction i think. but thats mostly grounded on their stupidity when it comes to technical details.05:30
rohbut its quite clear when it comes to 'continuity'.. its still netscape ;)05:31
wpwrakroh: at the end of the day, nothing is "free". someone always pays.05:33
rohexactly. if you ask me. be truethful. earnest. dont advertise stuff you cant to.05:36
roheh do05:36
rohbtw.. warranty laws are something which came after planned obsolescence05:36
wolfspraulwpwrak: so you say Mozilla should stop taking google money, and instead sell a proprietary (pro) version of the browser to make the money their developers need each month?05:37
wolfspraulin order to be truthful about who pays?05:37
rohyes. never lie to your customer.05:38
rohi think mozilla needs to find a proper way to make money.05:38
rohlike work with vendors similar to like opera does. get paid for integration work. for special stuff.05:38
wpwrakwolfspraul: "i" say ? i think you mean roh ;-)05:39
wolfspraulwpwrak: no you said "nothing is free. someone always pays"05:39
wolfspraulthat reminded me that Mozilla Firefox is indeed not free.05:39
wolfspraulGoogle pays every month.05:39
wpwrakwolfspraul: i think they should take google money. why not :)05:39
wolfspraulit sounded like you wanted that to stop and the end users to realize that Firefox is not free.05:39
rohbtw.. the new macbooks seem to have not user replaceable batteries... we'll see how that plays out.05:40
rjeffriesa $20 or$50 phone, no subsidy, is very close to reakity. maybe this year as a less capable Android phone with a not huge display wasily $75 USD no contract05:40
wpwrakwolfspraul: no, not at all. what i meant was that "totally free" is an illusion.05:40
rjeffriesroh the iPhone and iPad have alreday proven non user replaceable battery is an OK idea05:40
rohrjeffries: nah. it made lots of people angry and even sue apple.05:41
wolfspraulah ok, but unlike roh you think it's ok if the money is not made from the end user always, but can be made somewhere 'behind' in the system.05:41
wpwrakwolfspraul: the question is just what sacrifices (your own and those of others) you accept05:41
rohrjeffries: it also made sure i never buy one. screws are ok. no spareparts isnt.05:41
rjeffriesyou get a thinner lighter product, and every two years you send it to a shop and they replace teh batteries05:41
wolfspraulthe trend will go towards non-user-replacable batteries05:41
wpwrakwolfspraul: yes, of course. i actually think the direct money from end users approach is very difficult and may distract from the real objectives05:41
wolfspraulit goes well together with shorter life cycles, and better recycling05:42
rjeffriesroh you are indeed a special case. tens of millions of people have voted with thier hard earned money05:42
rohwolfspraul: only for some parts of the market. check nokia. they made much more money in the history because people could get fresh batteries everywhere.05:42
wolfspraulyes sure, but things are changing05:42
wolfspraullet's start with the thickness05:42
rohin some parts of the world an apple product is useless. you cant charge it. and you cant exchange the battery05:42
wolfspraulend user replacable batteries need more casing = thicker05:42
rjeffriesroh but that was then batteries are much improved and many people love VERY skinny phones05:43
wpwrakwolfspraul: (batteries) i like nokia there. if nokia die, ubiquitous small batteries will die with them05:43
wolfspraulthe next thing is product life span05:43
rohwolfspraul: as i said.. screws are ok. but unreplaceable isnt.05:43
rjeffriesNokia will not die,05:43
wpwrak(ubiquitous and with a long design life)05:43
wolfsprauluser replacable batteries make a lot of sense if a large percentage of users uses the product for let's say more than 2 years05:43
wpwrakrjeffries: or transform beyond recognition :)05:43
rohnokia fucked up big time with the windows decicion. i know lots of powerusers which i think will not switch like that idiot believes.05:44
wolfspraulhowever, if the product life span falls to 6 months or 'mostly' below 12 months anyway, there is no point in the user replacing the battery.05:44
rjeffriesbattery does not fail in 12 months05:44
wolfspraulon the recycling side, with fixed batteries you actually improve the recycling situation, because you create an incentive for people to go back to the store05:44
rjeffriesand apple has been smart the charge to replace battery is not bad at all05:44
wolfspraulour beloved Swiss Telecom did a lot of tests and study on that05:44
rjeffrieslook I like replaceable batteries05:44
wolfsprauland they clearly found that non-user-replacable batteries in real-life are more eco-friendly.05:44
rohwolfspraul: replaceable batteries is two-fold: one thing is replacing a defective battery, another is switching to a secondary one.05:45
rjeffriesbut EVERYTHING is a design tradeoff05:45
rohwolfspraul: the first one can be solved by spareparts for a fair money and being able to exchange it with 'some work' (may include screws)05:45
wolfspraulrjeffries: yes correct, the charge is genius actually.05:45
wpwrakwolfspraul: (recycling) because people will throw away a battery but bring a phone for recycling ?05:45
rohthe second one cannot be solved by screws. its about being able to use it when the battery is 'just empty'05:45
wolfspraulit improves the feedback they get on their manufacturing quality, it feeds the production machine with higher quantity (=lower per unit cost). It allows for an upsell and customer service opportunity. and so on.05:45
wolfsprauland it's also more eco-friendly, as studies by unrelated parties such as Swiss Telecom have shown.05:46
rjeffriesroh that is why they let you charge from USB;)05:46
wolfspraulwpwrak: yes, I think something like that. I only remember the bottom line they told me.05:46
rjeffriesbut this is a boring topic/;)05:46
rohrjeffries: apple doesnt. (proprietary connectors)05:46
rjeffriesI do not use or buy apple05:47
wpwraki don't like non-replaceable batteries, although i realize than most are de facto non-replaceable (because you can't get a replacement when they eventually die)05:47
rohrjeffries: also you need to be careful with apple. they are very much 'overpresent' in visibility due to their 'buzz'05:47
wpwraki wouldn't mind non-replaceable batteries in a phone if the platform was long-term stable05:47
rjeffriesthe real tragedy is everyonbe now thinks they must have an expenisve data plan to be a complete human being.05:47
rohin the real world of computing or phones in general.. they are only one of many.05:47
rjeffriesTHAT my friends is the big fraud. ;)05:48
rohrjeffries: carriers have to get that they are 'bit carriers' ;)05:48
rohelse people will tunnel/etc to use data, regardless of what sold05:48
rjeffriesI am sure the CIA and FBI and other security agencies are DELIGHTED that people check in on theior mobile phone when they go to a coffee shop05:49
rohive used a lot of voip myself. through openvpn via umts (when dsl was down till installed after moving)05:49
rjeffriesone good argument for removable battery roh did not mention:05:49
wpwrakrjeffries: (data plan) that's why i make a point of having phones with extremely poor or no data capabilities ;-)05:49
rjeffriesit is the only way to disable your phone revealing where you are in geo space05:50
wpwrakrjeffries: i only use data-over-phone in emergencies when my isp is down05:50
rjeffrieswpwrak i do not have a data plan at all05:50
wolfspraulrjeffries: watch out there may be secondary batteries in your thingie.05:50
rohrjeffries: that too. yes. ;)05:50
rjeffriesand I keep using a many years old Pal Treo 700p that looks like hell but Just Works05:50
rohrjeffries: but only because there is no proper 'mechanical switch'05:50
wpwrakrjeffries: (privacy) an open phone design and a reliable off switch would do that as well. we went to some length at openmoko to ensure that (although we didn't have the gsm side open)05:51
wpwrakroh: that's so 19th century :)05:51
rjeffriessp we all agree it would seem05:51
rohrjeffries: i have a new (used) razr v3i now. my second one after i had a S25. its a phone. ;)05:51
wolfspraulthe privacy should be guaranteed through the openess of the design and software, not by hoping that removing the battery will do it.05:52
wolfspraulit also renders the phone useless, as a minor side-effect :-)05:52
rjeffriessome paranoid people (??) in USA think Google may be infiltrated or anextensu=ion of out intelklugence services CIA etc05:52
rohrjeffries: its a us company so you can be sure their 'three letter services' will do what they want.05:52
rjeffriesI guess I need a lead case to slip my phone in while the battery is out.)05:53
rohthe only way to make sure they don't is: dont give your data to a us company or someone dealing with em'05:53
roh'host in europe' ,)05:53
rjeffriesindeed and they do and never EVER use any lubrication either05:53
wpwrakroh: as if data in europe wouldn't be shared with the good friends overseas ;-)05:53
Action: rjeffries nods to wpwrak they are all in this game together05:54
rjeffriesbut when I move to Chad or Niger, things will be different.05:54
wpwrakroh: if you really want to be sure about who is spying on you, you'd probably have to host in north korea :)05:54
wpwrakrjeffries: you can probably lead a pretty comfty life there with your dollars. spend them while they're still worth something ;-)05:55
wpwrakrjeffries: of course, the PIGS crisis in europa has bought the dollar something like a 5-10 years life extension :)05:56
rjeffriesthe world economy is in such a ditch05:57
rohwpwrak: wait till the chinese are done exchanging their dollar reserves to euro or so ;)05:57
rjeffriesChina and Saudis are our banker. what could possibly go wrong? ;)05:58
rjeffriesThank God China mainly uses bicycles, right?05:58
rohwpwrak: when the chinese stopped buying them to compentate the trade deficit, the fed in the us started buying them with money from nothing. its a us self-made recession.05:58
wpwrakroh: yeah. the clock is ticking05:58
wolfspraulhey, it almost feels off-topic here, but has anyone noticed how good our latest NanoNote OpenWrt release is? 02-2305:59
rjeffrieswe are so fscked05:59
wolfspraulit's the first time I cannot immediately find lots of details that are crashing or unusable05:59
wolfspraulit slowly feels like a whole little computer actually, of course still with endless usability issues05:59
wpwrakwolfspraul: then you must not be trying hard enough ;-)05:59
wolfspraulthere are a number of little games, interesting and powerful math apps, some attempts at dictionary, music player, etc.06:00
wolfsprauleven nice little things like metronome or ascii art06:00
rjeffriesI did go look at instructions for reflashing and realized that wolfspraul is right (again)06:00
wolfspraulof course I can still see this improving 10 times over, but as a first step 02-23 is actually somewhere, I feel.06:01
rjeffriesBen is not a mass market product. built by engineers for software geeks who enjoy challenges06:01
wolfspraulgreat work everybody!06:01
wpwrakrjeffries: so you reflashed successfully ?06:01
rjeffrieswpwrak oh no not at all. when I figured out more or less what I need to learn I put Ben back in charge battery mode06:02
rjeffriesneed to keep battery charged, just in case06:02
wpwrakrjeffries: it's basically just power up while holding the U button, then running a script :)06:03
wolfspraullet me see what I think needs to improve next:06:03
rjeffriesbut first I need to undersatdn the flash partition layout and blocks and all that and whatever06:03
rjeffriesit is not hard stuff but is a lot of stuff06:04
wolfsprauleasier upgrading, larger rootfs without increasing bootup time, jlime dual booting, document viewer - pdf/info, offline wiki06:04
wolfsprauleasier way to download music, maps06:04
wolfspraulupstream u-boot06:04
rjeffrieshow simple did OpenMoko eventually ake teh s.w upgrade process?06:04
wolfspraulusb gadgets, i.e. storage, keyboard, others?06:05
wolfspraulintegrate software for ubb, ben-wpan, hoperf module06:05
wolfspraulben-wpan stack06:05
wolfspraulthe 'easier upgrading' should also be possible from a fat-formatted memory card, and other ways like xbboot, not just opkg upgrades06:06
wolfspraulyeah, that's about it I think :-)06:07
wolfspraulthen we need a real test plan, better documentation both in the wiki and on the device06:07
wolfspraulif we have a good offline wiki reader we could even just have our own qi wiki on the device, we'll see what works well06:08
wolfsprauldid I forget anything? does anybody feel those priorities are wrong?06:09
wolfspraulah, I also like voip, and text2speech06:09
wolfsprauland better dictionaries06:09
wolfspraulmore love games06:11
wolfspraulrjeffries: Werner may remember the details of Om upgrade paths better than me, it feels like a long time ago for me now, I start to forget the details or what I remember is mostly wishful thinking.06:12
wolfspraulI don't think back then we had a strong focus on making upgrades particularly easy.06:12
wolfspraulI think by now upgrading the Ben is easier than upgrading a freerunner ever was.06:12
wolfspraulI still want to upstream the Ben u-boot patches, for example.06:12
wolfsprauland offer more ways to upgrade via fat-formatted memory cards, and other ways.06:13
wolfspraulbut it's already getting quite good (on the Ben). the foundations are there.06:13
wpwrakupgrading the openmoko phones was easy if you stayed with the images: just dfu-util the new kernel/rootfs and you're done. that is, unless you need an u-boot/u-boot enviroment change, then you're in hell :)06:13
wolfspraulyeah, that was bad. and booting from sd was also bad I think.06:14
wolfspraulbut you could reflash from Windows and Mac, at least at some point and to some degree of user-friendliess.06:15
wpwrakas long as you could use dfu, which was most of the time, it was pretty good06:17
wpwrak"human bean" ;-))06:19
sistem_error:-D06:19
wpwrakfwiw, atusb will also feature DFU capabilities for firmware upgrades :) the protocol is actually amazingly simple. not sure why it was such a protracted battle to get this into u-boot ...06:25
wpwrakwell, the mad partitioning scheme of the openmoko phones certainly helped to complexify things there06:25
rohwell.. sdboot on the mokos was bad because it couldnt do it itself.06:44
rohalways needed the bootloader for that06:45
kyakkristianpaul: (w3m image support) it's gdk-pixbuf drawing in fb07:45
wolfspraulroh: from a quick glance over some of those anonymous payment services it seems bitcoin may be the easiest to implement07:50
wolfspraulhttp://en.wikipedia.org/wiki/Bitcoin07:50
wolfspraulthe EFF accepts bitcoin donations, it says there07:50
wolfspraulI will see what it takes to bring bitcoin to the NanoNote, or to accept bitcoins when ordering a Ben or Milkymist One :-)07:50
kyakwolfspraul: may i ask you a question. Do you still have some Spectec SDW-823 left at your warehouse that you can offer with Ben?07:52
Action: wpwrak really needs to get those atben/atusb to adam07:52
wolfspraulkyak: I don't think so.07:53
wolfspraulkyak: btw - thank you so much for all your incredible help with our software - the result in the form of the 02-23 image is really amazing!07:53
wolfspraulI feel so good about our priorities now :-)07:54
wolfspraulwhat we have achieved in the barely 12 months since the Ben is shipping is really amazing, considering what a small group we are07:54
wolfsprauland I think it's not a one-trick pony, I have a feeling what we achieved so far is quite maintainable. in other words nothing will stop further software improvements, without regressions.07:55
kyakyea, Ben is becoming much better from the swoftware side07:55
kyaki will even be able to show it to my friends soon :)07:55
wolfspraulhe07:56
wolfspraulwhen I was playing with my Ben last night I was almost shocked07:56
wolfspraulI suddenly had this "wow, this is really starting to work" feeling :-)07:56
wpwrakkyak: quick, break something ! ;-)07:56
wolfspraulthere are still endless details in usability, of course. but that we got it this far is quite amazing, I think.07:57
wolfspraulI think the help screen in zvg was garbled - unreadable on my ben (strange artefacts)07:57
kyakmaybe we can start using the issue tracker more actively07:58
wolfspraulI did see little things like that here and there.07:58
kyakright now it's mostly reporting in mainling lists and irc07:58
wolfspraulyes. issue tracker is neglected.07:58
kyaka really great usability issue, already mentioned in emails, is separation of gui and console apps07:59
wolfspraulkyak: can I do anything to make Milkymist One attractive to you?07:59
wolfspraul:-)07:59
wolfspraulwe need some more contributors on that big challenge...07:59
kyak(separation in gmenu2x, so gui apps could be distincted from gui apps)08:00
wolfspraulkyak: [console apps] for sure. that's quite annoying.08:00
kyakwolfspraul: i just don't have an understanding about possible use case of MM for me! It could be used to hack and learn FPGA, maybe08:01
kyakthere was a nice proposal from jirka to mark console apps with a small "C" letter in the corner of the icon. Or maybe put them in a separate gmenu2x tab?08:02
rohwolfspraul: anonymous payment seems not help so much as long as you sell physical stuff ;)08:03
rohso.. i wouldnt invest too much work at implementing it for ben08:04
wpwrakroh: you could specify a geographic drop-off location. "in the park, behind the bush left of the 3rd bench"08:29
wpwrak"in a brown unmarked bag"08:29
wolfspraulroh: why not? physical stuff is paid with anonymous payment (cash) all the time.08:30
wolfspraulto me these are all just payment systems08:30
wolfspraulI'm not saying I would stop accepting paypal, or visa/mastercard08:31
wolfspraulbut you know what - I've recently had a Ben NanoNote customer who insisted on anonymous payment. So he got himself a prepaid one-time credit card, which seem to be available in Australia.08:31
wolfspraulthen he paid with that credit card08:31
wolfspraulhe paid for the prepaid one-time cc with cash08:31
wolfspraulthe Ben was delivered to some drop-off location08:31
wolfspraulall fine by me08:31
wolfspraulI'm very interested in alternative payment systems, so customers have more choice in which way they want to pay. it needs to be efficient though.08:32
wolfspraulI've sold Bens for cash, right Werner? :-)08:32
wpwrakthe argentine way, yes ;-)08:33
wolfspraulkyak: yes, I think the fpga opens up new possibilities, a real frontier for free software too08:33
wolfspraulright now the software is still bleeding edge everywhere, but I think/hope like with the Ben, every month it will improve in all areas08:34
wolfspraulrtems, gcc, uclibc, qemu, flickernoise, linux, openwrt, etc.08:34
wolfspraulyou could start on the qemu path, no need to buy anything08:35
wpwraki don't quite believe in qemu for a first contact. lacks the visceral aspect.08:36
wpwrakqemu is fine for optimizing whatever you do with the real device, though.08:36
wolfspraulfair enough. yes, you need to have very strong visionary drive to actually enjoy qemu as the first step into something new.08:36
wolfspraulfor a number of reasons I believe qemu will be more important in conjunction with Milkymist One than I think it is for Ben NanoNote though08:37
wolfspraulwhether it's the first step into Milkymist land, or some time later, is another question08:37
wpwraknot sure. part of the joy of mm1 would be the fpga. does qemu emulate that ? ;-)08:38
wpwraki view these emulators are something very technical. a tool for a complex problem. e.g., regression tests or cross-platform development. but not really as something that substitutes the actual device. particularly not if the device is a fully featured computer.08:40
wpwrak(as opposed to, say, a palm)08:40
wolfspraulkyak: I've just found the 2 spectec sdw-823 cards I still had...08:42
rohhm. i think qemu would make sense for the nanonote. (to do automated regression tests of build for example) .. but for the mm? what to emulate? the emulated cpu?08:42
wolfspraulthe problem with both is that some small part of the plastic came off, I think that is a problem/defect with all of them08:42
larsci totally agree with wpwrak regrading qemu08:42
wolfspraulI don't know whether they still work or not - the contacts also look a little 'rusty' maybe/08:42
wolfspraul?08:43
wolfspraullarsc: qemy not useful in conjunction with m1?08:43
wolfspraulqemu08:43
larscwolfspraul: it is08:43
larscbut imo it is quite boring if you don't have the real hw08:43
wolfspraulah yes, that I agree with too08:44
wolfspraul:-)08:44
larscroh: the development cycles are faster with qemu. i used it quite extensively when I was tracking down bugs in the lm32 kernel08:45
wolfspraullarsc: have you noticed any problem with your jtag-serial board? I mean slowness... we have found the cause and fix for the high-speed bug.08:46
wolfspraulso one day I can give you a fixed one. if there is no rush, we can do that at 28c3 :-)08:46
rohlarsc: true. i am just questioning if and how far it makes sense emulating the rest of the soc.08:46
wolfspraulif you see any slowness or problem, we can expedite it08:46
larscwolfspraul: I think I can apply the fix myself08:47
larsc(when i have the need for high-speed transfers)08:47
wolfspraulah ok. so you know the pics already I guess... http://en.qi-hardware.com/wiki/JTAG/Serial_Cable_run_1_for_Milkymist_One#Bugs,_Steps_of_Rework_and_Improvements08:48
larscyes08:48
rohmy reason is: the soc is(?) in flux, so qemu needs to follow soc development. thats a lot of work.08:48
wolfspraulif you don't get to it (or don't need it), I will exchange it for you next time we meet somewhere08:48
rohi can also do the rework in berlin if needed.08:48
wpwrakwolfspraul: btw, do you already know how much of a real-life difference it makes ? :)08:49
wolfspraulI didn't get answers, I think it makes a small or no difference at all.08:50
wolfspraulit's more that me and Adam don't like to have this kind of defective hardware our there08:50
wpwrakphantom pain ;-)08:50
wolfspraulthe ftdi ic has no voltage reference, etc.08:50
wolfspraulnot sure08:50
wolfspraulAdam likes it even less than me, I tend to be careful overriding the people that understand things better than me.08:50
wpwrakdoes he think it could have other effects ?08:51
wolfspraulThe ground pads of C3, R14, C27 and C28, which their grounds are all connected together to a locally same net and which doesn't connect globally to system ground. So that functionalities of these four parts become invalid:08:51
wolfspraula) C3, which is for the 12MHz Xtal purpose(Oscillator output), although I have never met any one time fail on 12MHz pulse during 97pcs tests. It'll still have a potential failure that will be happened on end user one day.08:51
wolfspraulb) R14, for FDTI chip current reference.08:51
wolfspraulc) C27, C28; for USB VPHY power supply noise reduction.08:51
wolfspraul---08:52
wpwrakah, i see. affects quite a lot then08:52
wolfspraulto me it's simple - it's broken and needs to be fixed. but I don't want to cause more damage with an overly dramatic and expensive 'recall' that will hurt people who are happily using their boards right now even more.08:52
wolfspraulso I offer to exchange/fix this at any time in the future. one by one we get the bad ones replaced with good ones.08:53
wolfspraulthe good news is that not many are out now08:53
wpwrakwolfspraul: all fine, but i think your message should be more specific: it's a complex breakage that currently manifests itself in high-speed not working (only full-speed), but that could conceivably also make the board fail completely under some circumstances.08:54
wolfspraulI'm not sure whether 'could fail under some circumstances' adds much in specifity.08:55
wpwrak"it'll be faster but not really. so how can we get yours fixed ? no pressure. asap ?" sounds a bit incoherent ;-)08:55
wolfspraulmy message is: I will replace the boards, sooner or later. :-)08:55
wpwrakit adds that the scope of the effects is not known and doesn't have to be limited to high-speed not working08:56
wpwraka) it may make people more interested in a fix/replacement, and b) it warns them that, if their new laptop/hub/whatever can't talk properly to the jtag board, their new laptop/hib/whatever may not be the one to blame08:57
wolfspraulwe should have a set of recall/fix terminology like in the aviation industry08:58
wpwrak"controlled flight into terrain" ? ;-)08:59
wolfspraulwith defects being rated by known (or unknown) seriousness08:59
wolfspraulno, to make communication this kind of fix/defect/improvement more consistent and easier to understand08:59
wpwrakplain english may work faster than teaching everyone a new terminology ...09:00
wpwrak.. that is, unless you plan to produce such issues in vast quantities ;-)09:00
wpwrakwolfspraul: at least you won't have the legal department to worry about that produced this beautiful description in very codified language: http://www.xilinx.com/support/answers/22471.htm09:03
wpwrakwolfspraul: (i'm still not quite sure what it really says :)09:05
wolfspraulwhen I insert my spectec-823, I get a jz4747-mmc.0: sparse irq: 80, about every 1.4 seconds09:12
wolfspraulafter that another 4-5 lines from ks7010_sdio.c:ks_sdio_interrupt [573,639,647,651]09:13
wolfspraulI meant "jz4740-mmc.0"09:13
wolfspraulthat's with the 02-23 image09:14
larsci guess we should get rid of that warning. the jz4740 mmc core seems to have problems with sdio irqs. but we can handle that in sw09:24
larscactually that warning isn't in the current driver anymore09:29
wolfspraulhow about the 4-5 lines from ks7010_sdio.c - will they go away as well then?09:39
wolfspraulon the console, the Ben is hardly usable with all that stuff scrolling by every 1.4 seconds09:39
qi-bot[commit] Lars-Peter Clausen: [package] ks7010: Remove debug printks http://qi-hw.com/p/openwrt-xburst/e231a3109:46
qi-bot[commit] Lars-Peter Clausen: [xburst] jz_mmc: Drop warning about spares irqs. http://qi-hw.com/p/openwrt-xburst/e86a9a009:46
wolfspraulhe, nice!09:46
wolfspraulthanks a lot!09:47
dvdkmorning09:54
kyakhi David09:54
wolfsprauldvdk: hi good morning! the 02-23 image has so many goodies from you - thanks a lot for your wonderful efforts!09:59
dvdkwolfspraul: don't mention it :)10:00
dvdkkyak: finally managed to make mplayer work.  need -vo fbdev10:00
dvdk-vo sdl hangs on startup.10:00
kyakhm! -vo sdl is the only output working correctly for me :)10:00
kyakfbdev and fbdev2 are somewhat half-screen10:01
dvdk-vo fbev have some kind of artifect that looks a little like dither.  maybe a bug with mplayer's non-assembly version of the yuv->rgb convernsion10:01
wolfsprauldvdk: in zgv, when I go to the help screen the output is garbled10:01
dvdkalso performance is somewhat underwhelming (using theora)10:01
kyakdvdk: do you play 320x240 videos? do you prepare it for Ben before ?10:02
wolfspraulmaybe we can make use of the xburst simd instructions10:02
dvdkwolfspraul: some of the menus in zgv are designed for the 640x480 std vga mode.  can't be easily fixed10:02
dvdkkyak: i transcoded a video to theora 320x180 (16:9)10:02
wolfspraulingenic has patches, not sure how easily they apply to our sources10:02
wolfspraulof course there is work in generating the right machine codes etc.10:03
dvdki have a little experience w/ mplayer hacking.  thought about adding a vo driver that uses the ingenic's hardware support for yuv->rgb and scaling10:03
kyakdvdk: right, i found theora/ogg have some performance issues. I use mpeg1video+lame for best peformance (mplayer is built with PATENTED)10:03
dvdkdon't think that intrinsics would help too much.  at leasts its _a_lot_ of effort relative to the yield10:03
kyakdvdk: there is mplayer release from ingenice, btw10:04
dvdktheora was never too performant.  getting ffmpeg's theora driver to work might give quite some performance10:04
dvdkkyak: yeah, heard it's terribly hacky10:04
kyaki had a look at their patch, the mostly work around some floating point calculations and sound issues10:04
kyaki have a feeling it's not very useful10:05
kyakdvdk: those UBB things, which you are going to distribute.. do you ship to Russia? :)do you think it can be fitter inside the envelope and sent as a usual mail?10:06
dvdkkyak: sure, should work10:07
kyakdvdk: btw, you can have a look at /root/.mplayer/config and change some options. the -vo sdl is set as default (as the most working video out)10:07
dvdkbtw ggi also works a little10:08
dvdk(-vo ggi)10:08
kyakdvdk: the UBB board and cable are pretty thin. But there are a matching connector in the other side (like COM port connector). I'm worried if it can be fitted in the envelope.10:09
dvdkoops with -vo ggi i get "bad packet in stream" etc.  memory corruption?10:09
dvdkkyak: i thought ubb was without any cable.  you only get the board10:09
kyakdvdk: same as -vo aa :) btw, svga output is not working, which is a little strange10:09
wpwrakdvdk: unless you yourself add something ;-)10:10
kyakhm, wait a second...10:10
dvdkkyak: there is some problem with svgalib's 'linear' mode afair.  maybe mplayer uses that10:10
kyakdvdk: so you will be shipping this: http://www.tuxbrain.com/img/davidblog/ubb_proto_sadowed.jpg10:11
kyak?10:11
Action: dvdk doesn't know. he just pre-ordered.10:11
wpwrakdvdk: there's your change to enter big business - UBB soldered to whatever other connector. 100 mil header, DB9, PS/2, RJ-45, SCART, ... ;-)10:11
dvdk:)10:11
kyakhehe10:11
dvdkwpwrak: scart sounds nice.  think we can bit-bang it?10:11
wpwrakkyak: that should be what you get, yes10:12
kyakwpwrak: argh, then it's soldering again10:12
wpwrakdvdk: would be an interesting exercise ;-)10:12
wpwrakkyak: *easy* soldering :)10:12
dvdkkyak: if you need something simple, i could pre-solder it here (have some equipment, but didn't use it for quite some time)10:12
kyakwpwrak: do you believe that i don't have a soldering iron at home? :)10:13
dvdkbtw i'm already planning to use a propeller to build a nanonote->vga adapter10:13
kyaki have at work, but my soldering capabilities are.. less than minimal10:13
dvdkprobably not true-color but might be cool anyways10:13
kyakdvdk: only if you could solder the trail :)10:14
dvdkjust periodically copying /dev/fb0 over the mmc bus (probably via bitbanging)10:14
kyakdvdk: from the other hand, i will need to solder anyway.. so don't bother :)10:14
wpwrakkyak: basic soldering is easy and fun to learn :)10:15
dvdkkyak: ok.  just tell me how many you want so i can update my files (and p.m. me a postal address)10:15
wpwrakkyak: besides, if there's soldering equipment at work, there's probably also someone around who's good at using it.10:15
dvdkkyak: russian postal system.  hmm.  how high do you think is the loss rate?10:15
kyakwpwrak: last time i tried an sd/mmc hack for my linksys router. It requires soldering of sd card connector to sdio pins. You wouldn't like to see the results, i had to ask some more soldering-fluent colleguaes to fix that :)10:16
wpwrakdvdk: if you need the exact loss rate, then i'd suggest wrapping the UBB(s) into a few Euro bills :)10:16
kyakdvdk: it's 75% probability it wil get lost10:17
dvdkwprwrak: 100%?10:17
wpwrakkyak: lemme guess: you didn't have or didn't use flux ? :)10:17
dvdkkyak: ok, then order 10, maybe you'll get 2.  forward-error-correction :)10:17
wolfspraulkyak: seriously? 75% chance it will be lost?10:17
kyakwpwrak: no, i used all the proper tools for soldering :) It's just my hands, they are not eqipped for that10:18
wpwrak(75%) my thoughts regarding the argentina postal system are already a bit less murderous10:18
kyakwolfspraul: i'm not sure about regular (paper) mail. But i'm hearing all the time about lost parcels10:19
wpwrakkyak: (hands) hmm. that can happen.10:19
kyakdvdk: i think i will order 3. In three separate e-mails :)10:19
kyakmaybe even separate courries10:19
wpwrakkyak: maybe drink more/less vodka before the soldering ? :)10:19
dvdkkyak: a RAID shipment10:19
kyakhehe10:20
wolfspraulI've sent a few parcels with NanoNotes to Russia and Belarus.10:20
wolfspraulboth with fedex and EMS (=national postal service). all arrived.10:20
wolfspraultotal was 5-10 I think.10:20
kyakwolfspraul: if i rememmber correclty, it took around 3-4 weeks for Ben to arrive via EMS. And i thought it was a good timing10:21
dvdkanother topic: had a look at NAND driver10:22
jekhorBelarus and Russia post service quality are very different :)10:22
dvdkhere is some stuff I found out: http://en.qi-hardware.com/wiki/UbiAttachTime10:22
dvdkAnybody thinks it.s an especially bad idea to try to use software-ecc10:22
dvdk?10:22
larscyes10:23
larscand the driver should support subpage reads10:24
larscbut the nand chip we use does not10:24
larsciirc10:24
dvdkwhat about performance?  the driver does busy wait for the nand's ecc.  any chance that software might actually be faster?10:26
dvdklarsc:  yeas means "bad idea"?10:27
larscyes10:27
dvdklarsc: why?10:27
larsci assume that it is slower10:28
wpwrakdvdk: (sw ecc faster) of the same data ? rather unlikely.10:29
larscand it the default linux ecc is weaker then the jz4740 hw eec10:29
dvdkahh.  you mean ecc is done by the ingenic chip, not the nand flash IC?  didn't realize that when reading the sources.10:30
rohdvdk: thats usually that way.10:30
rohthe nand controller or sw does the badblock management and ecc10:30
dvdkok, but maybe we can restrict the ecc on only the part thet we're interested in.  currently it reads and checks the full page.  instead read full page, but only check subpage?10:31
larscdvdk: the error correction is based on 512 byte blocks. so reading at that size should be fine10:31
dvdkfrom the sources it looked like ecc-size=72 or something.10:31
Action: dvdk looking for the source10:31
dvdkok so ecc-size is just the number of check-bytes?10:32
dvdkhttp://projects.qi-hardware.com/index.php/p/openwrt-xburst/source/tree/master/target/linux/xburst/files-2.6.32/arch/mips/jz4740/board-qi_lb60.c10:32
rohdvdk: i think ecc is done on full pages... in the end thats the smallest entity that can be written as once10:32
dvdkroh: looking at the source it is clearly done in smaller steps.10:32
rohs/once/in one chunk10:32
larscas i said ecc calculation is done on 512byte blocks10:33
rohbut i think in one thing you may be right. sometimes sw ecc is faster than hwecc. mostly because some hwecc implementations are bad to integrate/generate additional overhead in io/waitstates etc.10:34
tuxbrain_awaywolfspraul: I'm planing to create a project called ArduNote having the work I'm doing porting the avr-tools+ a todo list including a graphical front end "a la arduino"(one-button compile/upload+code-editor) that fits on NanoNote, do you think the qi wiki is the right place to start? or you prefer I mantain it only on tuxbrain domain?10:34
rohbut thats usually true one soc with fast cpus.10:34
rohand 'old' nand controllers10:34
dvdklarsc.  ah well, found it.  yes 512 bytes.10:36
dvdkstill much smaller than one page.10:36
dvdkpage size is 8k?10:36
rohi think 2k is usual on nand.10:36
larscit's 4k for the nanonote10:37
rohlarsc: ah. thanks.10:38
dvdkyeah, datasheet:  :34 <roh> i think 2k is usual on nand.10:38
dvdk> sorry.10:38
dvdk:34 <roh> i think 2k is usual on nand.10:38
dvdk> http://downloads.qi-hardware.com/hardware/datasheets/qi_lb60/U9~K9GAG08U0M~~2GB-FINAL_NAND-NOT_ON_SOME_PROTOTYPES~~.pdf10:38
rohwell.. i just guessed with the 2k. thats what ive seen on lots of soc so far10:39
Action: dvdk d trouble pasting a link into emacs :)10:39
dvdkso we could reduce the amount of ECC work by factor 8?10:39
dvdkduring UBI attach?10:39
dvdkby only checking the first 512 bytes of the page?10:39
larscits not the ecc calculation thats slow10:39
larscit's reading the data10:39
larsceven if you'll only read the first few bytes of a page you won't gain much of a speedup10:40
dvdkfrom the datasheet ' The 4,224 bytes of data10:41
dvdkwithin the selected page are transferred to the data registers in less than 60us10:41
dvdk'10:41
dvdksure these 60us are the part that takes most of the time?10:42
dvdkok, so with 60us page read time i get 245ms (!) for a complete UBI attach.  where is the other 99.9% of time spend?10:44
dvdkall the other parts of the operation can implement partial reads, can't they?  only transfer the first 512 bytes of the data register, only do ecc on the first 512 bytes etc.10:45
larscprobably reading the bytes, then10:46
dvdkwell, directly connected to the SD-RAm data bus, isn't it.  but only transferring 8 bits at a time?10:46
larscthe nand chip?10:49
dvdklasc: from the schematics.  yes afair.10:50
dvdks/lasc/larsc10:50
larscthe nandchip is connected to the soc via an 8bit bus10:50
dvdklarsc: isn't it sharing address and data pins with the sd-ram.  or do the schematics use the same names for different signals?10:52
dvdksharing data pins, not address pins afaics10:53
dvdks/pins/signals10:53
rohhm. maybe we should use a faster nand connection next time. 8bit isnt much at once.10:53
dvdkso how do we up the clock rate of the 8-bit bus?10:55
dvdkschematics say it can transfer 8 bit in 25ns10:55
dvdkor is there some dma unit that could automatically transfer more than 8 bit at a time?10:55
dvdks/schematics/data sheet10:55
larscnope10:58
dvdklarsc: nope what?  pick (a) no dma unit?  (b) no way to make the bus go faster?10:59
larscwell, you can tell the dma controller to do 8bit reads on the nand bus and do 32bit writes into main memory10:59
larscbut last time i tried that it was slower then just doing io with the cpu11:00
dvdklarsc: ok.11:00
larscand timings are already at max iirc11:00
dvdklarsc: i just relaized that you are the person that wrote all the code :)11:01
larscbut you might indeed gain some improvements if you do not read the whole page11:01
larscnand_do_read_ops does all the work11:02
larscyou could extend the read_page callback to take a bytes argument and stop when enough bytes have been read11:02
Action: dvdk needs to look at the code to understand what larsc is writing about11:03
larscmight save a second or two during bootup11:03
dvdklarsc: hmm, thought we might really drop 7/8 of the ubi attach time.11:04
dvdklarsc: you mean with the small ubi partition?11:04
dvdklarsc: ok thanks that was really helpful and probably saved me a lot of futile tinkering.11:07
dvdklarsc: I'll have a look at implementing partial reads.  but need to wait for a sufficiently long 'timeslice'.  maybe not before april.11:08
Action: dvdk is going to paste larsc comments into the wiki page11:08
rohhm. wtf is librewrt?11:13
wolfspraulif it's a second or two with a 512mb rootfs, it may still make the difference that we can go to a 2gb ubi partition without bootup time becoming unacceptable.11:13
kyakdunno. but it does support beeping with a speaker! ;)11:14
wolfspraultuxbrain_away: feel free to use Qi resources as they can help you. wiki or projects server or downloads.11:14
dvdklibrewrt is to openwrt what gnewsense is to debian :)11:14
wolfspraulsounds like your stuff is all free software, so totally fine. you can also export and move elsewhere later, we have a precedence for that as well with linuxencaja :-)11:14
larscdvdk: it might be helpful trac the actual read commands send to hw, to see if ubi really only reads the first 512bytes11:14
wolfsprauljekhor: you say belarus and russian postal service are very different. in which way?11:14
Action: dvdk has troubles doing any real work. as soon as he sits down his son starts crying.11:15
wolfspraulI will continue to maintain the wiki (mediawiki), projects (indefero), and other assorted services. no big new installations on the server planned, just upgrading, maintaining, cleaning, polishing, etc.11:16
rohdvdk: so you mean librewrt will be completely broken and inferior for stupid reasons?11:16
rohwell.. debian already is partially also ;) but not that far.11:16
dvdkroh: show more respect for the extremists :)11:16
wolfspraulI met the librewrt guys in person, they are nice fellows.11:17
dvdklarsc: hmm, technically ubi attach should only read the first 128 bytes or so.  need to do more source code browsing for whether it can use partial reads.11:17
wolfspraulthe bootup screen shows the GNU logo11:17
rohdvdk: i am an extremist from a few pov myself.. but what happens there is just plain stupid. (like removing fw blobs etc)11:17
dvdkroh: stupid only if you depend on those blobs.11:18
rohi am 'extreme' in some cases for technical reasons. i dont see any intelligence in being that for idealistic ones.11:18
dvdkroh: what would you say if >50% of linux were blobs?11:18
rohdvdk: everyone does. try booting a computer without acpi ;)11:18
rohdvdk: i wouldnt use it. besides. these blobs aren part of linux. they are part of the hw.11:18
dvdkroh: i don't like that.  i'd actually install coreboot, if there were any chance that it actually worked.11:18
rohdvdk: not in a rms-way (he doesnt differenciate between on which cpu something runs, or simply doesnt have a clue about hw)11:19
rohdvdk: you will never get coreboot running without either acpi reverse engineering and or detailed board specs (which boardmakers usually dont document)11:19
wolfspraulI hope we are growing some more Milkymist fans here... taking free software to the next level(s)11:20
rohall the gpios for example are 'wired' there11:20
rohwolfspraul: heh. get me a free toolchain ;)11:20
dvdkroh: no the problem is if you destribute blobs you are bound by the blob's license.  that's very non-nice.  it effectively means that you don't really own the hw you bought because any license in the blobs may remove right that a normal buyer would have.11:20
jekhorItems posted with ordinary tracked Airmail from HK to Belarus usually are shipped after 1-1.5 week after posting. To Russia --- 3-4 weeks. And Russian people told about problems with reliability. Sometimes items aren't delivered or even are replaced by another items.11:20
rohxilinx and their fucked up sw is the reason i dont play with fpga every day11:20
wolfspraulroh: you know people work on that, in small steps and in many places.11:21
wolfspraulhttp://en.qi-hardware.com/wiki/File:Nanonote_librewrt.jpg11:21
wolfspraulhere it is, I saw it. LibreWrt booting on Ben NanoNote.11:21
dvdkyeah was quite amazed to find out about the llhdl stuff  http://www.milkymist.org/llhdl/11:21
rohdvdk: the blobs they remove they dont because of licensing. they do because of purely idealistic reasons. its mostly blobs running on seperate cpus communicating with the free driver. my freedom requirements are 'everything which is on the cpu i run linux, and or being able to access its memory space freely11:22
rohwolfspraul: sure i know ;) just wanted to make a point. currently the toolchain issues make mm not any better than if one buys a documented soc like in the nanonote. i would 'rate' both having the same 'level of freedom' currently.11:23
dvdkroh: idealistic is not brain-damaged.  they have these ideals for a reason.11:23
rohdvdk: i dont see that. ideals are for getting one further. not stand in your way.11:23
rohdvdk: to be fair i wouldnt have called it 'wrt' then. it seems it doesnt even support (the completely proprietary) wrt hw line11:24
dvdkroh: these blobs may become a problem to free software in general (restrictiev licensing).11:24
dvdkroh: of course you could ignore the issue until it is too late.11:25
dvdkroh: see how sony was able to just remove functionality from the h/w people owned (ps3 linux support).11:25
dvdkroh: with blobs similar stuff may be possible for your PC in the future.11:25
rohdvdk: i dont. but these people are completely ignorant and irrealistic if they believe they get somewhere that way. their energy would be better used to write code instead of splitting the community without reason.11:26
rohdvdk: the binarys we are talking about are mostly DISTRIBUTED WITH THE KERNEL SOURCE. and have yes, no source, but a license explicitely granting stuff like re-distribution.11:26
dvdkroh: i don't think they're splitting the community. they just try to create a branch with different design goals and see how far they get.  take it as an experiment.11:27
rohso no, there is no reason to remove them besides to 'break support for existing hw'11:27
rohdvdk: i dont see branches there. i see copies.11:27
rohbranches can be merged back.11:27
dvdkroh: copies can be either.  note that debian packages are not handled by any revision control anyway (wrt gnewsens)11:28
rohlibrewrt is a fork. not a branch.11:28
rohdvdk: sony will be tried in court.11:28
dvdkroh: you sound like these people are hurting your experience. they're not.  their tinkering just have different goals.11:29
dvdkroh: sony already won in first instance.11:29
rohdvdk: what they did is computer sabotage and data modification. atleast here thats a criminal offense.11:29
wolfsprauljekhor: ah ok, so you say Belarus is better. interesting. Belarus has a bad reputation but maybe it's not true after all.11:29
Jay7morning11:29
rohdvdk: nah. they didnt. the real trials will come. juristic stuff takes time.11:29
wolfspraulwell my shipments so far all arrived, a little faster or slower, a little more or less expensive, but they all arrived.11:29
rohdvdk: btw: the real problem are people buying hw which they know needs binaries. so.. basically every nvidia customer.11:30
jekhorwolfspraul, yes, Life in this country is not buitiful, but post service works good :)11:30
wolfspraulbut I think I never shipped 2 or more nanos to either Belarus or Russia, in a few cases it was close but then we hesitated (also the buyers).11:30
wolfspraullarge shipments so far only went to USA, Europe and India (large means >= 50 units)11:31
Jay7hehe11:31
dvdkroh: yeah luckily there are alternatives.  how are the people writing free nvidia drivers better than the gnewsense guys :)11:31
dvdkafter all the free driver is pretty broken, too :)11:31
rohdvdk: also one has to see... sony is like a screaming and beating child. they are desperate. why? because they know they have lost (factual. the hw is out there)11:32
jekhorMy friends ordered few ZipIt devices in one post item, and they have no problems. But problems are possible, yes.11:32
Jay7wolfspraul: people here just 1) don't know much about NN 2) don't know what to do with NN :)11:32
Jay7and $99 are relative sensible money11:32
dvdkok trying to do some work (emacs still needs a bugfix)11:32
dvdkcu11:32
rohdvdk: writing drivers for hw where the vendor actively tries to avoid you is bad for the community. i would rather have the people use their time on stuff which gets us all further not only people who gave the wrong company money.11:33
Jay7but seems I understand now how to raise NN sales here11:33
rohits 'honorable' but in the end the nvidia foss driver writers support the wrong side of the industry. they should know that.11:33
wolfspraulthe issue of how low we have to take the free software before we can be relaxed about buying hardware and not falling into proprietary traps is difficult. I guess we are also still learning what the best boundary line is.11:34
wolfspraulI think once the same component can be bought from two separate, independent companies, it should be ok for sure. even if it is not exactly the same, but easily replacable.11:34
Jay7wolfspraul: I'll at least try to help you with sales in Russia11:34
rohwolfspraul: low? in price?11:34
wolfspraulno, in hidden proprietary knowledge/software11:34
wolfspraulJay7: great! really? that's really fantastic news then.11:35
wolfspraulI think Russia has real potential, where with 'real' I think we could have sold 50 Nanos to Russia by now, instead of 5-10. If we only had a local distributor.11:35
wolfspraulthe problem is that whether we sell 10 or 50 units, that distributor cannot get rich in any way, so it's hard to find the kinds of people like David from Tuxbrain :-)11:35
rohwolfspraul: well.. my rule is: as long as the vendor supports free drivers with either documentation and or code and support. its ok. when it needs blobs: its only ok when they can be distributed without any restriction (same as gpl) and the whole 'hw interface' as well as how to upload the binary is known. also the 'chip' may only share a hw-limited memory window and or documented interfaces with the host cpu.11:36
wolfspraulroh: because you talk about nvidia.11:36
wolfspraulthe proprietary part will grow. if their business model is to differentiate over that part you are screwed sooner or later. I don't need to wait for that day.11:37
Jay7wolfspraul: I can't sell it directly but I'll try to explain things people can do with NN11:37
wolfspraulif the redistribution license today is not good, my experience is that it is very hard to get them to change their redistribution terms to something that can easily be shipped alongside free software.11:37
wolfspraulI've spent a lot of time trying to convince companies to change to free software friendly redistribution terms for their proprietary blobs.11:37
kyakwolfspraul: first of all, it's hard to import 50 Bens to Russia. It is a bureaucratic hell, i'm not even sure if a private individual can do that, and if such devices need to be licensed or whatever11:37
wolfsprauleventually I decided that I value my time too highly to continue with that :-)11:38
Jay7people are thinking that NN is bad because it have less RAM, slow CPU, etc11:38
rohwolfspraul: nvidia is fucked from my pov. the reason they are not supporting foss is afaik that its not their decision. from what i've learned most of nvidias past is cluttered with sgi and their patents and 'ip11:38
roh'11:38
Jay7but they should consider things they may do with NN11:38
wolfspraullooks like nvidia is doing great11:38
Jay7that is the point from my POV11:38
rohso even if they want to.. its not their 'ip' they built the company on. nvidia is in the end only the 'consumer electronics outlet' for sgi inventions.11:38
wolfspraulwow, market cap 13.5 billion11:39
rohor atleast it was. thats already a 'old' story.11:39
Jay7kyak: we shouldn't even try to import ben's officially11:39
Jay7and non-officially too :)11:39
wolfspraulstock up a solid 50% in the last 12 months11:39
Jay7just let's help other to buy direct from wolfspraul e.g. :)11:39
Jay7and provide support11:40
Jay7docs/service11:40
Jay7translation :)11:40
wolfspraulroh: I think nvidia is just doing fine, really :-) they can live with you thinking they are ... while they are trying to keep overview over all the money coming in :-)11:40
rohwhen ati/amd doesnt go belly up in the next few years i think at some point nvidia will be bought by some major chipvendor.11:40
rohlike via or even intel.11:40
Jay7via have very hard times11:41
rohi hope ati/amd survives. they are the only ones keeping intel in balance when it comes to freedom.11:41
rohJay7: well.. not enough interresting products11:41
kyakJay7: when i first ordered Ben, i was thinking about buying 10 pcs. to distribute it locally afterwards. And you know what? I think i would still be having them all at hands by now :)11:41
Jay7yeah... I'm sure problem of management, not R&D11:41
rohJay7: ack.11:41
tuxbrain_awaywolfspraul: I'm planing to create a project called ArduNote having the work I'm doing porting the avr-tools+ a todo list including a graphical front end "a la arduino"(one-button compile/upload+code-editor) that fits on NanoNote, do you think the qi wiki is the right place to start? or you prefer I mantain it only on tuxbrain domain?11:42
Jay7kyak: yeah.. but now I understand what is NN good for at least :)11:42
rohone cannot beat giants in what they are good at. one needs to be innovative. i think thats what ati did. thats also why they have (imho) the better hw. (not sw)11:42
kyakJay7: tell me what it's good at? :)11:43
Jay7anyway, I'll try to write some articles about NN, MM and possibilities :)11:43
rohtuxbrain_away: not much to do: get avr-gcc and avrdude compiled, install gnu make and some text editor.11:43
Jay7kyak: to 1) self-education 2) make other wonderful things (with UBB, atBen, etc)11:43
wolfspraulnvidia is worth 13 times more than via. without looking into further details (balance sheet), it's safe to say that without a genius-level financial maneuver, via cannot buy nvidia.11:44
tuxbrain_awayroh:I kwow is feasible that why I will do :)11:44
tuxbrain_awaythat's why11:44
rohwolfspraul: maybe. pick another giant.11:44
kyakJay7: i thought, that instead of inventing possible use cases for Ben, it is better to give people a link to Applications oage on qi's wiki. Then everyone could figure out a use case by his own.11:44
Jay7kyak: sure11:44
wolfspraulyeah wait, I am just trying to get a few facts into those theories :-)11:44
wolfspraulalthough I cannot keep up with the speed one can suggest new ideas.11:45
wolfspraulnvidia is doing great, really11:45
wolfspraulmaybe that's because they have a good proprietary strategy, not although they have one11:45
wolfsprauldon't know11:45
rohwolfspraul: in the end nvidia cannot compete with ati-amd when cpus, gpus and other simd stuff gets into the melting pot with soc and io as well.11:45
wolfspraulok, amd. one sec.11:45
Jay7kyak: other problem - we have no good HW devels here.. because of Russia have almost no HW industry..11:45
wolfspraulwow11:46
wolfspraulamd is worth half of nvidia!11:46
wolfspraul:-)11:46
Jay7but NN is very good for learn it :)11:46
rohwolfspraul: amd fusion will hit nvidia hard. they dont have cpus at all. and intel 'accepts' nv.. but isnt its friend.11:46
wolfspraulso nvidia could indeed buy amd (without looking into details, but they could definitely go to a bank and see what is possible, if they wanted to)11:46
rohintel on the contrary failed on designing a state of the art gpu. so i could imagine they would be interrested in nv at some point11:46
wolfspraulamd is worth 6.3 billion USD, nvidia 13.511:47
Jay7wolfspraul: nvidia can't buy amd because of anty-monopoly things at least11:47
rohwolfspraul: the 'value' isnt the only thing.11:47
wolfspraulamd just lost their CEO, right? whereas nvidias ceo is running the show for several decades, no?11:47
wolfspraulwell you talk about who buys whom or merges or what11:47
kyakJay7: it's not quite true. There is almost no CHIP production in Russia, but a lot of bright people who can figure how and what can be done with these chips :)11:47
wolfspraulso I just look at the very first number I can find, market capitalization.11:47
wolfspraulagain: nvidia is doing great, right now11:47
Jay7kyak: well.. seems you are right more than I'm11:48
rohrussia has quite some vhdl/fpga design labs i think. seen offers more than once11:48
rohwolfspraul: i would be interrested rather in 'whats the vendors cash cow'11:48
wolfspraulintel could buy nvidia, or at least they could try :-) but that would be a huge acquisition, not easy. I doubt that will happen.11:48
wolfspraulyes ok, sure. I just look at one number now because it's the easiest to look up.11:49
wolfspraulI also never like nvidia11:49
wolfspraulsince the 90's11:49
wolfspraulnever11:49
wolfspraulbut that's not the point here, you talked about who is merged/bought by whom11:49
wolfspraulnvidia screwed me over so many times with graphics drivers, I am done with them as a Linux user11:49
wolfspraulI don't even care anymore whether they are 'better' now.11:49
rohi think its a question of attitude. nv doesnt have a nice one. they dont like their customers (their hw-fail series with the broken chips showed that)11:50
wolfspraulfine. but it seems to be a great business (business defined as a money making machine).11:50
wolfspraulso it won't easily be bought or disappear11:50
rohbah. money is boring.11:50
wolfspraulbecause many people benefit big time from its existance11:50
rohit sucks if you dont have any. if you have too much is even more annoying.11:50
wolfspraulok. still I suggest you don't wait for its disappearance. that might be even more boring.11:50
wolfspraulthey will be around for a loooong time, it looks like11:51
Jay7well.. time to implement test-mode UI into kexecboot..11:52
rohwolfspraul: btw.. you know why intel dumped pulsbough?11:52
Jay7I hope I'll try to run kexecboot this weekend - today or tomorrow11:52
Jay7@NN I mean11:52
roheh poulsbo11:52
wolfspraulno idea, don't even know what that is11:53
rohbecause of GMA500. its a sgx from powerVR they licensed11:53
wolfspraulI just read more about nvidia, the founder who founded it in 1993 is still running it. amazing.11:53
rohpoulsbo is a intel x86 chipset with integrated graphics.11:53
wolfspraulunless he steps down or gets to old, expect them to be around and to be aggressive.11:53
rohbut not their low-end stuff, but the sgx high end variant (well.. high end for powerVR)11:53
wolfspraullast fact check - he's only 47 years old now.11:54
wolfspraulfounded the company when he was 30.11:54
rohintel has the guideline to provide documentation and or drivers even for linux for all their products. powerVR got in the way (and was dumped)11:54
wolfspraulno way, nvidia will not be bought the next few years. I've made my bet :-)11:54
rohwolfspraul: not neccessarily the ceo to say that. ive seen more than one unfriendly takeover now11:55
wolfspraulborn in Taipei, look at that11:56
cfyhi all,thanks all,i can run scheme on nn. :)11:56
Action: Jay7 still imagine NN-based book reader..11:56
wolfspraulroh: if a company was founded by someone and run by that same person the last 17 years, and is profitable, has cash reserves, and and and, then there is no way any sane person would attempt a takeover without getting that person on board.11:56
wolfspraulso yes, I can take a brief look at that guy if I speculate about nvidia's future, imho11:57
rohwolfspraul: hehe.. you know porsche and VW? ;)11:57
wolfspraulcfy: did you port any additional software for that?11:58
wolfspraulor you just want to tell us that it worked out of the box?11:58
wolfspraulif you needed extra software, can we include it in the future? what was missing?11:58
cfywolfspraul: no..11:58
rohwolfspraul: well.. nv could have a future.. maybe even for foss.. but they need to change their attitude first. maybe their technology isnt just 'videogamer optimized' .. but who knows.11:58
rohon ati i know that they outperform nv as soon as you dont only use single precision but double and 32bit color. atleast it was like that for a loooong time.11:59
rohmuch more quality tuned. i know some tv station broadcast hw which used amd chips due to that. better colors12:00
cfywolfspraul: if it can run some common lisp interpreter,such as sbcl.that's would be great :)12:00
wolfspraulcfy: are you saying you can run scheme, or are you asking whether you can?12:00
cfywolfspraul: i'm learning lisp,then i find guile can run on nn12:01
wolfspraulwe have guile on it right now12:01
cfywolfspraul: i can run guile12:01
wolfspraulok12:01
wolfsprauland now you want more? what?12:01
wolfspraulI don't know the exact differences between those things like lisp, scheme, guile12:01
cfyno,i just want say thank you12:01
wolfspraulor what implementations exist, or which implementations make sense for the Ben12:01
wolfspraulah great, got it12:01
wolfsprauleven better!12:01
wolfspraulzedstar was behind the guile idea for a long time...12:02
wolfspraulcfy: thank you for buying a Ben!12:02
cfywolfspraul: hehe.thank you for developing:)12:02
wolfspraulI did the least. only reflashing in the factory, really.12:03
wolfsprauland where you immediately overwrote my hard work the moment you got your Ben, I assume :-)12:03
larsci just implemented dvdk idea and it reduces bootime by 3 seconds when the rootfs is on the large partition12:04
wolfspraulwow that's nice. 3 seconds is a lot!12:04
wolfspraulI guess he can scratch his plans for April then :-)12:04
larsc1/312:04
wolfspraul1/3rd second, or 3 seconds?12:05
larsc1/3 of the boottime12:05
larscna actually more 1/3 of the mounttime12:05
wolfspraulit takes 8 seconds on my ben to mount the 1.5 gb data partition12:06
larscwithout the patch:  10.940000] VFS: Mounted root (ubifs filesystem) on device 0:11.12:06
larscwith the patch: [    7.860000] VFS: Mounted root (ubifs filesystem) on device 0:11.12:06
rohnice. more than i'd have expected.12:09
wolfspraulthat's with a 2 gb ubifs partition?12:09
larscyes12:10
wolfspraulI get to that point at 4.23 seconds on my Ben, with a 512 mb partition12:10
wolfspraulso with your patch, that means with a full 2gb partition, boot time would increase by 7.86-4.23=3.63 seconds12:10
wolfspraulthat's bearable I think, maybe we can make that back elsewhere :-)12:11
larscwell, the small partitions would mount faster as well12:11
wolfspraulsure. I just look at the bottom line and what we can deliver to the user.12:11
wolfspraulthere is not much point in having inaccessible nand space, so if they have to run some script or whatever mount hook later, those seconds will still be spent.12:12
wolfspraulwith an increase of only 3.63 seconds, maybe we can go to a full 2gb partition, and look at werner's pivot_root at the same time to get jlime (or other) dual booting going12:12
wolfspraulanyway I think the speedup is great news12:12
wolfspraulwe will find a way to use the new power :-)12:13
wolfsprauland you are right, there is no way we will waste several seconds just for the fun of moving the Linux kernel into ubifs, so that will stay outside for sure.12:13
kyakwolfspraul: remember the safe place across reflashes...12:14
wolfspraulwhat safe place?12:14
wolfspraulyou mean some space on nand not occupied by anything?12:14
wolfspraulor a small partition just so the partition layout is more flexible?12:14
kyakthe safe place for images/video/music/scripts/files/documents/whatever that won't be swapped off during reflash12:15
kyaki don't like the idea of backing up the whole Ben before i reflash12:15
dvdklarsc: wow.12:16
kyakwolfspraul: right now, i reflash the rootfs and all my files are kept in datafs. This is very reasonable12:16
larscdvdk: news spreads fast?12:17
kyakdvdk: are you reading the irclogs on web? :)12:17
dvdklarsc: openwrt compilation is slow.  gives me lots of time to read irc logs12:17
dvdkkyak: yup12:17
larschehe12:18
dvdkwaiting for an emacs rebuild12:18
larscdvdk: ubi reads the first 64bytes of the first page in an eraseblock and the whole second page12:18
dvdklarsc: now if we could detect and optimize *that* :)12:19
dvdklarsc: probably not possible.12:19
larschttp://pastebin.com/Ku6wmced12:19
kyakdvdk: note that weblogs are generated only 11,41 * * * *, so for real time experience you should stay here ;)12:19
dvdklarsc: wow again.12:19
Action: dvdk reading log output12:20
dvdklarsc: are you sure it's that pattern throughout all of flash?12:20
Action: dvdk doesn't understand why it has to read the second page12:21
dvdkor does it only read the second page depending on the contents of those 64 bytes?12:21
larsci think the second read is it reading the vid header12:22
larscand since we don't support subpage writes the vid_hdr is aligned to page_size12:23
dvdkbut online i find "UBI stores 2 small 64-byte headers at the beginning of each non-bad physical eraseblock"12:23
dvdkhttp://www.linux-mtd.infradead.org/doc/ubi.html12:23
dvdksecond read unneccessarily long?12:23
Jay7btw, are ubifs partitions showed in /proc/partitions?12:23
larscdvdk: thats the code:12:25
larsc    read_err = ubi_io_read(ubi, p, pnum, ubi->vid_hdr_aloffset,12:25
larsc              ubi->vid_hdr_alsize);12:25
larsc"vid_hdr_alsize: size of the VID header aligned to @hdrs_min_io_size"12:25
dvdkwhat's the difference ot the code of the first read?12:26
dvdkand what's the value of hdrs_min_io_size12:26
larschdrs_min_io_size: minimal I/O unit size used for VID and EC headers12:27
larscwhich is the page size in our case12:27
larscbut replacing the ubi->vid_hdr_alsize with the actual header size should work, i guess12:28
dvdkyou're too fast.  i can't really follow you :)12:29
larsc[    4.520000] VFS: Mounted root (ubifs filesystem) on device 0:11.12:29
larsc:)12:29
wolfspraulthat's 0.3 seconds slower than on my 512mb rootfs :-)12:30
larscso now mounttime is one by 2/312:30
dvdkwolfspraul: that's a timestamp12:30
dvdkyes, for me too12:30
larscwolfspraul: but the 512mb rootfs will probably be done after 1 second now12:31
wolfspraulyes, I compare with the same timestamp on my Ben12:31
dvdkfor me it's [    4.320000] VFS:12:31
wolfspraulsure sure, I know12:31
wolfspraulkyak: yes, but we want to move away from whole partition reflashing anyway12:31
wolfspraulof course we could still have multiple partitions (the moutn speed improvements apply no matter in which sizes the ubifs partitions come)12:31
dvdkyou mean larsc attached a full 2G flash?12:31
wolfspraulthe advantage of one large one is that working with multiple rootfses (dual-booting) is easier, and the boundary between code size and data size is more flexible (read: there is no fixed boundary at all)12:32
wolfsprauldvdk: yes, I'm sure that's what he did12:32
larscdvdk: yes i did12:32
wolfspraullet's see how this pivot_root stuff works, I really don't know and need to understand it first12:33
wolfspraulright now I mount my Ben, then I go to a console and manually run 'mtd.nn mount data /data' each time after booting. that's pretty stupid I think and not a good 'out of the box experience'.12:34
wolfspraulthe speed improvements will make many new things possible, that's really cool I think12:34
wolfspraul(of course I know I could automate the mounting by putting it into the right place in the startup scripts, but we could do this right from the beginning...)12:35
qi-bot[commit] David Kühling: emacs: force it to not use gpm and libasound http://qi-hw.com/p/openwrt-packages/b091aa612:35
qi-bot[commit] David Kühling: emacs: fix load-history problem by disabling gpm-mouse-mode by default http://qi-hw.com/p/openwrt-packages/bf91beb12:35
Jay7Zauruses have 'system' area (non-FS, raw nand + level wearing), 'root' partition and 'user' (or 'home') partition12:35
Jay7reflashing changes kernel in system area and rootfs in root partition12:36
Jay7reflashing routine is called by loading second kernel + initrd from system area12:36
Jay7by special key combo at boot12:36
Jay7JFYI :)12:37
wolfspraulI don't like our current usbboot-based reflashing business. never did. that's not the right way to treat nand either. is the default still to discard all bad block info?12:37
kristianpaulwolfspraul: (bitcoin) acepting bitcoin too in your shop? :D12:37
wolfspraulI'm looking forward to the day where we can control the nand experience entirely through the Linux kernel, not the stage1/stage2 usbboot stubs.12:37
rohi am hoping for the day we get rid of that pestilence ;)12:38
wolfspraulkristianpaul: yeah! you like that? I don't know much about bitcoin, but I will investigate a bit more.12:38
roh(nand flash sucks in handling)12:38
kristianpaulwolfspraul: I like12:38
Action: kristianpaul dont trust on banks12:38
Action: kristianpaul dont trust on paypal money 12:38
Action: kristianpaul dont trust on amazon money 12:38
kristianpaulso, yes12:38
wolfspraulso once we do that, there should be no reason to reformat the ubifs partition over and over anyway. we can mount a linux rescue/reflash/update os into memory, and do a software upgrade from there, preserving /data whether it's in a separate partition or not12:39
Jay7kristianpaul: you should go and pay cash then :)12:39
wolfspraulI've learnt of the existance of one-time pre-paid credit cards recently, but that bitcoin stuff takes it to a totally new level of course. I spent 5 minutes reading about it, I like what I read.12:40
kristianpaulJay7: I do most of time, but i can travel to taiwan or USA just for that. So a electronic solution is needed12:40
wolfspraulwill do more (so many things to do...)12:40
rohhm. how does one get papermoney to bitcoin?12:40
wolfspraulfor sure I'm crazy enough to sell my physical goods (nanonotes and milkymist one), for a p2p backed digital currency. yes I am. so that's a prerequisite fulfilled right there.12:41
kristianpaulJay7: Actualy cash is printed by Banks... thats other issue, but i'm not going to be extremists here12:41
Jay7kristianpaul: hehe :)12:41
rohkristianpaul: s/banks/states/12:41
qi-bot[commit] David Kühling: emacs: another patch necessary to disable gpm-mouse-mode http://qi-hw.com/p/openwrt-packages/e16211312:41
rohbanks dont print money. they just are the 'frontdesk'12:41
wolfsprauli'm sure there is some way to convert cash into bitcoins, and also some way back. well I need to buy eatable food once in a while, so I hope there is.12:42
Jay7roh: I'll say even 'united states' ;)12:42
wolfspraulcannot carry the p2p digital bits to the market.12:42
rohJay7: nah. they are not that important (i dont take us$ for example. i make sure i get payed in euro (currently)12:42
kristianpaulroh: Okay, they give credit too ;-)12:43
kristianpaulJay7: If the bitcoin model of trust is big oh things are moving..12:43
wolfspraullarsc: those ubifs speed improvements are in the 2.6.37 openwrt tree? (or will be) How hard/easy is it to get them into 2.6.32 ?12:43
Action: Jay7 still need to implement ability to detect UBI and UBIFS partitions into kexecboot12:45
larscwolfspraul: they are on my hard disk.12:45
larscwolfspraul: the patches are rather small, so it should not be much of a problem to backport them12:45
wolfsprauleven better.12:46
wolfspraulactually in Hong Kong there are different paper notes printed by different banks.12:50
wolfspraulI don't understand how the UK system works, but every time I see this it makes me think "why can't I print my own money" or "I want to do that too" :-)12:50
dvdkbtw config.full_system still fails to compile with a pango error.  tried to scripts/feeds uninstall pango without success.12:51
wolfspraulyou see notes from this bank or that bank - hey. open a bank???12:51
dvdkhttp://mosquito.dyndns.tv/david/nanonote/buildlog-20110226-1021.bz212:51
wolfspraul3 banks in HK have a license to 'print money' (I'm simplifying), 8 in the UK12:54
wolfspraulanyway. bitcoins, we see...12:54
wolfspraulI read that the EFF accepts bitcoins as donations, that sounds trustworthy to me.12:54
rohthe uk system is weird. here no bank prints money.12:54
rohi wonder who fucked over the state to get that through ;)12:55
Action: kristianpaul want to acept bitcoins as donations too12:55
kristianpaulas NO bank in Colombia allow me to receive money by using paypal..12:55
wolfspraulgood point. bitcoins is probably egalitarian.12:56
kristianpauland i have "rich" here to apply fot such us features :-(12:56
kristianpaulhave to be*12:56
wolfspraulno discrimination against people in Colombia, Nigeria, or other places for whatever reasons.12:56
larscdvdk: i just build a kernel with ecc disabled and mounttime was the same as with ecc12:56
dvdklarsc: you mean ecc has infinite speed? :)13:00
larscno significant overhead13:02
larscit's just a few dozens instrutions anyway13:03
dvdkuploaded fixed emacs packages to http://mosquito.dyndns.tv/~spock/openwrt/13:04
dvdk(emacs-data has the more important fixes)13:04
kyakdvdk: an update of pango: https://dev.openwrt.org/ticket/8813. You can adapt the patch for yourself13:05
larscyou'll get more overhead if there is actually an error to correct, but just checking if an error occurred is negligible13:05
dvdkkyak: thanks for the info.  going to look at it .13:06
dvdklarsc: ah ok, so older nanonotes will be slower.13:06
larscdvdk: but anyway, you are the hero of the day13:06
dvdkkyak: wanted to write: going to look at it later, have to go now :)13:06
kyaksure :)13:06
dvdklarsc: you are the hero, implementing that stuff in an hour or so.  wold have taken me a full day at least :)13:06
dvdkkyak: irc-multitasking=bad13:07
dvdks/=/==13:07
larsc[    3.970000] VFS: Mounted root (ubifs filesystem) on device 0:11.13:29
kyakso it's getting faster and faster :)13:34
larscyeay, but i guess thats now about as fast as it can get13:35
kyaki'm looking forward to trying it soon13:36
wolfspraulwhere did that speed improvement come from?13:37
larscmagic13:38
wolfspraulthat's another 0.5 seconds less - great!13:38
kyakbtw, it's still 12.6 seconds between [    4.610000] Please be patient, while OpenWrt loads & and the next message [   16.000000] ks7010_sdio : Feb 17 2011 06:10:34.. I think a lot can be improved in openwrt's startup scripts13:39
wolfspraulyou are mounting a 2gb partition faster now than I am mounting my 512mb partition :-)13:40
kyakindeed!13:40
kyakvey cool13:40
larscthis is with a cleanly unmounted fs: [    3.880000] VFS: Mounted root (ubifs filesystem) on device 0:11.13:48
larsc20 seconds till gmenu2x13:50
wolfspraulit's perfect. the 2gb are not the lowest hanging fruits anymore, after your optimization.13:53
wolfspraulyou brought it down from 10.9 to 3.9 seconds, in a few hours13:55
rohnice one13:56
larscok, one last minor improvement and it's now at 3.73000014:00
wolfspraulhmm. even that doesn't sound minor. another .15 seconds, it's real stuff. seems to be a very healthy code area for optimizations indeed.14:01
larscmore interesting, now with mount time being down it actually makes sense to improve in other areas as well14:03
roh;)14:05
wolfsprauldo you have any ideas where it might be worth looking next?14:16
wolfspraulI vaguely remember some posts on the openwrt devel list recently about speed of startup scripts, but I forgot the details.14:17
wolfspraulit also seems gmenu2x is becoming a bottleneck, with more icons14:17
wolfspraulI still hope we can find a total replacement one day, a super minimalistic, small and fast launcher.14:17
wolfspraul100 kb memory footprint, and super fast14:18
wolfspraulok 100kb is a little extreme maybe, let's say 300 KiB :-)14:19
wolfspraulI don't know whether it's easier to slim down gmenu2x, or to find a replacement. which also has to do with the question whether gmenu2x is forked or not and if it is, what is happening at the other/upstream gmenu2x14:20
larscfor all i know qi-hw is gmenu2x upstream now14:23
rohhm.. how fat is it?14:28
rohi mean.. it looks like not that much code.. well.. c++14:28
rohwhat does it need xml for?14:30
rohah. good. i was mistaken. thought ive seen tinyxml in there14:32
wpwraknice ... dvdk the philosopher who postulates the speedup, larcs the wizard who implements it, and soon, I guess, wolfgang the bard who sings its praise on the list :-)14:35
kristianpaulqimenu? :-)14:37
rohhrhr. well.. i think it could be slimmer. and not c++.. but thats just my taste.14:37
rohhavent seen anything too bad. but there are header with nearly 100kb in it which i dont get what they are for14:38
wpwrakroh: cpp performance testing ? :)14:39
rohthere are ifdefs for msvc++ in there. looks like 'crap' to me.14:39
wpwrakseems that we really need qimenu :)14:39
rohmy guess is that recompressing the icons could make most optimisation useless ;)14:40
wpwrakprecompute the icons. then it's just readv(icon, iov, icon_lines);14:40
rohi mean.. how big should the binary be? loading all images for sure takes more time.14:40
wpwrakone step further: precompute the screen :)14:41
rohthats even slower (icons are smaller than the whole screen)14:41
wolfspraulwhat is the lowest overhead way to get the icons to the screen? directly use the linux framebuffer?14:42
kyakprecompute user's actions - and take it further: precompute that the user will exit and exit immediately!14:42
rohwolfspraul: well.. basically. yes. ot dfb or such.14:42
rohdunno about sdl overhead.14:42
mthit's probably best to benchmark first before guessing where the bottleneck is14:42
wpwrakroh: (slower) depends on the cost of file access in relation to the reads14:42
mthI'm sure gmenu2x can be done more efficiently14:43
rohmth: for sure.14:43
mthbut I'm not sure what parts are actually the slowest at the moment14:43
wolfspraulso just a small c app that blits icons to the framebuffer and allows scrolling around14:43
rohi dont know how it works in detail. does it parse _all_ freedesktop startfiles or only the ones shown?14:43
wolfspraulok if qi is upstream for gmenu2x already, and we cannot find an existing super small framebuffer launcher, then we may as well optimize/cut down gmenu2x sources more (I think there was some cleanup already)14:44
mthSDL is very useful for development, since you can run the program on your PC for testing14:44
wolfspraulthe best would be to find an existing actively maintained small launcher project14:44
rohmth: sdl is usually a good way to keep me away.14:44
wpwrakdoesn't gmenu2x also have some status displays ? battery, clock, and such ?14:44
wolfspraulyes, and lots of settings too14:44
mthroh: why?14:44
rohit also has shitloads of other stuff in there as well.. as audio volume and usb stuff14:45
wpwrakroh: sdl is quite friendly14:45
wpwrakqimenu then :)14:45
mthyou can init only the SDL subsystems you actually need14:45
mthand audio mixing and USB are not part of SDL itself afaik14:45
rohmth: i dont like the paradigma. also its not possible to use any acceleration anywhere. its just another senseless feature removing compatibility wrapper usually. (on x/win32/mac)14:46
wolfspraulwe have found and collected some really obscure but neat apps already, I cannot imagine that there are no existing small launcher projects somewhere14:46
rohthe usb and audio stuff is in gpmenu2 not sdl. (sorry for being unprecise)14:46
mthyou can use accelerated blits, if the driver inside SDL supports that14:46
rohwolfspraul: well.. there is the console ;)14:46
wolfspraulncurses launcher may be a bit radical, also would have trouble with icons14:47
mthonce thing that could be improved is to not let gmenu2x use dictionaries all the time14:47
mthI don't know if it has a measurable performance impact though14:47
wpwrakroh: sdl is nice for development. you make it work on the pc, then it runs on the ben. e.g., atrf-rssi was "ported" just like that14:47
rohwolfspraul: i like the ncurses idea ;)14:48
rohi still want a phone with a bw lcm which does mostly text (like featurephones 10 years ago) running something like a text-only shell14:49
mthI think if you init only the video subsystem of SDL and ask for a HWSURFACE, it's pretty much the same as opening a frame buffer and mmapping it14:49
rohmth: why then not mmap a fbdev instead?14:49
rohmemset and its black.14:49
rohioctl and the cursor is gone. not much more work than using sdl.14:49
mthto make testing under X11 easier14:49
rohbah. kids and their problems14:50
roh;)14:50
mthwhat I mean is, SDL is not a fat abstraction layer14:51
rohtrue. but its also mostly useless on a fbdev14:51
mthand while it does hide features, the features exposed by SDL map pretty well to the frame buffer14:51
rohatleast in our case14:51
wpwrakmth: roh doesn't like to chat with us. so he wants us to use development environment where we can't run a chat in parallel14:51
mthunlike if you would have GL ES capable hardware14:52
rohmth: but you are right. sdl isnt our real problem here i guess. its rather logic (load too much (all)stuff on start) or bad implementation of details14:52
rohmth: huh? do you think  devices with gl es are faster?14:53
wolfspraulgmenu2x is about 7800 lines of .cpp code, not as bad as I thought. another scary 4600 lines of .h though. thank you C++...14:53
mthI already removed the use of the google hash library from gmenu2x, now it uses libstdc++ for that14:53
wolfspraulsomehow I also think like roh a super small launcher could go directly onto the fbdev and be done in 1000 lines of C14:53
rohdo you get what FastDelegate.h does?14:53
mthbut I think it wouldn't actually need to use dictionaries (maps) all that often14:54
wolfspraulmth: if we accept that we are upstream for gmenu2x, we can go through that code with the big axe.14:54
mthwolfspraul: the original upstream seems to have rebooted and included the patches from qi git14:54
wolfspraulah, interesting14:55
mthbut I don't know how feasible it is to work with the original author14:55
wolfspraulis it still active?14:55
wolfspraulwell I also wouldn't want to ruin the name of the software by cutting it down to something quite different14:55
wolfspraulthat's not fair to the original project, especially so if it is still active14:56
rohhuh? why does he implement stuff like 'max' and 'min' ?14:56
mthhttps://github.com/mtorromeo/gmenu2x/commits/master14:56
rohall that 'powerfull language' c++ madness and then this? i dont get it14:56
rohhttp://projects.qi-hardware.com/index.php/p/gmenu2x/source/tree/master/src/utilities.cpp14:57
wolfspraulok that looks very active to me14:58
mthroh: there are std::min and std::max in C++, but it seems the author didn't know about those14:58
wolfspraulthat means we should be careful with further changes to the qi gmenu2x, unless we somehow have a joint plan with the upstream author14:58
wolfspraulotherwise we are creating a second software under the same name, and lots of confusion14:58
rohmth: heh ;)14:58
mthwolfspraul: I think we should decide soonish whether to join him or to fork + rename14:59
rohmth: maybe the crappy win32 compilers he tries do keep portability to need it14:59
kyakwhy do you care about gmenu2x upstream? just take it, and cut it to pieces :)14:59
rohwho knows.14:59
mthroh: the most recent versions of MSVC++ aren't that bad, it's the ancient ones that were horrible14:59
wolfsprauldon't know. if the upstream code works on Ben, I would always prefer that approach.14:59
wolfspraulif we do 'cut it down', then we need to give it a new name for clarity and to protect the name of the upstream software15:00
rohmth: well.. i dont care. i use gcc and maybe some time clang. but i dont care for any win32 compiler at all. their trouble, not mine15:00
wolfsprauland in that case we may either search for an actively maintained replacement first, or indeed write one from scratch directly with C on top of fbdev15:00
mthI prefer to continue from working code instead of rewriting from scratch15:01
mtheven if the end result is nothing like the current code15:01
mthat least there will be something working every step of the way15:01
rohmth: maybe we should benchmark first.15:02
mthand since C++ is a superset of C, you could even migrate the language without restarting from scratch15:02
rohcould be that libstdc++ is out terror on startup fail.. then we know that reusing is much work15:02
rohanyhow. i need to leave.. food. getting crushed ice. birthdayparty. :)15:03
mthlibs are paged in on demand, so the less you use of it, the less is loaded15:03
wolfspraulI agree with your approach ("something working every step of the way"). but we may need to rename it, and/or talk with upstream first.15:03
mthwolfspraul: if we want a full-featured gmenu2x, but more portable, more maintainable and more optimized, we should try working with upstream15:04
mthif we want a menu like gmenu2x bug significantly lighter in features, it's probably better to fork + rename15:04
mth*but15:05
wolfspraulanother approach would be to 're-base' on upstream, by converting our fork to a set of patches, and building from upstream+patches15:05
wolfspraulthat would preclude a big 'cutting down' though, that's more to bring gmenu2x back in line with the upstream author15:05
mththe amount of changes still needed is too large to make that practical, I think15:05
wolfspraulto make what practical? the fork + rename?15:06
mthto work with patches15:06
mthunless you mean start a new branch from upstream and cherry-pick our changes on that15:07
mths/on/onto15:07
wolfspraulsure, any way to create the initial set. nobody cared about upstream-syncing for over a year I think15:07
wolfspraulbut if we do that (upstream syncing), at least I will try hard to find a much leaner replacement. if I look at the upstream Changelog - there are all sorts of features there that I wouldn't want in a minimalistic launcher.15:08
wpwrakwolfspraul: (cut it down) if you really just need icons, scrolling, and selection, maybe it's better to start from scratch ?15:08
mththe initial set would be quite small, as it seems the original author accepted most of our changes15:09
mththe main reason no-one cared about upstream is because upstream had been dormant for years15:09
mthbut that changed recently15:09
wolfspraulsounds like we should get in touch and try to get the rest upstream, or switch to building from upstream+patches15:10
larscimo we should stick with gmenu2x. it's good enough15:11
Jay7iirc you need to launch some X apps as well15:11
Jay7so framebuffer isn't sufficient for you15:11
Jay72 wolfspraul15:11
larscso instead of dividing our (human) resources further by starting a rewrite we should collaborate with the new old upstream15:12
wolfspraulif we can get some speed and memory footprint optimizations upstream, and switch to building upstream+patches, maybe that's 80% of what we need anyway15:12
Action: Jay7 have some kind of fb-drawing library in kexecboot now15:12
Jay7but no X then15:12
wolfspraulok sounds like a plan. let's end the undeclared gmenu2x fork and bring it back inline with upstream, and try to get the things we care about upstream (speed and memory footprint mostly, as far as I'm concerned)15:14
Jay7is matchbox heavier?15:14
Jay7as I see jlime is using matchbox15:14
mthmatchbox is a window manager under X, afaik15:14
Jay7yes15:14
mthheavier than what then?15:15
Jay7do we supply console/fb apps only then?15:15
Jay7I'm just not familiar with OpenWRT-based 'firmware'15:15
mthit doesn't boot in X, but I don't know if X is optional or excluded15:16
Jay7ah, well then15:16
larschm, i wonder why gmenu2x startup is so slow on the nanonote15:17
mthhow slow is it?15:18
mthon Dingux it's actually not that bad imo15:18
larscwell it needs 15 seconds or so15:18
mthhuh?15:18
wolfspraul15 seconds? it's not that bad here.15:19
mthI don't have a working rootfs right now, but I think it was about 1s on Dingux15:19
kyaki declare it the "Saturday of optimisations" :)15:19
mththe entire boot sequence of Dingux is below 15 seconds15:19
wolfspraulit's also very slow, maybe 3-4 seconds (hard to say when exactly gmenu2x starts)15:19
larschm, so it's the openwrt init scripts whats so slow?15:20
kyakopenwrt initscript take around 6.6 s15:22
wolfspraulI see in dmesg 4.4 seconds until "please wait while openwrt loads"15:23
kyaki recently measured15:23
kyak*init scripts15:23
wolfspraulI have about 3 seconds from pressing power-on to hand-over to Linux kernel, there's a lot of u-boot slowness there15:23
wolfspraulthen 4.4 until "please wait while openwrt loads"15:23
wolfspraulthen the 6.6 kyak just mentioned sound about right, until the blinking cursor goes away and the screen goes totally dark15:24
wolfspraulI assume that's when (or soon thereafter) gmenu2x takes over15:24
wolfsprauland another 4 seconds or so until all the icons show up in gmenu2x15:24
kyakhow i measured: mv /etc/rc.d/S* somewhere.15:24
wolfspraulso that's a total of 3+4.4+6.6+4=18 - maybe the part in the middle is another 1-2 seconds slower here15:25
kyakand compared with and without these scripts15:25
wolfspraulsince lars already vaporized the 4.4 into 1 or so, the next big ones would be openwrt scripts, and gmenu2x15:25
larscit will be more like 2.515:26
wolfspraulok I exaggerated :-)15:26
wolfspraulgmenu2x now feels like needing a solid 3-4 seconds, it should be < 1 second for sure.15:27
larsci've finished backporting the patches and flashing the lastest rootfs now. so soon we'll know15:27
wolfspraulanother thing is u-boot, but that can wait until xiangfu manages to get the patchset upstream15:27
wolfspraulthen the openwrt scripts, although I'm not sure how easy it is to bring that down without diverting too much from backfire15:27
kyakmaybe the whole iniscripts can be re-written, as they are generic15:28
wolfspraulthe proprietary dictionary software needs 4-5 seconds from pressing the power-on button to fully functional device15:28
kyakdo they need to hold the button?15:29
wolfspraulno15:29
larscwell there will be new stable openwrt release in may. maybe it's time to upgrade then as well15:29
kyakcool15:29
wolfsprauljust press briefly, by the time your finger goes up the screen is already lighting up15:29
larscwolfspraul: so why do we have to press the button longer?15:30
kyakit gives us a reference point, and pretty requiring one...15:30
wolfspraulI'm not sure we do. u-boot is just slow, and LCM initialization (or rather displaying something on the lcm) is not done at the very beginning.15:30
wolfspraulbecause there is no visual feedback, one doesn't know whether the button was actually pressed. so I also tend to keep my finger there (pressed) until I see the screen going on.15:31
kyaki'm trying to hold for ~1 sec - no effect15:31
wolfspraulwhereas if the lcm would immediately show some feedback, people would press the button much shorter, and pull back as soon as the lcm reaction shows up15:31
kyakeven 2 sec - no effect15:31
larscsame here15:32
wolfspraulhmm15:32
wolfspraulsame here :-)15:32
wolfspraulso much for my theory15:32
kyakso really you _can't_ release your finger untill you see the screen lights up15:32
wolfspraulI haven't had the old software on a device in ages.15:32
wolfspraulwell I don't have one here now or can reflash it back, but unless my memory totally fails me, that's how it is with the proprietary software.15:33
kyakwolfspraul: where can the prop. sw can do downloaded and how to flash it?15:33
wolfspraulno idea, I doubt you can find it anywhere.15:33
wolfspraulI'm not even sure I have it. also not sure whether our reflashing tools could flash it easily, because of nand chip differences, page size, etc.15:34
larscit probably wouldn't boot on the nanonote15:34
wolfspraulthen the lcm differences15:34
kyakah, ok.. i forgot for a moment that these are different devices15:34
wolfspraulI will ask Xiangfu tomorrow, he may have used the proprietary software the most.15:36
wolfspraulI'm pretty sure you don't need to press the power-on button for more than a second there.15:36
wolfspraulbasically you just solidly press it, and when you lift up your finger the screen goes on already.15:36
wolfspraulthe power supply circuit is the same as on Ben15:36
wolfspraulstrange15:37
wolfspraulwell my memory may be wrong too, maybe that's the explanation.15:37
wolfspraullarsc: didn't know about the next openwrt release in may. that's great!15:37
wolfspraulwhen you say "it's time to upgrade then as well" - what do you mean? upgrade what?15:38
wolfspraulif we did our homework well, I hope we can forward-port whatever patches or packages we have easily.15:38
wolfspraulin fact, if the May timeline is really a hard date, maybe we can do that even now already, and have an NanoNote OpenWrt image on the same day the upstream release comes out?15:39
wolfspraulsounds a bit risky to me, maybe the date will slip, or there will be several -rc releases first15:39
wolfspraulso maybe we do another backfire-based release first, and then start working on a trunk-based one15:40
kyakwe tried to follow the upstream once, and it lead to problems with NPTL :)15:40
wolfspraulwith 'follow upstream' you mean trunk?15:40
kyakyes15:40
wolfspraulthat's not what I mean15:40
wolfspraulwe learnt our lesson there15:41
wolfspraulwe just released an image, 02-2315:41
wolfspraul02-23 is backfire based, in my understanding15:41
kyakand the binary compatibility was broken, too15:41
wpwrakmaybe u-boot checks that the power button is pressed and shuts down again if it isn't. similar to the "debouncing" we had in openmoko15:41
kyakat that time15:41
wolfspraulmay is 3 months out, plus some slippage15:41
wolfspraulso it's too early now to switch to trunk15:41
wolfspraulbut maybe another backfire-based release, let's say April?15:42
wolfspraulor even if late March15:42
wolfsprauland then, knowing that a new openwrt release is going to happen, start moving to trunk, with the goal being to release a ben image asap after the next openwrt release comes out, ideally on the same day or just a few days later15:42
wolfspraulmeanwhile, back to bootup speed, if the openwrt scripts are 'slow' and we feel they can be optimized, what's the best way to go about that?15:43
wolfspraulwpwrak: interesting idea [debouncing]15:44
wolfspraulu-boot may do this kind of thing?15:44
wpwrakwolfspraul: it can be made to do it. in openmoko, we did this in the board-specific init15:45
wolfspraulif someone has a scope, we could find out when power comes up to the cpu15:45
wpwrakwolfspraul: so when you hit POWER only quickly, the gta0x would turn on, load u-boot, and then power off again. you'd see it all on the serial console.15:46
wolfspraulwhat was the point of that?15:46
wolfspraulso it wouldn't turn on accidentally in your pocket?15:47
wpwrakyup15:47
kristianpaulwpwrak: how we can debug that theory on the nanonote?15:48
kristianpaulwatch serial console as when pressing power button?15:48
wolfspraulbest is check with scope when cpu is powered, I think15:48
wpwrakkristianpaul: hmm, i guess wpwrak could plug a cable into idbg and have a look ...15:49
kristianpaulIf is faster for you :-)15:49
kyakCPU Temp: 101.0 C (min = 15.0 C, max = 75.0 C) -oh-oh.. shouldn't do make -j2 :)15:50
kyakit's 48 C in the normal15:50
wolfspraulhow can it go above max?15:51
wpwrakhmm .. and find a battery with a non-zero charge first ...15:51
kyakwolfspraul: the "Max" is just a recommended value, set in lm_sensors config :)15:51
wpwrakwolfspraul: (above max) ever heard of "cloud computing" ? ;-)15:52
kyaksensorsd would notify when a parameter is going out of the tube15:52
kyaksmartd[1016]: Device: /dev/sda [SAT], SMART Usage Attribute: 194 Temperature_Celsius changed from 49 to 5215:54
kyaki need to move my build host from wardrobe somewhere :)15:54
qi-bot[commit] kyak: libnsfb and tic-tac-toe: fix missing dependencies http://qi-hw.com/p/openwrt-packages/861d3a216:00
kyakmake -j2 + make distclean = missing dependencies revealed :)16:03
kristianpaulokay i got the test env (uboot + battery)16:11
kristianpaulwhats the speed?..16:11
kristianpaulhmm wait a min16:12
wpwrakah .. i almost forgot. hmm, doesn't power up yet16:14
wpwrak3.5 V battery voltage. let's charge it some more then16:15
kristianpaulnor 9600 or 115000...16:16
wpwrak57600,i think16:16
kristianpaulah16:16
kristianpaulyeap16:16
kristianpaul quick power on, dint gaveme log..16:17
kristianpaulNAND Boot is a uboot message i guess?16:25
kristianpaulwpwrak: i could not confirm you theory on my side16:25
kristianpaulnand boot is first msg just before LCM power on16:27
wpwrakstill needs more charging here :-(16:27
kristianpaulquick press of power button for 1 and 2 seconds dint genereate debug messages16:27
wpwrakcould be that the reset pulse is very long16:28
wpwraknaw, not really. simulation says ~0.1 s16:31
wpwrakand no data sheet for U6 (BL8509). well, i'd be surprised if the reset generator had any crazy timing16:33
kristianpaulwpwrak: you mean this16:54
kristianpaulwww.belling.com.cn/2010p/BL8509_en.pdf16:54
kristianpaul BL8509-263DNRM <- 2.63V detect 200mS delay17:16
Action: kristianpaul needs to learn how measure and freeze for certain voltage threshold in the scope17:17
kristianpaulwow ettus have a gpsdo-kit17:18
kristianpaulgetting ready for faking GPS signals? :D17:18
B_Lizzardlarsc, I was told that 2.6.37 is stable for the NN17:28
B_LizzardDo I need to apply the patch for the modifier keys?17:28
B_LizzardOr the patch for usbnet?17:28
larscif you want to use them, yes17:29
kristianpaulwpwrak: may be the other firmware dint shutdown, just get in to a sleep/suspend mode and thats the trick17:30
B_LizzardOK, thanks.17:33
kristianpaulwpwrak: Jz4740_ds page 3417:34
kristianpaulwpwrak: Yes, is a different state power up the ben the first time it is energized (ie when locating battery)17:44
kristianpaulthan after you shut it down by software and "power on" again17:44
qi-bot[commit] Lars-Peter Clausen: [xburst] Improve mounttime. http://qi-hw.com/p/openwrt-xburst/630a78617:45
dvdkhi17:46
kristianpaulhello mate :-)17:46
dvdkjust looking over docs to find out whether the ingenic's hw scale could be used by mplayer video-out17:46
dvdkthe schematics say jz4720.17:46
dvdkbut i only have a full datasheet for the jz472517:46
dvdkare those the same or close enough?17:46
kristianpauli think share some similarities yes17:47
dvdkFound this datasheet (maybe unofficical): http://www.gmun.unal.edu.co/cicamargoba/embebidos/Jz4725_pm.pdf17:47
dvdkthis seems to be pretty complete (qi-hardware.com doesn't seem to host a complete manual for the ic)17:47
dvdkit also documents the video scaling unit.17:48
dvdka litttle mmap(/dev/mem) seems all there is to access it.17:48
dvdki guess the video scaler is using physical memory addresses?17:49
dvdkthat will be the difficult part (from userspace anyway)17:49
kristianpaullarsc: linux is resposible of shutdown powersuply to the nannote? can you tell me more about what registers and procedures are involded on that?18:05
Action: kristianpaul misss a doc from lars about linux for nanonote integrations and remarks on the process18:05
larsckristianpaul: take a look at arch/mips/jz4740/reset.c18:07
kristianpaullarsc: I mean poweroff now, means just disable the main power of the Xburst disabling by the use of PWRON pin?18:07
kristianpaulah ok !18:07
dvdkdoes anybody of you know whether pages locked via 'mlock()' are guaranteed to never change their physical address?18:09
dvdkthen i could just mmap /dev/mem scan for the memory (using changing signatures) to get the physical address18:10
dvdk.. and try to use that for the scaler18:10
dvdk(one page at a time)18:10
larscdvdk: if you want proper support for the scaler you should probably write an video4linux driver18:13
dvdklarsc: yeah might be cleaner.  but might also mean more work :)  note how most of the Xorg stuff is in user-space as well.18:14
kristianpaullarsc: you know that a halt is diferent than a powerof? :) or think it should18:14
larsckristianpaul: yes it is18:15
kristianpaullarsc: So i can say we dont have a proper shutdown sequence in linux? And by then a "long" and usual main power delay boot?18:30
kristianpauls/shutdown/poweroff18:31
larschm?18:33
kristianpaulhe, just especulatingc ;)18:34
larscshutdown works fine here18:37
kristianpaulsure, i was thinking about the power on18:39
kyaklarsc: before your patch: 20.76s (gmenu2x), 4.51s (rootfs). After your patch: 15.90s (gmenu2x), 2.56s (mount rootfs). This is awesome! Overall boot time, including mounting of datafs is 17.10s19:15
kristianpaulkyak: And without mounting datafs?19:32
wpwrakkristianpaul: yeah, u-boot starts only with a delay. once it does start, it's good.19:44
wpwrakkristianpaul: so the mechanism is different from what we had at openmoko19:44
kyakkristianpaul: almost the same. datafs is mounted after gmenu2x start19:46
kristianpaulwpwrak: I wonder know is the delay is just normal and the other firmware handles shutdown differently19:48
kristianpaulbrainless is nice game (i love his font) !19:51
larsckyak: looks as if there is an measurable performance improvemnt for regular data access as well, thats good :)19:52
kyaklarsc: i'm writing to datafs at ~400 KB/s. As i remember, it was always like that19:57
kyak(when copying via scp)19:57
kyakkristianpaul: you might want to have a look at jfbterm for "brainless font" :)19:59
rjeffrieskyak re proposal from jirka to mark console apps w/ small "C" letter in corner of the icon and maybe in own group. YES PLEASE!20:00
kyakrjeffries: are you on the latest image already? :)20:00
rjeffriesnot a chance ;) I fail the Ben NN IQ test20:01
kyakhehe20:03
kyakso how do you think, should it be a "C" or a separate tab?20:03
rjeffrieskyak seperate tab would be easy, and less work20:04
rjeffrieswpwrak remind me, will code to communicate with UBB be in user space, or will a small driver be (also) needed?20:07
rjeffriesI assume driver...20:07
dvdkrjeffries: nope, no driver needed as long as you only need general purpose I/O20:12
rjeffriesthks dvdk your idea of a propeller based video "card" is clever20:13
dvdkkristianpaul: kyak: yeah selected jfbterm for brainless _because_ of the font20:13
dvdkneed unicode glyphs for the chesspieces, and fbterm default font was way too small20:13
rjeffriesI am in ash right now20:15
kyakdvdk: hint: fbterm -s 12 :)20:15
rjeffrieswould someone suggest a setfont command that will give me a larger font? my eyse are no longer young20:15
rjeffriess/eyse/eyes20:16
kyaksetfont /usr/share/kbd/consolefonts/ter-v12n.psf20:16
rjeffriesthanks kyak20:16
kyakor 14, 16 etc20:16
kyakdepending on how bad your eyes are :)20:16
rjeffrieskyak is it posssible that font is not in the October 2010 image?20:22
kyakit is not there20:22
kyakhm, i thought you were kidding about reflashing. Why don't you reflash?20:23
xMffkyak: jast fyi, the autotools backport is done now, just poke me if you run into troubles. I'll write a troubleshooting guide tomorrow I think20:24
xMff*just20:24
rjeffrieswell when I read the wiki it seems there is a lot to do. I will at some pont20:25
kyakxMff: ok, thanks for info. Will wait for someone to merge it into qi branch20:25
kyakxMff: are problems expected in all "PKG_FIXUP:=libtool" pacakges?20:26
kyakrjeffries: should be as easy as ./reflash_ben.sh20:27
xMffthey udnergo a full autoreconf now, the most common issues are missing macros (must be hunted down/patched out), broken macros (syntax issues, autofail api changes) and in some rare cases static libs are not built anymore which could break packaging20:27
xMffa "clean" automake package should have no issues at all20:27
kyakok. at least we will know which packages are "clean" and which are "Crappy" :)20:28
xMffin some cases people only patched generated configure scripts and/or Makefiles, when autoreconf is in effect, the corresponding .in templates must be patched (too)20:28
kyakhmm, yeah20:28
kyakxMff: seems that all patches for "configure" are non-sense now?20:29
xMffall patches for configure in a program that uses and actually needs PKG_FIXUP20:30
kyakthey will be applied before autoreconf will overrwirte configure20:30
xMffpackages without pkg fixups do not undergo autoreconf, so nothing changes20:30
kyakok, i got it20:30
xMffthere are some candidates like mysql which I haven't cracked yet20:31
xMffautoreconf fails because their autocruft is a horrible mess20:31
xMffit must be basically redone20:32
xMffwe'll see20:32
kyakcan we just not do autoreconf for such pacakges and use/patch theconfigure?20:32
xMffyes20:32
xMffthat what we do in openwrt atm for such packages20:33
xMffthe downside is that it will reveal some ugly libtool issues we tried to fix in the first place with autoreconf, like attempts to relink into /usr/lib20:33
rjeffriesok got the font changed they font names were different back in the Oct 2010 day20:33
xMff... if the projects embed an ancient/broken/experimental libtool20:34
kyakdo i understand correctly that you wanted to remove libtool fixup completely?20:34
xMffnot in the near future20:34
xMffI still want packagers give the control over whjen to fixup and when not20:35
xMffbecause the autocruft behaves so erratic that it is impossible to fix everything with a simple programmatic aproach20:35
kyakwhat does libtool fixup actually do? so far i noticed that it gets rid of "-c: command not found" thing and allows some libraries compile20:35
xMffit throws away whatever ltmain.sh, libtool, ltconfig libtool.m4, libltdl sources etc. is embedded in the package20:36
kyakhm! why didn't it work for pango then?20:36
xMffand replaces it one common, patched libtool from the opewrt env20:36
xMffbecause, the old aproach just replaced libtool resources without regenerating surrounding makefiles and configure scripts20:36
xMffthis led to shell inconsistencies - makefiles calling libtool with /bin/sh while libtool itself was built with /bin/bash20:37
kyakoh, ok! if we want to use both libtool and autoreconf fixups, how should we do? separate them by spaces?20:37
xMffadditionally, the replacement of the libtool wrapper script itself never worked (it only replaced ltmain.sh) which led to version inconsistencies within libtool20:38
xMffthe old fixup is not available anymore, however if it turns out we need it I'll reintroduce it20:38
xMffright now "autoreconf" and "libtool" fixups do the same20:38
xMffnow that backfire and trunk are in sync I will clean that up during the next week20:39
xMffits ~400 packages or so20:39
kyaklooks like a big job..20:39
xMffwell, its a sed job + some review20:39
xMffnothing that should take longer than a couple of hours20:40
xMffthe most work was done within trunk during the last months, the package situation there was actually better than in backfire20:40
xMffwe're down to 10-15 broken packages or so20:40
xMffcompared to around 100 in backfire20:40
kyakhm, i'm confused. Aren't backfire and trunk using the same feeds?20:41
xMffyes, and the same feed worked better with trunk than with backfire :)20:42
xMffnow backfire got the same fixups as trunk so it should be capable to compile everything correctly again20:42
kyakok :)20:42
xMffthis whole thing might sound more dramatic than it actually is, I do not expect spectacular fallout in the qi packages20:43
Jay7btw, what is default rows/cols on NN with OpenWRT?20:43
Jay7in console20:43
xMffbut sometimes autofail has the potential to produce errors that take hours to track down, so I'm offering my experience here20:44
kyakJay7: 53x21 with kernel font 6x11, 53x24 with setfont2 6x10 font, 53x20 with terminus-12 font :)20:44
Jay7so, 53x20 is minimal20:45
kyakxMff: thanks a lot! your support is very appreciated :)20:46
dvdkkyak: i thing fbterm also had that nasty startup time overhead, so i went for jfbterm20:46
kyakdvdk: btw, i've hunted down the startup time of fbterm. It is due to fontconfig generatign its font cache for the first time. I'm thinking to supply the cache in nanonote-files, but for this i'd need to see your md5sum /usr/share/fontconfig/cache/* :)20:47
kyakin thoery, it should be the same if our /usr/share/fonts is the same, and then this pre-generated cache can be installed...20:50
xMffkyak: oh btw, is there a qi devel list or something to send patches to?20:53
xMffor just to discussion?20:53
kyaki think it's discussion20:53
xMffokay20:54
kyakroot@ben:~# time fc-cache20:54
kyakreal1m 13.51s20:54
kyak(after removing of /usr/share/fontconfig/cache/)20:54
kyaknot everyone would wait for so long for fbterm to start for the first time...20:55
xMffmaybe all such tasks should be made a firstboot action with some progress bar or so?20:58
xMffkind of a system setup20:58
xMff(font cache etc.)20:58
kyakyeah, this would be better..20:59
kyakthere is uci-defaults/01-optimization-for-ben-nanonote21:01
kyakthe problem is, there is no console at that moment..21:02
xMffthere is but it is hidden I believe21:02
xMffbehind the bootup artwork21:03
kyakand again, someone might jsut think that his Ben is not flashed successfully if he has to wait >1min21:03
xMfftrue21:05
xMfftherefore it should happen as first action after the graphical environment launched21:06
xMffhas gemnu2x some kind of notification facility?21:06
kyaki'm not aware of any21:08
xMfftoo bad21:09
Jay7use psplash21:09
Jay7with progress bar21:10
xMffpsplash looks nice, indeed21:10
kyakxMff: the scripts get removed from /etc/uci-defaults. If i want to test it, i just put script there? (i don;t want to reflash every time to test is, of course:)21:11
xMffkyak: yes, drop a script there and reboot21:11
xMffon next boot its executed once and then deleted21:12
kyakgood21:12
kyakhm, good news21:13
kyakit won't prevent gmenu2x from starting21:13
kyakgmenu2x is called by init itself (tty1::respawn:/usr/bin/gmenu2x)21:15
kyakand init script are running in background21:15
kyakso it's safe to put fc-cache in uci-defaults21:15
qi-bot[commit] kyak: generate fontconfig cache from uci-defaults http://qi-hw.com/p/openwrt-packages/bb20bc321:20
kyakxMff: thanks for your idea :)21:21
xMffhowever now you have to deal with the possibility the the application is launched while font caching runs21:22
rjeffriesanybody living near Ilmenau, Germany ??21:27
rjeffrieshttp://www.thesycon.de/eng/company_overview.shtml21:27
rjeffriesI want to get a USB driver for Ben working on Windows (no philosphical rants required, I know everybody dislaikes the idea)21:28
rjeffriesif Ben is attached to a Windows lappie, via ssh one has a nice human interface in terms of full size keyboard and regular size scree21:29
rjeffriesI think schools might be interested in using Bens as small inexpensive way for peopel to learn a real linux environment21:30
rjeffriesand yes I know about and use VirtualBox21:30
rjeffriesBen could e.g. be used as a little dedicated web server for local use21:31
zedstarrjeffries: yeh i have pondered this idea...a lot of colleges have locked down windows machines but would be nice to connect a Ben and ssh in and do a programming class using a dynamic language directly on the Ben and take home your work in your pocket21:50
Jay7hm.. I wondered that there is no free if_cdc implementation for windows21:53
rjeffriesJay7 if you find it please let us know21:57
rjeffriesI have written to the guy at TheSyscon.de asking him what sort of deal I might arrange to act as a reseller for Ben Nano only21:58
rjeffrieszedstar where are you in geographical sense21:58
zedstarrjeffries: uk21:59
rjeffriescool which city? I love London22:00
zedstarrjeffries: im from london22:01
Jay7"Recent Linux kernels (2.6.14 and later) include experimental support for the RNDIS protocol. Since that's the only USB networking protocol built into MS-Windows, it's interesting even though it's a proprietary protocol with only incomplete public documentation."22:02
Jay7http://www.linux-usb.org/usbnet/22:02
rjeffriesI live in California USA city is Arroyo Grande, about 80 miles north of Santa Barbara22:02
zedstarrjeffries: nice....been to santa barbara22:05
Jay7http://wiki.maemo.org/USB_networking#Windows22:05
Jay7http://www.linux-usb.org/gadget/22:05
Jay7rjeffries: there are lot of info about connecting to windows22:06
rjeffriesi think the ben problem is it is client22:06
Jay7nokia n700/800/810 are clients as well22:10
rjeffriesinteresting22:15
rjeffriesI wonder if thos Nokia products required a custom driver on Windows, or simply uses RNDIS22:17
Jay7http://www.kernel.org/doc/Documentation/usb/22:17
Jay7there are pair of *.inf22:17
Jay7but seems it should be already in XP22:18
rjeffriesJay7 maybe but plugging Nanonote into Windows gives (after a whiel) a message to insert CD that came with device22:38
Jay7hm..22:39
Jay7may be you need adjust vendor/product in that linux.inf file22:39
Jay7or may be things are more complicated..22:39
kyakJay7: RNDIS is broken22:40
kyaktherefore it is completely disabled in ben to aboid windows driver confusion22:41
kyak*avoid22:41
kyakyou have to use cdc_ecm driver for windows22:41
kyakwhich is proprietary and runs only 4 hours in a row, but at least it works22:41
Jay7that demo driver?22:41
Jay7ugly22:41
kyakit was patched by some guy to run with this limitation :)22:42
kyakbut it's lost now when nanohacks.org is down22:43
kyakxMff: you are right, but still this would work in most cases. Better than before22:45
Action: kyak is off22:45
wpwrakrjeffries: (font name changes) so you reflashed ?23:15
kristianpaulhmm,how i can guarantee C reads a file as sing-mag and not one complement..23:55
wpwrakkristianpaul: by implementing the integer reading function yourself ? :)23:58
--- Sun Feb 27 201100:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!