#qi-hardware IRC log for Monday, 2010-10-11

tuxbrain_awaywolfspraul: hey01:41
tuxbrainI have read you message about someone bulid 10.20 SIE01:42
wolfspraulyou mean if someone wants to?01:42
wolfspraulwhere is that someone?01:43
tuxbrainmaybe me01:43
tuxbrainI can ask for price to two local companies (well one on Italy)01:43
wolfspraulof course, go ahead01:44
wolfspraulyou have a lot of data01:44
wolfspraulgerbers, ai files, bom, etc.01:44
wolfspraulall there01:44
wolfspraulyou can make a list of parts that are hard to get for you, and you can buy them from me, at cost01:44
wolfsprauli.e. price I paid + shipping01:44
tuxbrainI would need to know the list of the parts you already hava and cost to evaluate01:45
wolfspraulwe can also do pcb on our end, so you focus on smt01:45
wolfspraulup to you01:45
wolfspraulbut the less I have to do the better :-)01:45
wolfspraulour entire Inventory is in the wiki :-)01:46
wolfspraulone sec01:46
wolfspraullet me rename the page to Sharism_inventory01:46
tuxbrainyes I remember somthing like that, there can I take the stimated price for valid?01:47
wolfspraulI don't know01:48
wolfspraulsourcing is very messy01:48
wolfspraulnormally I would not want to start selling out of my inventory, I'm not trying to become a components dealer.01:48
wolfspraulbut in your case, if there are some parts that are hard to get, or have a high MOQ, of course I help you and just 'move them' to you at whatever cost I paid + shipping.01:49
wolfspraulbut if you have a run, at the end you will have some excess parts anyway01:49
wolfspraulmaybe you start Tuxbrain_inventory :-)01:49
wolfsprauland btw, Carlos is already making a number of changes, towards V3 I think01:50
wolfspraulso you might want to at least think about whether you want to make more V2, or the newer V3/in-progress-V3 right away01:51
wolfspraulI am actually thinking about the components/ordering process01:51
wolfspraul(that's unrelated to your proposed run)01:52
wolfspraulwhen wpwrak is back from his barbecue sleep, I'll see what he thinks01:52
wolfspraulwe should work on that process a bit, starting from the components in kicad, then generate some kind of list, then automatically query distributors/stock levels/moq/lead time from octopart.com or so, then automatically generate shopping lists for mouser/digi-key/arrow/etc.01:53
wolfspraulYanjun Luo (the milkymist jtag-serial guy) was complaining about components/bom handling in kicad, and I remember werner working on some scripts in that area as well01:54
wolfspraultuxbrain: btw, check this http://en.qi-hardware.com/wiki/File:Jtag-serial-board-on-m1.jpg01:55
wolfspraulnice picture of the jtag-serial daughterboard sitting on Milkymist One01:55
tuxbrainthank for the info, I don't think I can progress on this any ealier than next week01:56
tuxbrainbut now I know where to start :)01:56
tuxbrainnice pic!01:56
wolfspraulit's hard work01:57
wolfspraulhardware is called like that because it's hard :-) (stealing a quote from Harald)01:58
wolfspraulso don't underestimate01:58
wolfspraulbut in the long run it can be good for you, for sure01:58
tuxbrainI'm remodelating web site removing some crap like OM, Netwalker, GP2x.... and include SIE, Milkimist as comming next01:58
wolfspraulwe can work together, you make some things, I make some things01:58
tuxbrainno really I don't understimate at all01:58
tuxbrainalso adding some linux service we where also done to some costumers ,01:59
tuxbrainthat work has priority to anything els02:00
wolfspraulsure make some money!02:00
tuxbrainin paralel we have to do a course, a presentation for Hackmeeting in two weeks, host an Arduino Meeting at end of November, introduce my self in the fedora community, learn how FEL works, work on spi SDIO interface for ben, and time to time some sleep ,02:03
tuxbrainalso dayli common taks, as now, gonna prepare to ship some orders, c u, and thanks for the info.02:05
kyakwolfspraul: out of interest, what is the warranty period for Ben?03:02
wolfspraulhe, I don't know the official policy from our HK store03:03
kyakso it depends on where i bought it?03:03
wolfspraulit depends, for example Tuxbrain in the EU has to offer at least whatever minimum the EU requires03:03
kyaki see03:03
wolfsprauldo you have a problem with yours?03:03
kyakno-no, just asking :)03:03
wolfspraulso far there hasn't been a single warranty case that we would not have resolved03:03
wolfspraulkeep in mind how small our community still is03:04
wolfspraulso there are few people that are trying to take advantage etc.03:04
kyakso a defect rate is very small, that good03:04
wolfspraulohs ure03:04
wolfspraulout of the first 1000 we sold, we had to replace maybe 6-8, worldwide03:04
wolfspraulmostly screen issues (fpc)03:04
wolfspraulmost of those were 'refurbished', for example by replacing the fpc section03:05
wolfspraulsome of the refurbished ones were sold at 50% off, some were donated for specific hacking projects03:05
kyakhow do you repair? is there a repairing workshop?03:05
wolfspraulI wish.03:06
wolfspraulnot really03:06
wolfspraulalthough of course little by little we are accumulating pictures and what not in the wiki. or werner's case scans.03:06
wolfspraulbut definitely we are going in that direction03:06
wolfspraulI will stick with this form factor.03:07
wolfspraul(there may be other form factors later but this one is nice for a pocket computer)03:07
wolfspraulin terms of warranty, my actual goal is to make electronics that last 20 years03:07
wolfspraulthat is definitely not the case with the Ben03:07
kyakoh lol :)03:07
wolfspraulthe hinge will become loose03:07
kyakare you sure that the planet will still be there in 20 years? :)03:08
wolfspraulsome mechanical parts are really bad, such as the microSD connector and battery connector03:08
wolfspraulkyak: oh it's no problem03:08
wolfspraulyou just asked about warranty so explain how I see it03:08
wolfspraulcan i offer 20yr warranty for the ben - probably not03:08
wolfspraultotally sure03:08
wolfspraulso back to electronics, I think devices built for longevity (mostly mechanical longevity) go together with free software best03:08
wolfspraulfree software needs time to mature, like good cheese or wine :-)03:09
wolfspraulof course technology advances, more mhz, more memory, higher resolution screens03:09
wolfspraulthat is fine03:09
wolfspraulbut when we sell a device, yes, I think it should work for 20 years03:09
wolfspraulnot the Ben though, but with the Ben we are learning why that is not the case03:09
kyaki'd like to believe you, and i generally admire your ideas.. Though i think they are mostly far from real life03:10
wolfspraulso the Ya will be better03:10
wolfspraulkyak: how long do you think a computer should last?03:11
kyakhm... i think it _could_ last for a long time, but what's the point? the hardware will become outdated in several years anyway03:12
tuxbrainkyak define outdated , wolfgang has interesting thoutghs about it, that I start to understand :)03:14
kyak"outdated" means that that you can buy twice more powerfull device in two years for the same money03:15
kyakso what's the point of keeping the old one for years?03:15
tuxbrainbtw about patent and not allowed material in qi, I find that page very inspiring, and fedora has some lawyers to assess them so I think it will basically fine03:15
kyakhow old is your mobile phone?03:15
kyak(asking both tuxbrain and wolfspraul)03:15
tuxbrainabout 3-4 years I don't remember03:16
kyaki suppose it is not 12 years old huge brick03:16
tuxbrainI think more on hardware like a car,03:16
kyak3-5 years is a general warranty period for devices with no mechanical parts03:16
tuxbrainallways there are new carss, more bautifull , more faster, more confotable03:17
tuxbrainbut meanwhile your car woks and is useful and doesn't estar to waste money in repairs you hold it until this moment arrives03:17
tuxbrainat least on computers I work the same way03:18
tuxbrainmeanwhile I can do my job on them, they stay, if not I try to upgrade the bottleneck part if not then is time to change03:18
kyakyou are right, but computer is a bit different.. a new computer doesn't mean better ecology, safety etc03:19
wolfspraulkyak: I think you are lumping a lot of features into 'hardware'03:19
wolfspraulso of course many companies try to sell you something that they can declare outdated asap03:20
wolfspraulthat's their business model03:20
wolfspraulwhen you become their customer, don't complain (only about your own stupidity maybe)03:20
wolfspraulif you make something bigger, it's easier to declare it outdated faster03:20
wolfspraulI'm looking at all this from a free software perspective.03:21
wolfspraulmy mail client is mutt, and I don't think for email I will ever change, in my life03:21
wolfspraulthat's my feeling03:21
wolfspraulthere are not many more innovations in email. you tell me what happened in the last 10 years. spam filters?03:21
kyakthere is a time period for deprecation that makes sense. Of course i won't buy new computers every year, but i also won't stick with any device for 20 years :)03:21
wolfsprauloh you probably shouldn't03:22
wolfspraulok let me ask you this way then03:22
wolfspraulthe microSD connector on Ben03:22
wolfspraullet's say it costs 15 US cents03:22
wolfspraulaverage life expectancy 2 years03:22
wolfspraulthen there is another one, costs 80 US cents03:22
wolfspraulaverage life exepctancy 10 years03:22
wolfspraulwhich one should the manufacturer you like to buy computers from use?03:22
wolfspraulthen there is a third one, let's say 1.20 USD, which will last 20 years, on average03:23
wolfspraulyou pick...03:23
kyakwell, in this case you should declare a life time warranty, most users will pick that03:24
wolfsprauldon't understand03:25
wolfspraulwhich connector would you want your computer manufacturer to use?03:25
wolfspraul80 cents / 2 years03:25
wolfspraul15 cents / 2 years03:25
kyakof course the most reliable one03:25
wolfspraul80 cents / 10 years03:25
wolfspraul1.20 USD / 20 years03:25
wolfspraulahhh :-)03:25
wolfspraulhe he03:25
wolfspraulmaybe we mean the same thing all along03:25
wolfspraulI am not telling you you have to use 20 year old computers for some bizarre reason.03:26
kyakah ok, i see your point03:26
wolfspraulI love high-tech.03:26
wolfspraulso I look at this from the manufacturer's perspective03:26
wolfsprauland I have issues with the Ben03:26
kyak20 years warranty doesn't mean that you will be actually using it in 20 years03:26
wolfspraul1. microSD connector03:26
wolfspraul2. battery connector03:26
wolfspraul3. hinge03:26
wolfspraulI want a hinge that after it becomes loose, can be tightened again.03:26
wolfspraulthe ben's hinge cannot, it's a throw-away design.03:26
wolfspraulyou want that?03:27
wolfspraulprobably not...03:27
wolfspraulthat's a separate question, first as a manufacturer I try to make devices that will have very few failures over as long a timespan as reasonably possible.03:27
wolfspraulI think that goes along best with free software.03:27
wolfspraulother manufacturers have other goals, that go along better with their software :-)03:27
wolfspraulso for example on Milkymist One, we spend a lot of time to identify the 'right' connectors03:28
wolfspraulDMX from Neutrik, USB and Ethernet from Molex, etc.03:28
wolfspraulwe actually even did connector comparisons, and surprisingly the more expensive ones really turned out to be the better ones.03:29
kyakbut frankly speaking, don't you want that a user who already has Ben to buy Ya?03:29
wolfspraulI think the happier he is with Ben, the more likely he will buy a Ya.03:29
kyakwhy would he buy it, if he's satisfied with Ben and it is working for years?03:29
wolfsprauland the Ya has to have something new/exciting anyway.03:29
wolfspraulno I am not worried about that03:29
wolfspraulI don't think the Ben has to be crappy and fail to increase my chances to sell a Ya.03:30
kyakyes, maybe that's a bad comparison because Ya has some obvious advatanges03:30
wolfspraulif there is nothing he needs in Ya, he can continue with the Ben, and maybe comes back in the future when we add something he actually needs.03:30
wolfspraullike I said, I am building a manufacturer that goes along best with free software03:30
wolfsprauland I am not worried about free software not supporting old devices anymore03:31
wolfspraulmany other companies don't even have that option, because supporting old devices is terribly costly for them03:31
wolfspraulnot on the mechanical side, but on the software side03:31
kyakyeah, they usually just stop suppoting it03:31
wolfspraulnow, we say, OF COURSE, because they came out with crazy hacked-together vendor ports that are totally unmaintainable03:31
wolfspraulwell they cannot, they maneuvered themselves into a dead-end road, primarily on the software side03:31
kyakthis is a way to make people buy more03:32
kyakif you can endlessly update my devices sw, it will always satisfy y needs03:32
kyak*my needs03:32
wolfspraulyou mean I should use artificial software obsolescence to sell new hardware?03:33
wolfspraulwell that strategy won't work too well with free software either, because whatever trick I use to make my new software not work on the old hardware, people will just take it out, or run 100% free software on the old device in the first place03:34
kyakno, cause you are a good guy :) but this is what others do, and this is how they survive03:34
wolfspraulah no03:34
wolfspraulI don't see myself as the 'good guy'.03:34
wolfspraulI just try to explain the business model, the one of Sharism, and other manufacturers.03:34
wolfspraulyou need to understand that many other manufacturers just do not possess (!) the software skills needed to execute a strategy such as the one I hope will eventually work for Sharism.03:35
wolfspraulso the fact that they have unmaintainable vendor ports is something they don't know how to avoid03:35
wolfspraullook at the mess that is already clearly visible in the Android scene now03:35
wolfsprauldo we believe this is still fixable?03:35
wolfspraulmaybe it's already too late03:35
wolfspraulthe manufacturers are again too fast, they let software companies talk them into quickly throwing this or that device on the market, even though software will be unmaintainable later03:36
wolfspraulbut to me, 'hardware' is just a bit of mechanical, silicon, metal, etc. around free software03:37
wolfspraulso we try to dig into those various things that are 'hardware', and then build them in such a way that they perfectly fit the needs, of today and tomorrow, of free software03:37
wolfspraulmakes sense?03:37
wolfspraulI don't have these ideas in very polished sentences, sorry...03:37
wolfspraulhope I get it across03:37
kyaki understand what you are talking about03:38
kyakbut a simple example: i own a linksys router03:38
wolfspraulso in that view, I also say 'hardware' should last 20 years.why not? are electrons wearing out the wires they are running through? I doubt it.03:38
wolfspraulthe problems in longevity come from cheap mechanical, you wouldn't believe how rusty (!) many cheap Chinese connectors are even after 6 months03:39
wolfsprauland they have not been in a humidity/temperature test chamber, no no03:39
kyakthe first thing i did is throw away the manufcaturer firmware and put openwrt inside.. so now i'm greatly satisfied, and the device is working fine for years.. i don't need to buy another router03:40
kyakbut if they prevented me from updating the custome software, I would just go and buy another router when i need some feature not supported by this router03:40
kyakso by giving such great freedom to their users, they reduce their sales :)03:41
wolfspraulyou are trying to convince me with a hypothetical story that you aren't even following yourself?03:41
wolfspraulmaybe if software would be un-updatable, you would not have bought their device in the first place03:41
kyakgood point03:42
ezdagorAnyone know if there is a keymap file for the Nanonote?06:57
wolfspraulezdagor: there was lots of keymap talk on the mailing list a while back07:14
wolfspraulare you using OpenWrt/Jlime/Debian?07:15
wolfspraulezdagor: found this http://gitorious.org/mzcode/nanonotes/blobs/public/keymaps/us_bnn.kmap07:24
wolfspraulI'm just looking in the discussion mailing list archives, I don't know myself never ran Debian on my NN07:24
ezdagorAwesome. Thanks, man.07:25
wolfspraulit's from the thread here http://lists.en.qi-hardware.com/pipermail/discussion/2010-August/thread.html#480907:25
wolfspraulif nobody else comes forward in irc, definitely also try emailing discussion@lists.en.qi-hardware.com07:25
wolfspraulbest if you subscribe first, I need to moderate non-subscriber mails and sometimes it takes time, or you don't see the answer unless people remember hitting reply-all etc.07:27
wolfspraulyou can subscribe at http://lists.en.qi-hardware.com/mailman/listinfo/discussion07:27
ezdagorBleh. I am unable to save it any other way than HTML.07:30
wolfspraulhmm, too late http://gitorious.org/mzcode/nanonotes/blobs/raw/public/keymaps/us_bnn.kmap07:40
wpwrakwolfspraul: (automatic component list) well, ben-wpan/atusd has that already :) (adapted from the process we used in gta02)07:53
wpwrakwolfspraul: (bad usd connector) ah, you had bad experiences as well ? (i have one where my experiments dislocated one of the mechanical parts, so it doesn't "click" anymore. still works electrically, though)08:01
wolfspraulwpwrak: not only do I have bad experiences, but as I described to kyak earlier, I do believe that high quality mechanical and free software go together well, an ideal pair08:10
wolfspraulso I spend quite some time to dig into this, understand what is a good connector, what a bad connector08:10
wolfspraulwhat brands are actually good, which are just playing patent games and try to lock you in08:10
wolfspraulwe probably have to define some edurance tests at some point, kinda a like a 'free software mechanical quality' standard :-)08:11
wolfspraulbelow that is just crap...08:11
wolfspraulwpwrak: [component lists] ok, I start looking in ben-wpan/atusd then08:11
wolfspraulthere are many things to improve, but that one seems like a reasonable next target08:12
wolfspraulwhat do you think?08:12
qi-bot[commit] Niels: let the user select a subdirectory to mass download tiles into. Use this for tiles you want to delete after usage. Just delete the subdirectory in $HOME/Maps. http://qi-hw.com/p/nanomap/c4d718f08:52
qi-bot[commit] Niels: show download progress only during the download http://qi-hw.com/p/nanomap/4c4a96008:52
qi-bot[commit] Niels: fix layout of download widget http://qi-hw.com/p/nanomap/1d81db008:52
wpwrakwolfspraul: (bad quality) yes, agreed. i was just curious whether you had run into problems specifically with the uSD connector in the ben. this would become important when we start encouraging people to stick little pcbs in the uSD slot.09:26
wpwrak(bom) it's a relatively low-hanging fruit since much of the work is already done, yes09:26
wpwrakmy bom processor isn't perfect but it does get the job done with a moderate amount of pain09:27
wpwrakthe tricky bits are the substitution scripts and of course the component equivalences (e.g., digi-key 123-456-nd == murata xyz789)09:30
wpwrakalso, one decision the bom system, doesn't take for/force on you is whether you want to introduce a qi-hw-specific layer of part numbers, and use these to pick parts.09:31
wolfsprauljust sent another Sisvel mail to Chris09:44
wolfspraulhe will shoot back I would think :-)09:45
wolfspraulhe keeps saying that 'all devices' infringe, which is very unfortunate because he adds more FUD to the discussion09:46
wolfsprauloh well :-)09:46
wolfspraulhe got burnt anyway, so he has the right to be angry...09:46
wolfspraulwpwrak: I don't want to introduce qi-hw specific anything, if possible09:47
wolfspraulI'll just look at what you have there first.09:47
wolfspraulis the bom coming out of the kicad files?09:47
wpwrakyeah, i don't believe that interpretation either09:56
wpwrak(bom) just a sec ...09:56
wpwrakhere's the big picture: the stuff lives in ben-wpan/bom/09:57
wolfspraulPulster could do a great service, he could publish the entire documentation he has about his lawsuit.09:57
wolfspraulbut he will not, claiming all sorts of threads, confidentiality agreements, etc. etc.09:57
wpwrakthe scripts come from svn.openmoko.org/trunk/eda/boom/. there's a path to it in ben-wpan/bom/Makefile09:57
wolfspraulit's a pity, he could really help others coming full circle on this09:58
wolfsprauloh well09:58
wolfspraulwpwrak: oh I see you want to continue the good tradition of KiCad to be the application with the most extensions?10:08
wolfspraulin the ben-wpan/bom directory I see .inv .gen .equ .sub .chr .inc10:09
wolfspraulalmost every file has its own extension :-)10:09
wolfspraulI just run 'make' and see what happens :-)10:09
wolfspraulboom not found, but it did quite a few things10:10
wolfspraulhave you heard of octopart.com ? do you think they are any good?10:10
wolfsprauloh nice! boom/README!10:12
wolfspraulhmm, as usual this is seriously good work. I will take some time to read through it first.10:15
wpwrak(octoparts) hmm, more or less. i use them when i can't find a part at the regular places. but often, they don't find stuff that is easy to find and they produce hits that are difficult to use. (e.g., manufacturer or "call for more information" brokers)10:17
wolfspraultwo questions: are there standardized 'query APIs', and are there standardized 'shopping list' formats?10:23
wolfspraulI wouldn't want a digi-key only thing, we should be able to include farnell, mouser, arrow, and others if easily possible10:24
wolfspraulalso if octopart has too many sources, but a good API (?), maybe we can include some?10:24
wolfspraulsorry I mean exclude some10:24
wolfspraulif possible, it would be great if things like MOQ and lead-time could be factored in as well10:24
wolfspraulwell and stock-level10:25
wolfspraulstock-level, moq, lead-time10:25
wolfspraulwpwrak: some of the boom stuff seems to be around FIC-specific issues, that will probably never again appear in this way. true?10:33
wolfspraulfor the rest, you start with the .lst file that comes out of KiCad.10:33
wpwraki don't think there are any standard APIs. but the script that talks to the API isn't all that complex.10:33
wolfsprauldo you see any overlap in what you do outside/after KiCad and what KiCad should or might do inside of KiCad one day?10:33
wolfsprauljust trying to understand where the lines between tools/packages could be10:34
wpwrakhmm, not even sure there's anything FIC-specific in boom itself.10:34
wolfspraulis 'boom' a separate tool in and of itself, like fped10:34
wolfspraulah OK, I look at the workflow in xfig10:34
wpwrakthe inventory (.inv) does include MOQ and stock. it doesn't have lead time, though.10:35
wpwrak(boom vs. kicad) there was a bit of discussion about this. the general understanding is that BOM/inventory processing is very site-specific and kicad shouldn't try to get too deep into it10:36
wpwrakboom is separate, yes10:38
wolfspraulsure there is always a line10:44
wolfspraulif both sides (KiCad & boom) know about each other and define the line clearly, and communicate with each other, that would be perfect10:45
wolfspraulwpwrak: so you see boom similar to fped?10:47
wolfspraulin terms of packaging/independence I mean10:47
wpwrak(boom like fped) yes. fped is a bit more mature, though. boom is still somewhat experimental. also, the pattern matching syntax may been some simplification10:54
wpwrak(line) i just track whatever kicad does :)10:54
wolfspraulyes but it sounds like they are aware of boom10:54
wolfspraulthat's encouraging10:54
wolfspraulsame as with schhist, I am very interested in getting this to work on the projects server, automatically and server-side run10:54
wolfsprauldoesn't need to be tomorrow, I think the problem boom tries to solve is a bit more messy than schhist, for example10:55
wpwrakthe discussion was mainly about how to handle things in kicad and what the general requirements are.10:55
wolfspraulbut you seem to be dancing around the right problems in boom, after a quick look10:55
wolfspraulI think I will play with it a bit locally first.10:55
wolfspraullead-time is important, but we can look at missing features later. first I will try to understand what we have already, then how I can get it to run on the server10:56
wolfspraulin this case, bom stuff, I don't believe in the 'perfect' solution10:56
wolfspraulsourcing is always a bit messy, full of surprises10:56
wolfspraulso the system just needs to be flexible and robust and allow for easy modifications10:57
wpwrakyes, boom needs much more context. you need to define which parameters describe a part, you need to define how you represent this in the schematics, you need parts lists for manufacturers, you need translation tables for distributors, and you need scripts that poll the distributor's database10:57
wolfspraulbut it seems to be like that already...10:57
wpwrakand you also need some system to manage your own inventory10:57
wolfspraulok one use case I am thinking about is this:10:57
wolfspraulI commit something into git10:57
wolfspraulan electrical design10:58
wolfspraulnow I want to do a run of xx units, let's say 2010:58
wpwrak(lead time) yes. for now i simply use zero lead time = inventory. boom doesn't use parts that aren't in stock.10:58
wolfspraulideally the server can present me with some data / price list, so I can quickly send out an order to digi-key for some parts, to farnell for others, to arrow for yet others.10:58
wolfsprauland maybe source some key parts totally elsewhere10:58
wolfspraulfor me it's about making the life of someone who wants to execute a run easier10:59
wolfspraulhow can he efficiently generate and upload and order a shopping list10:59
wpwrakfor that query, you would (now) do this:  make KITS=20 again show-atusd10:59
wolfspraulnot click around websites forever10:59
wpwrakyes, definitely11:00
wolfspraulalso prices, availability and lead-times are constantly in flux11:00
wolfspraulso it needs to be able to draw data from multiple vendors11:00
wolfspraulyou think boom fits the purpose I am describing (now or in the future, just want to make sure I am looking at the right tool)11:00
wpwrakyou'd also need vendor preferences. e.g., for me, electrocomponents.com.ar would be quite convenient. for you, not that much :)11:01
wolfspraulthe process will be full of exceptions/customizations11:01
wpwrakyou can control this in boom by what vendor lists you feed to the system11:01
wolfspraulsometimes you have some parts in stock from an earlier run11:01
wolfspraulor the price drops significantly if you order 50 instead of 20, so significantly that you are better off ordering the 50 today11:02
wolfsprauland so on11:02
wpwrakbut once a vendor is included, it will be treated as equal to all others. so that's not very realistic.11:02
wpwrakboom does that kind of calculation11:02
wolfspraulwell it cannot know about planned future runs :-)11:02
wolfsprauland the size of them11:02
wolfspraulunless you tell it somehow, of course11:02
wolfspraulbut like I said I am not looking for the perfect system11:02
wolfsprauljust a starting point, then we can refine it slowly for real runs, real shopping activity11:03
wpwrakno, that not :) but if 40 x MOQ 1 price > 1 x MOQ 50 price, it will get 5011:03
wolfspraulsure but if your run is 20, but at 50 the price drops a lot, and you have another run of 80 planned later, you may want to order 50 now already11:03
wolfspraulor even 100 right away, especially if there is another price drop there11:04
wpwrakyes, i understand. that's something it doesn't know :)11:04
wpwraki think it would be useful for have some tool for interactively exploring the information/database. there are other areas as well, where the batch approach doesn't work so well, e.g., all the rewriting rules11:04
wpwrakbut i think it's safe to get started with it as it is now. you can always convert data and add features later11:05
wolfspraultotally it needs to be able to easily make manual overrides, not everywhere but at some key places11:05
wpwrak(overrides) yes. there are some overrides you can do via equivalences. e.g., atrf.equ11:06
wpwrakthere i'm substituting the expensive wuerth balun with the less expensive johanson part11:07
wolfsprauldoes it currently assumes the inventory is always there together with the parts data?11:08
wolfspraulfor me they are different11:08
wolfspraulthe server doesnt' know about inventory11:08
wolfspraulthe inventory is local to whoever wants to do a run11:09
wpwrakit needs the inventory to decide what to pick11:09
wolfspraulso you upload your inventory, then the scripts can tell you what is missing11:09
wpwrakotherwise you get things like atusd.par, with a list of all possible choices. pretty much useless for humans.11:09
wolfspraulor the scripts will always run based on an inventory assumption of 0, then you need to deduct what you have already at the end, either manually or with a separate script that is run on the 'client' (i.e. the machine of the person doing the run)11:10
wpwrak(right now, the list is still manageable. but that's just because i didn't add a lot of sources)11:10
wolfspraulsure but I think about running this on the projects server now11:10
wolfspraulwhat 'inventory' does the projects server have?11:10
wpwrakinventory assumption 0 means that it will not produce any output11:10
wolfspraulwell let me look at the whole thing first11:11
wpwraki think what the project server can do for you is to give you a "reasonable" shopping list11:11
wolfspraulyou have thought about it much more than me, you need to give me some time to catch up...11:11
wpwrakif you want to customize, the you need to run it locally11:11
wpwrakit would be possible to generate a list of alternative sources from the existing data. but that needs a different representation. e.g., with html output and when you click on a position, it shows you all the possible sources.11:13
wpwrakeven that wouldn't be perfect, though, as there may be equivalent parts. and sometimes, parts for distinct components get lumped together.11:14
wpwrake.g., if you need a 100 kOhm 1% resistor and some 100 kOhm 5% resistors, then it may be cheaper to buy everything 1% than splitting11:15
wolfspraulok but that example would only matter in higher volumes11:15
wpwrakso you get one entry in the shopping list. but if you change the suppliers, that may change as well, so the number of positions in the shopping list is variable as well.11:15
wpwraknaw, particularly for small volumes :)11:16
wolfspraulwhether a resistor has 1% or 5% accuracy?11:16
wpwrakif MOQ for a resistor is 1 at one place and 10 at the other, that may already make a difference. of course, it's cents.11:16
wolfspraulhe, that's what I mean11:16
wolfspraulfractions of cents11:16
wolfspraulok anyway, boom should try to deal with those cases, even to document them is good11:17
wolfspraulit's not infinite after all11:17
wolfspraulthat's why I like boom, it tries to bring order into this mess11:17
wpwrakthere a bit the problem of computational complexity. also, perl isn't the most efficient language for such things.11:17
wolfspraulk great I think I got it somewhat11:18
wpwrakbut so far, it's holding up well, also for a larger project like gta02-core.11:18
wolfspraullet me read through the scripts and think a bit11:18
wpwrakhave fun ! :)11:18
wolfspraulI want to get this up and running on the qi projects server, for sure11:18
wolfspraulone-click shopping list11:18
wolfspraulmaybe you also need to tell it your income level11:18
wolfspraulif you are rich, just order everything from digi-key, done11:19
wpwraksounds good. also, people often wonder what the material cost of some project is. that would provide a quick answer.11:19
wolfspraulmaybe you need to give it your self-declared hourly salary11:19
wolfspraulah yes11:19
wpwrakdigi-key aren't that expensive ...11:19
wolfspraulbut frankly, if I get that type of 'bom' question I already know it's not going anywhere11:19
wolfspraulthe people that are asking these questions have a strange mindset11:19
wpwrakmany things there are quite okay.11:19
wolfspraulwell you know what I mean11:20
wolfspraulI just tend to say 500 USD nowadays, to make them go away.11:20
wpwrakwell, it's an indicator. like "are we talking about a 1 USD or a 1000 USD part"11:20
wolfspraulhardware is hardware, every 1k has new surprises11:20
wolfspraulsure, but you know what I mean11:20
wpwrakyeah. it's not the price they'll see in the shop :)11:21
wolfspraul'bom' seems to be something that even people that should have never heard about this term have heard about, somehow11:21
wolfsprauland then it gets nasty11:21
wpwrakthe bom is okay for those who actually do want to DIY. of course, few do, many talk :)11:21
wolfspraulwhat I want is probably mostly an ordering system11:23
wolfspraulso someone looks at project X, and decides he wants to order the parts for a run of 2011:23
wolfspraulthe server should make it possible for him to actually send out those orders within let's say 4 hours, and not paying totally too much11:24
wolfspraulin other words comparing a few distributors, making a few manual overrides for stuff he found who knows where, etc.11:24
wolfspraulso the scripts should help to bring down the time it takes to issue actual orders (credit card in hand), while at the same time not letting costs go through the roof11:25
wolfspraulthat's how I look at it11:25
wolfspraulthen we can slowly improve it...11:25
wpwrakthe algorithm should probably have some estimate of the S&H cost per distributor.11:25
wpwrakof course, that gets complicated on the computational side. i'm not sure brute force goes a long way with this. (brute force would be to try all 2^#distributors-that-have-at-least-one-non-unique-part-we-need combinations)11:27
wolfspraulmaybe an interactive javascript would also not be bad11:29
wolfspraulyou just click around your multi-source order form until you are happy, then press the big BUY button11:29
wolfspraullogging out for today, n811:31
qi-bot[commit] kyak: gcc: strip debug, install cleanly, both static and share linking works http://qi-hw.com/p/openwrt-packages/3e4e58712:20
kristianpaulwpwrak: what is your paid_work? (i think i never asked before)15:18
kristianpauli wonder if is related with the CNC machine and 3D scanner?..15:18
qbjectkristianpaul: might have something to do with this? http://en.wikipedia.org/wiki/Openmoko#History15:21
wpwrakkristianpaul: heh, at the moment, qi-hw is my full-time hobby ;-)15:22
qbjectwpwrak: and we love you for it.15:22
kristianpaulyeah i had learn lot from your chats and code :)16:03
kristianpauli hoep do more in the future16:03
kristianpaulin openmoko times i just was learning about microcontrollers so i missed all their history16:17
kristianpaulALL of then, really16:17
wpwrakah, you missed a good one then. not sure how it looked from the outside, though.16:24
Action: kristianpaul wonders if is the ony here with out openmoko16:24
kristianpaulheh look what i found http://people.openmoko.org/werner/bootmenu.jpg16:25
wpwrakkristianpaul: yup, been there, done that :) and i doing it in u-boot didn't make me like u-boot better :)16:31
kristianpaulso what you sugguest for dual booting?16:31
kristianpaulno dual booting?16:31
kristianpauliris.. ?16:31
wpwrakadapt the idea of the "qi" boot loader16:32
wpwraksomething small and specialized16:32
kristianpaulwell is getting close16:32
wpwrakiris looks like it's going in the same direction of u-boot. very complex and not very mainstream. so maintenance can easily become a problem.16:33
wpwrakif you really need maximum flexibility, use a linux system :) you could make a small recovery system that can also boot something with kexec. of course, it would be relatively large, but then everything would be standard tools.16:34
wpwraki did something like this as well, http://kboot.sourceforge.net/ (never made it past proof of concept, though)16:34
kristianpaulwhat is uboot for right now, just boot linux and setup soem devices at startup?16:35
larscon the jz4760 I more or less directly boot into linux and have a shell in one second after pushing the poweron key. something similar should be possible for the nanonote16:36
kristianpaulso you mean linux can boot it slef and do the other job?16:36
kristianpaullarsc: :O16:36
kristianpaulthats great16:36
wpwraklarsc: wow, 1 s is excellent !16:36
kristianpaulso the bootloader was moved to hardware some how?16:37
kristianpaulwhat power do we lost about hardware initialization with the jz4760?16:37
wpwraklarsc: how much of the user space comes from initramfs ?16:37
larscwpwrak: none16:37
larscbut in theory you could put a minimal openwrt with nand tools initrd + kernel in 4 MB or so16:38
kristianpaullarsc: what stops current xbusrt in the ben to wont be able to same as jz4760?16:39
larscwell, the cpu is clocked slower, so it probably will take a moment longer16:40
wpwraklarsc: the video seems choppy. (4 mb) yup, that sounds about right for a minimum system. a few years ago, i put kboot (kernel + user space) on a 1.44 MB floppy. the only way to boot my Sony C1 ;-)16:41
wpwrakif we can get linux to boot in ~1-2 s, the first-stage boot loader can be kept very very simple. doesn't even need debugging output and such. just initialize the cpu, the ram, load system from nand and go.16:43
wpwrakno need for user input to select the boot partition, no need for uSD support, no file systems, etc.16:44
larscthe bootrom loads the first 8k from the nand and executes them16:45
wpwrak8 k is plenty for this kind of work16:45
kristianpaullarsc: there is documentation aboout this "fast boot" method somewhere?16:52
larsckristianpaul: there is nothing special about it. the first 8k contain the spl loader. on the nanonote the spl loader loads u-boot and executes it. for the jz4760 I modified it to load the kernel and execute it16:54
wpwraklarsc: what fs do you use ?16:56
larscit's the only one which scales with nand size16:57
wpwraklarsc: okay, we wouldn't need scalability in this case. but a very fast root fs mount.16:58
wpwrakever looked at logfs ? it did sound promising. but i didn't follow what happened with it.16:58
larsci wanted to look at it but i couldn't find a tool which generates a rootfs17:00
larsca rootfs image17:01
larscwpwrak: do you thing it is possible to do zlib decompression in 3 or 4k?17:21
wpwraklarsc: i don't really know. but i think the code isn't all that large. needs lot of ram, but we have that.17:38
larscthat way we could support compressed uImages17:39
wpwrakyup. ideally, decompression would run in parallel with nand loading.17:40
larscnand access is right now the biggest bottleneck17:41
wpwrakwe can do nand with dma on xburst, can't we ?17:44
larscbut it's slower :/.17:46
larscor i did it wrong17:46
wpwrakthat's weird17:49
larscwell there is overhead in setting up the dma transfer17:53
larscand if the cpu is otherwise idle it might indeed be slower17:53
wpwrakhmm, can't find any description of bus priorities17:54
wpwrakif the cpu just runs off the cache, then dma should be as fast as non-dma (in a reasonable architecture)17:55
larscwell you get additional dealys from starting the dma transfers17:56
wpwrakcan you set up the dma when initiating the nand operation or do you have to wait for it to enter the data transfer phase ?17:56
wpwrakin the former case, there should be no significant delay17:57
larschm, not sure. i don't remember the details17:58
larscbut i guess i could overwrite {read,write}_page and setup the dma descriptors while the nand chip executes the command17:58
wpwrakyup :)17:59
wpwrakwhat documentation do you use for the MIPS core ? also the xburst core manual doesn't seem to have bus priorities :-(18:00
wpwrakactually, the core wouldn'18:00
wpwrakt care about it. it's all in the bus18:00
wpwrakthat would be a question for ingenic then: how to give DMA higher bus priority than the cpu ? (or give it has much priority as possible)18:01
wpwrakplan B: see if one could insert instructions into the busiest loops that let the cpu release the bus for a while do that the dma can write its buffers. hmm, gotta check if it actually has write buffers ...18:02
wpwraksigh. where's the reconfigurable SoC when we need it ? ;-) this would be so much fun if we could mess with the soc architecture.18:03
larscwell, if the cpu would be otherwise idle and dma was running in the background the cpu should go to sleep18:04
larscnot accessing the bus at all18:04
wpwrakyup. when it's really idle, yes18:05
wpwrakwith decompression, we'd want NAND DMA to have priority while the cpu uses whatever is left.18:05
larsci see18:07
wpwrakwhen nand is done, the interrupt wakes the cpu from sleep. cpu fires off the next transfer and goes to work on page N-1.18:07
wpwraki suppose you can command the cpu to go to sleep and if an interrupt is pending, it'll just stay awake. so the case where the nand finishes first wouldn't need any special-casing.18:08
wpwrakof course, the code that drives the nand would live somewhere in the "read the next byte" function of the decompressor. but proper layering is for whimps anyway ;-))18:09
wpwrakwith such a design, you wouldn't even need a real interrupt handler and things would still run nicely in parallel18:11
larschm, i think there is a flag in the dma descriptor which tells the dma engine to take the bus18:11
larscand not release it until the transfer is done18:11
larscthought that might not be what we want18:12
wpwrakyou mean DCMn.TM, block mode ?18:14
wpwrakyup, must be the one18:15
larscthe problem with nand dma is that it is a mem-to-mem transfer. so there is no fifo or whatever that gets filled up by the nand controller18:18
larscand we'll simply block the bus while the dma engine waits for the nand18:18
wpwrakso i think it should be "single mode"18:19
wpwrakDCNn.RDIL looks like the burst length. so that would be useful.18:19
wpwraki haven't found the DMA vs. CPU priority yet, though18:19
wpwrakvery often, dma is given bus priority over the cpu. so there's hope for this to be the case here as well.18:23
wpwrakthere's no mention of write buffers in the dma. this is bad.18:24
wpwraknor in the NAND controller, for that matter18:24
wpwrakhmm, Jz4740-pm section 3.5.2 says that only 4 kB are loaded by the boot loader18:25
larscno thats the internal rom?18:26
larscor maybe it was changed for the jz476018:26
larsclooks like they doubled the size for the jz476018:27
wpwrakwell, we survived with 4 KB on the S3C2442, so it'll be possible here as well :)18:27
larscif nothing else, we can always load the other 4kb by hand18:28
wpwraka test for dma behaviour would be to initiate a transfer, let the cpu sleep, and measure how long it takes from start to interrupt.18:32
wpwrakthen do the same but make the cpu hammer the memory. if they take the same amount of time, excellent. if not, things may need more investigation.18:32
Action: larsc looks forward to a future where don't have to speculate about the internals of the SoCs we use18:48
wpwrakindeed :)18:58
kristianpaulindeed :)21:21
kristianpaulwolfspraul: seems in the backlog for today there is a interesting question for ingenic at 17:0221:26
kristianpaulGMT -521:26
wolfspraulkristianpaul: what is the question, couldn't find it...21:47
kristianpaul17:02 < wpwrak> that would be a question for ingenic then: how to give DMA higher bus priority than the cpu ? (or give it has much priority as possible)21:52
kristianpaul17:03 < wpwrak> plan B: see if one could insert instructions into the busiest loops that let the cpu release the bus for a while do that the dma can write its buffers. hmm,  gotta check if it actually has write buffers ...21:52
kristianpaulit was a result of chat in wich uboot was some way kicked out byt the direct loading of linux kernel in order to get faster boot times21:58
mthwpwrak: section 4.5 of the pm says something about how long a DMA transfer will claim the bus22:04
wolfspraulyes I saw that video, nice!22:05
wolfspraullarsc: can I upload this into the qi wiki? may want to include it in the community news...22:06
larscwolfspraul: sure. but it's not really qi related. except that i borrowed the nanonote rootfs22:50
wolfspraulwell it's a nice 4760 video22:50
wolfspraulqi related in the sense of copyleft hardware? no. but qi related since we decided to use Ingenic XBurst for a few years to get us started, yes.22:51
wolfspraulwe may very well do some products with 4760.22:51
wolfspraulI doubt Milkymist can replace that chip fast enough :-)22:51
wolfspraulI need to go back kick these people (Ingenic), haven't met with them in a few months.22:52
wolfspraullarsc: thanks for permission, I'll upload it then and mention it as some 4760 related work. I think it's cool. If you want me to hold back until later or not mention it, also no problem. whatever is best for you.22:53
larscwolfspraul: it's ok. i just wanted to avoid that people might think that qi is activly working on a jz7460 based the next-gen nanonote.23:01
wolfspraulgood point, I'll mention that23:02
wolfspraulI think there are only 3 choices for the CPU in the Ya NanoNote: 4740, 4760, Milkymist (spartan-6)23:02
wolfsprauls-6 is a bit crazy, needs a lot of testing, more software, etc. maybe we can make a few boards for experiments, but I cannot imagine this to become real anytime soon.23:03
wolfspraulthat leaves 4740 or 4760.23:03
wolfspraulwe could do a lot with a 4740-based board that applies everything we learnt23:03
wolfspraulit also depends on whether we can get things like ben-wpan or gps_open_stack to work with the 4740, or whether they depend on some 4760-specific feature23:03
wolfspraulwell I think you know all this anyway, but good point if I include the video I'll mention this23:04
wpwrakwolfspraul: 4760 would be nice: 3 MMC controllers. that way, you can have two slots. also, two SPI controllers, like the 4720. (4740 only has one)23:06
wolfspraulwell the Ya could have 4720 too of course. I didn't know that 4720 has advantages over 4740 in some features.23:07
wolfspraul4760 would be nice, but it's a lot of work, well luckily larsc seems off to a good start. I will go back to Ingenic to see what's going on.23:07
wpwrak(4720 better than 4740) SPI seems to be the only one23:08
wpwrakwell, i haven't collected infos on all details. so there could be others. e.g., my table doesn't have ADC channels.23:09
larscwpwrak: are you sure?23:12
larscthat there are two spi ports?23:12
wpwraklemme check ...23:12
wpwrakhmm no. figure 1-1 (page 3/5) of jz4720_ds says there are two SPIs but the rest of the manual only talks about one23:15
larscthe block diagram mentions two spi ports. but if you check the pinout there is only one23:15
wpwrakyes. foolish me to trust in the overview diagram :)23:16
wolfspraulso both 4720 and 4740 have only one SPI?23:18
mthcompared to 4740, the 4760 has a 50% higher clock, floating point support and a coprocessor that can be used for video decoding or other pipelined operations23:20
mththe coprocessor is an xburst core with a small local RAM, similar to the Cell SPUs I think23:22
wpwrakmth: ah, so the two cores don't both have access to all the memory and peripherals ?23:23
mthwpwrak: all the info I have is a summary of the specs, but it looks like it has only a local RAM that can be DMA-ed to and from23:24
wpwrakmth: (summary) 4760_ds ?23:24
wolfspraulyeah quite a few 4760 related files leaked to baidu23:26
wpwrakyu[, that23:28
wpwrak's the one i meant23:29
larscand there is also a gpu core23:29
larscthere is no documentation though how to access either the vpu and the gpu.23:30
wpwrakso the 2nd core may not be able to do efficient i/o. that was one of the theories behind the idea of doing gps in software - that one core could be dedicated to gps decoding and (implicitly) would also do all the i/o23:30
larscand their offical kernel doesn't make use of the fpu either23:30
wpwrakdoes it even work ? ;-)23:30
wolfspraullarsc: got lost in the acronyms. there is no documentation on vpu, fpu and gpu, right?23:32
larscwell if they exits they are not part of the pm23:32
larscneither is documentation about the gps unit23:33
wolfspraulmy feeling is the gps unit will be a failed project23:33
wolfspraulthe way they work is they quickly throw together a lot of stuff23:33
wolfspraulthen they see what works, and only those features get marketed23:33
wpwraklarsc: gps unit ? where ?23:34
wolfspraulso over time the datasheets get lighter and lighter23:34
wolfspraulbasically the first one is a wishlist23:34
wolfspraulthen it slowly becomes reality, as features are removed23:34
wpwrakwolfspraul: nice approach. so much more efficient than fixing bugs :)23:35
wolfspraulI am still not 100% convinved that the 4760 will even be a 'success', i.e. the improvements over earlier chips (the ones that are actually working) are enough to convince customers to upgrade to it, or start using it23:35
wolfspraulthat question can only be answered very late in this game :-)23:35
wolfspraulin fact Lars has one of those potential customers asking him now to help get some stuff to work, that Ingenic cannot get to work (correct me if I'm wrong)23:35
rohwolfspraul if its easy enough to integrate and cheap enough not to get tackled by the earlier ones to much...23:35
rohuh. so integration isnt easy?23:36
larscwpwrak: iirc the ds mentions the gps unit as well23:36
larscwolfspraul: it's just boot time optimization23:36
wolfspraullarsc: the last ones I've seen have it removed23:37
wolfspraulI need to ask again.23:37
wolfspraulthe GPS front-end RF IC would come from a Shanghai startup23:37
wolfsprauland this is their first IC23:38
wolfspraulwell at least they hired an entire American (!) IC design team, so I was told23:38
wpwraklarsc: _ds only has "gps" in one place. and that's the kind of devices it's for, not a subsystem.23:38
wolfspraulit seems they think because the team is American, they must know how to design ICs23:38
wolfspraulin fact Ingenic is the only customer of that Shanghai startup23:38
wolfsprauland the Shanghai startup is funded by someone who made his money in real-estate and has no clue about high-tech.23:39
larscwolfspraul: i have one from sep. 16 which still mentions the gps module23:39
rohwolfspraul yummi ;)23:39
wolfspraulso there was a lot of let's say 'irritations' when the technology didn't behave like a skyscraper23:39
wolfspraullarsc: maybe it's back. I'll definitely ask.23:39
wolfspraulthe Shanghai startup is real23:40
wolfspraulall this exists23:40
wolfspraulmoney is changing hands23:40
wolfspraulbut whether there is a functioning chip in the end only god knows23:40
wolfspraulso I will ask about FPU, GPU, VPU, gps23:40
larscok. have to go to sleep now. need to be at work in 6h...23:41
wolfspraulsince I learnt more about the Shanghai startup and their first IC being this GPS front-end IC, I started to work with SiGe instead23:42
wolfspraul(on the GPS front-end RC IC)23:42
wolfspraulI will try to collect some info...23:42
rohi wonder if the the ublox guys would be ready to open their hw. (not meaning their fw for the arm7, just how to replace it with a free one)23:43
wolfspraulI highly doubt that.23:44
wolfspraulu-blox essentially is a software company23:44
rohexactly. but that way they could 'sell more chips'23:44
wolfspraulif our copyleft hardware project succeeds, 100% of what u-blox actually does will be openly documented, if copyrightable licensed under gpl or cc-by-sa23:44
wolfspraulah no, what's left for them?23:44
rohwolfspraul i doubt that any fre project would get near their results when it comes to performance soon23:45
wolfspraulthey use SiGe front-end RF ICs too btw, in some products (forgot which one)23:45
rohmeans one could develop sw on the same hw others use with their binary the rest of the time. win-win23:45
wolfspraulyes sure23:45
wolfspraulthat's why I work with SiGe, same as u-blox23:45
rohwe get documented hw, they get chips sold and the users can decide if they want to use all free or most free and some commercial code running23:45
wolfspraulkristianpaul has the evb, he's on it23:46
wolfspraulyes true, but all this is already happening23:46
wolfspraulonly that you are looking at the wrong partner if you look at u-blox23:46
wolfspraulu-blox is a proprietary software company, like Microsoft, Adobe, etc.23:46
wolfspraulyou think they would work with you?23:47
rohsure. but atleast they got proper hw. something on the level of sirfIII i wouldnt even start looking at23:47
wolfspraulI had a great meeting with them, with one of the founders and the VP of marketing there.23:47
wolfspraulwait a sec let me tell you this story23:47
rohi dunno what hw SiGe does. any pointers?23:47
wolfspraulyou know I actually try things in reality before I come to my conclusions23:47
wolfspraulone sec23:47
wolfspraulso I went there23:47
wolfspraulvery nice office at lake zurich btw23:47
wolfspraulcompany growing, stock market etc.23:48
wolfspraulit's funny, when Openmoko & u-blox started to work together, Openmoko was twice bigger (in employees) than u-blox23:48
wolfspraulafter 2 years, u-blox was 20 times bigger than Openmoko23:48
wolfspraullisted on the stock market, founders were millionaires, etc.23:48
wolfspraulI think the founder of Openmoko would have liked this exactly the other way round.23:49
wolfspraulbut back to the meeting23:49
rohheh. they had a working product already then, right?23:49
wolfspraulso it was kind of a 'let's look back at the last 2 years' meeting23:49
rohi mean.. then moko and ublox crossed paths23:49
wolfsprauland what they said made total sense to me. I asked whether the collaboration with OM met their expectations.23:49
wolfspraulthey said "YES!", and that they always saw the entire Openmoko project as 'extended documentation', and that was exactly what it turned out to be, and why it was 'nice' for them23:50
wolfspraulbrought a few customers here and there to them23:50
wolfspraulwell I thought about it, and realized - THEY ARE RIGHT :-)23:50
wolfspraulall OM created was extended documentation, that brought some random lost souls to u-blox technology23:50
rohwell.. thats what linux is when done right and in the spirit of gpl. the best sdk you can get23:51
wolfspraulthis extended documentation was very costly to create, but u-blox didn't pay, so why not...23:51
wolfspraulno come on. if you create 'extended documentation' you are lost.23:51
wolfspraulimagine the Linux kernel would not actually solve anybody's problem, but all computer science classes in the world would use it as 'extended documentation'23:51
rohthey already got quite cool stuff i i remember right.. layouts, and a really detailed documentation of their fw-api in some *sigh* windows format23:51
wolfspraulsure they believe in open API23:52
wolfspraulthey are very open minded, clear and sharp23:52
rohdocumenting hw isnt anything else. just the level of detail is different23:52
wolfspraulof course that also means they super clearly understand why they are in business, and why their customers are paying them, which is because they are holding proprietary IP, copyrights and patents.23:52
rohtheir sw would even gain because dual implementation on the same hw exposes more bugs which can be fixed in later revisions23:53
rohalso hw-bugs23:53
wolfspraulalso they foudn it a bit funny that OM was so happy with u-blox only because accidentally (their own words) their firmware was in ROM not RAM23:53
rohthats why drivers are often so hard23:53
wolfspraulI can see how that much look funny from their side.23:53
wolfspraulbut then again, hey, if someone wants to spend x million USD to write extended documentation for our proprietary technology, why not... more power to them! :-)23:53
wolfspraulroh: let's get back to positive things, SiGe23:54
rohwolfspraul well.. i knew that.. yes it is/was a blackbox. but none which shared ram or rom with the app cpu so it is/was fine by my definition of opensource23:54
wolfspraulif you look at what u-blox does in more detail, you realize they outsource all manufacturing/hardware, even RF stuff23:54
wolfspraulroh: yes they laugh at you for that.23:54
wolfspraulof course you can take that :-)23:54
wolfspraulif it provides only 1 penny advantage to them, they will move their proprietary tech from ROM to RAM23:54
rohwolfspraul my definiton of opensource in hw is different from stallmans there ;)23:55
wolfspraulso let's look at u-blox deeper23:55
rohwolfspraul as long as its seperate ram, i dont care23:55
wolfspraulthey are a software company23:55
wolfspraulthey like open APIs, of course23:55
wolfspraulfree SDKs23:55
roheven if i need to load the fw as blob (as long as i can distribute it)23:55
rohsee network chips, wifi, etc.. same there all again.23:55
wolfspraulthey won't even help you understand how they are making theri stuff, that would be pretty crazy23:55
rohbut i need the specs/documentation how to load it and the distribution rights for the fw also then (as well as its api documentation of course)23:56
wolfspraulroh: let's move to the next level http://www.sige.com23:56
wolfspraullet's call the u-blox firmware a 'driver' and the SiGe RF ICs the 'hardware' we want to write a GPL 'driver' for23:57
wolfspraula few re-definitions but I hope you follow me :-)23:57
rohmoving from a multi-cpu concept to a gps 'softmodem'23:58
rohsimilar to the way the gps in gta1 works23:58
wolfspraulsure you could describe it like that, same end result I think23:59
--- Tue Oct 12 201000:00

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