#milkymist IRC log for Monday, 2011-03-14

Fallenougn8 !00:02
Fallenousee ya !00:02
kristianpaulFallenou: wait !00:04
kristianpaulwell.00:05
kristianpaulokay, now something on my flickernoise broke..00:05
kristianpaulhttp://paste.debian.net/110599/00:05
kristianpaulanyway i can compile again..00:05
kristianpaulbut i still wonder why.00:06
Fallenouhumm it's not related to my changes00:06
FallenouI had this problem00:06
kristianpaulbut always i do make on rtems-bsp it happens00:06
FallenouI think I solved it doing some lm32-ranlib or something00:06
kristianpaulhmm00:06
Fallenoulekernel knows about it00:07
FallenouWe should document this on the wiki :x00:07
Fallenousorry have to go00:07
kristianpaulnp00:07
kristianpaulgo !00:07
kristianpaul:-)00:07
Fallenouif you did make in the bsp and then make install00:07
kristianpaulyes00:07
Fallenouyou should surely make install and mv the files from zlib libpneg and stuff00:07
Fallenoulibjpeg etc etc00:08
kristianpaulhmm00:08
Fallenoumaybe00:08
Fallenoui don't know00:08
Fallenoutry00:08
kristianpaulk00:08
kristianpauli solved before too, but not proper way tought..00:08
kristianpaulokay i recompiled zlib and linpng problem solved..00:11
Fallenouok00:11
Fallenoumaybe we should document that00:11
kristianpaulbut i do not understand why hapen this and better way to solved00:11
kristianpaulnot proper way i think, rtems should have a guideline about deal with issues related to libs00:12
kristianpauli wonder if this is the source of the problem00:13
kristianpaul"The libpng shipped with the RTEMS graphics toolkit is out of date, so we don't use it. But there is a problem with the latest vanilla libpng (1.4.4) which can't find the RTEMS zlib because all its functions are prefixed with "z_". Instead of messing with the GNU/Autohell scripts of libpng, we simply recompile zlib to do away with the prefix. We use the regular zlib. The instructions below work with zlib 1.2.5."00:13
kristianpaulanyway..00:13
kristianpaullets try this00:13
Fallenou*away sleeping*00:15
kristianpaullol00:18
kristianpaulNo rx buffers left in the pool! We are in big troubles!00:18
kristianpaul(ran nmap)00:18
kristianpaulFallenou: now i'm able to freeze to board and no overflow msg on screen ;-)00:21
kristianpaulokay that was ftp,. now ttcp just in case..00:21
kristianpaulah, yes for ftp test i was sending a 11Mb file00:25
Action: kristianpaul need a tap00:35
kristianpaulhmm or tell my swich do trafic mirror, but nah, i prefer bring the laptop ;-)00:35
CIA-37rtems-milkymist: Yann Sionneau master * rb121f8f / c/src/lib/libbsp/lm32/shared/milkymist_networking/network.c : initialize rx buffers busy flag with 0 - http://bit.ly/gGHQgl00:37
kristianpaulwhat was that??00:37
Action: kristianpaul looks at Fallenou 00:38
Fallenouforgot to say that by default the buffer is "not used"00:38
Fallenouit was ok with qemu though00:38
kristianpaulyeah, seeing diff tight now00:38
kristianpaulright*00:38
FallenouI guess this fixes nothing00:39
Fallenoubut better initialize the values00:39
Fallenouso you would say it's worse ?00:39
kristianpaulno00:39
kristianpaulactually i will like help, but no enought skill for give opinios on this topic00:40
Action: kristianpaul back to CSR bus topic00:40
Fallenouok00:40
Fallenouis it better ? no opinion ?00:41
kristianpauljeje00:41
kristianpaulno no00:41
kristianpaulintialize is always good00:41
FallenouI mean, is the driver working better now ?00:41
Fallenouor worse ?00:41
kristianpaulah00:41
Fallenouor the same00:41
kristianpaulwell..00:41
kristianpaulit crasehed board with a 12~ file..00:42
kristianpaulthats same as before00:42
Fallenouhum ok00:42
kristianpauli dint tried small files00:42
kristianpaulBUT00:42
kristianpauli dont get a feedback on screen that may be realted with the crash..00:43
Fallenoudi you try ttcp ?00:43
Fallenou+d00:43
kristianpaulyes00:43
kristianpaulis running now..00:43
Fallenouoh ok00:43
kristianpaulbut is on my home swich so i cant debug verywell..00:44
Fallenouin theory you wouldn't need to debug :p00:44
kristianpaulohhh00:44
Fallenouit should behave well00:44
kristianpaulohhh00:44
Fallenouwhat ?00:44
Fallenougood news ?00:44
kristianpaulhttp://paste.debian.net/110602/00:45
kristianpaulhttp://paste.debian.net/110603/00:45
kristianpaulnow i wonder what hapen with the other file..00:45
kristianpaulgmm00:45
Fallenouwow what a mess00:46
Fallenouthe output00:46
kristianpaulNOW you SEE ;-)00:46
Fallenouyes :p00:46
kristianpaulmy suffering :(00:46
kristianpaulthats from flterm btw00:46
Fallenouso if I understand all this mess00:46
Fallenouthe test completed ?00:46
kristianpaulat the amazing rate of 27.19 KB/sec00:47
Fallenouit's fucking slow00:47
Fallenoueven on my qemu i get better00:47
kristianpaulbut safe ? ;)00:47
Fallenoui get like 1,2 MB/sec00:47
kristianpaulhmm00:47
Fallenoureally strange00:47
FallenouI thought it would be faster on the hardware than on qemmu00:48
kristianpaulnot so00:48
Fallenouwell the hardware bus frequency is 80 MHz00:48
Fallenouand mine is a lot faster00:48
kristianpauli bet buses not behave same in qemu..00:48
Fallenouyes00:48
Fallenouttcp is behaving as before ?00:49
Fallenouor is it better ?00:49
kristianpaulno crash00:49
FallenouI mean there is no rx buffer overflow00:49
kristianpaulbut damn slow..00:49
kristianpaulrx buffer overflow00:49
kristianpaulgone..00:49
kristianpaulhahaah00:49
kristianpaulbad00:49
kristianpaul512K file by ftp to ramdisk00:49
kristianpaulcrashhh..00:49
Fallenou:(00:49
Fallenouwtf00:49
kristianpaulno erros00:49
kristianpaulthats bad00:49
Fallenouethernet is driving me crazy00:50
Fallenoulekernel thinks there may be a bug in hardware too00:50
kristianpaulyes i think same00:50
Fallenouthat could explain some of the bugs00:50
kristianpaulwith on chip ram00:50
kristianpaulbut this was working before btw..00:50
Fallenouit may be a good idea to run simulation of the minimac core00:50
kristianpauleven wit the already noticible  "rx buffer overflow"00:51
Fallenou:(00:51
FallenouI really doesn't understand00:51
Fallenoulet's see what lekernel thinks of the patch00:51
Fallenoumaybe there is something terribly wrong with it00:51
kristianpaul52294 bytes sent in 2.70 secs (18.9 kB/s)00:51
Fallenoulet's see if I can improve00:52
kristianpaulas i said ttcp work now, but no ftp..00:52
kristianpaulas before00:52
kristianpaulodd..00:52
FallenouI sent the patch to the ML   , to get feed backs00:52
kristianpaulrtems yeah !00:52
Fallenouoh do to MM ML00:52
Fallenouno*00:52
kristianpaulrtems guys could help too00:53
kristianpaulalso they claim the bug also00:53
Fallenouyeah but they would whine about the fact that we are not sync'ed with them00:53
Fallenouour git and their cvs00:53
kristianpaulhehe00:53
kristianpaultrue00:53
kristianpaul154128 bytes sent in 4.96 secs (30.4 kB/s)00:53
Fallenouwell sorry00:54
Fallenougotta go sleeping00:54
kristianpaulno no00:54
Fallenouwill try to improve00:54
kristianpaulsure !00:54
kristianpaulgo sleep00:54
Fallenou:) gn800:54
wolfspraulkristianpaul: when you guys say 'bug in hardware', you mean the Milkymist SoC, or you mean the actual physical hardware?00:56
kristianpaulyeah, system crashed with a 512K transfer..00:56
kristianpaulwolfspraul: fpga i think, internal on-chip-memory00:56
wolfspraulyes OK, but again "bug" here means something that can be improved in Verilog, or something that needs to be improved in the physical hardware?00:57
wolfspraulthat distinction is quite important to me :-)00:57
kristianpaulwolfspraul: lekernel already pointed some errata, and i noticed etherner bug was also manifesting  itself by ramdon  or after loading bitstream sometimes00:58
kristianpaulwolfspraul: verilog yes00:58
kristianpauli think00:58
kristianpauldunno what lekernel things00:58
wolfspraulok00:58
kristianpauli bet is avoid some custom memory arrangements that can trigger this00:58
wolfspraulI'm preparing for the next bigger run, with capital expenditure 20,000 - 30,000 USD00:58
wolfspraulnaturally I want to avoid creating a 25,000 USD broken art installation00:58
wolfspraulno gallery to sell it either00:59
kristianpaulas ethernet core have a small buffer the memory bug can manifest there i think00:59
wolfspraulthe milkyman, a sculpture made out of 80 m1 rc3 boards :-)00:59
kristianpaulbbl (dinner)00:59
wolfspraulkristianpaul: yes, understood. you get my point. I am most interested where we believe the bug can be fixed, or the improvement made.00:59
wolfspraulwhenever you have the feeling it reaches beyond Verilog, into physical hardware, shoot me a note right away01:00
wpwrakwolfspraul: (broken art) maybe RISD could be interested, as a backdrop for the gta03 that supposedly went there ? :)01:04
kristianpaulwpwrak: RISD?01:09
wpwrakkristianpaul: rhode island school of design01:18
kristianpaulwolfspraul: how many mm1 this run?01:27
kristianpaulfor this*01:27
wolfspraulcurrently I am planning for 8001:28
wolfspraulideally 75 as sellable result01:28
wolfspraulI may adjust this number upwards or downwards depending on what happens on the ground (=everywhere) :-)01:28
wolfspraulbut not much, because I am already sourcing, so especially if we adjust a lot upwards, we may have to source some components a second time01:29
wolfspraulif I have 75 sellable units, I feel equipped well enough for a real launch01:29
wolfspraulthat's 37,500 USD sales revenue, need to find customers for that much first - not easy01:29
wolfsprauland even if they do show up, I can react with another run01:29
wolfspraulthe size is double the rc2 run, which is a good way to grow manufacturing runs01:30
wolfspraulif I get very nervous, I will reduce the size of the run, or make two partial runs01:30
wolfspraulif some large pre-order customer shows up, I can increase it too, of course01:31
wolfspraulactually it's easy - the goal is to never run out of stock01:31
wolfspraul:-)01:32
wolfspraulwhich is the case right now, so what limits m1 today, is that we need to find additional customers that are paying money for the units we have01:32
wolfspraulnot the size of the next run01:32
wolfspraulbut I have too few units in stock (and in too customer unfriendly conditions - unassembled etc) for a real launch01:33
kristianpaulaw_: /aw02:35
kristianpaulaw_: hi02:35
kristianpaulaw_: How many or  wich kindof sd-like memory do you used for mm1 testing?02:36
kristianpaulxiangfu: morning02:37
kristianpaulxiangfu: new booting bios?02:37
xiangfukristianpaul: morning02:37
xiangfuafter flash the last bios and last flickernoise.fbiz. now MM1 boot again.02:37
kristianpaulgood02:37
xiangfumy build of flickernoise.fbiz is 3MB02:37
xiangfu(btw, I update the mupdf  to 0.8)02:38
aw_kristianpaul, you meant size? i used 2GBytes only.02:38
kristianpaulaw_: size and branf :-)02:38
kristianpaulbrand**02:38
kristianpaul2Gb okk02:38
kristianpaulxiangfu: 3Mb inclusing mupdf ?02:39
aw_brand: Transcend02:39
kristianpauljust that?02:39
kristianpaulonly*02:39
xiangfukristianpaul: yes.02:39
kristianpaulxiangfu: oh wow02:39
kristianpaulxiangfu: what you did?02:39
kristianpauljsut upgrade to 0.8?02:39
kristianpauli wonder if is also lighter when using, had you tried read some pdf?02:40
aw_kristianpaul, but I used this test program under http://www.milkymist.org/test_tool/02:40
kristianpaulaw_: yes02:40
xiangfukristianpaul: I follow the this wiki : "http://www.milkymist.org/wiki/index.php?title=Flickernoise_build_instructions"02:41
kristianpaulhmm..02:41
xiangfuin this wiki the mupdf is 0.7, I just update it to 0.802:41
kristianpauli see02:42
aw_kristianpaul, the source codes must be here: https://github.com/lekernel/m1testing02:42
kristianpaulyeah actually i put that version but feel fry to udate wiki :-)02:42
kristianpaulaw_: oh, sure02:42
xiangfuthe images file I download from "http://milkymist.org/msd/msd-dec2010.tar.bz2", flicknoise.fbi only 1M02:42
aw_but seems that the repository now was added a multi -times about dram already.02:43
kristianpaulaw_: he, just i got a 1Gb kinkstone.. dint work on mm1, bu i guess i need take a look software part first02:43
kristianpaulbtw02:43
xiangfukristianpaul: I think it's ok. the libmupdf.a is 6.6M for now. I need to reduce this size02:43
kristianpaulcan i tap a SPI just as anyone can think?02:43
kristianpaulxiangfu: ah,you building is not linked to mupdf, right?02:43
kristianpaultap = make a node for every ping.. and send it to the logic analizer..02:44
kristianpauls/ping/ping02:44
kristianpaulsorry :(02:46
xiangfukristianpaul: I type "make" under "flickernoise.git/src" from the makefile it's linked mupdf.02:47
kristianpaulxiangfu: if is from last commit, i guess i dint linked mupdf to flicernoise02:47
xiangfukristianpaul: yes. you right. my local is not last.02:54
Action: kristianpaul will update to mupdf 0.8 asap in order to have same build parameters02:55
kristianpauleventought seems this version of mupdf have better results in code size than mine...  3Mb is good02:58
kristianpaulpuf, now do sintesis again.. so many chnages something will break, but if not i got up 32 devices on the CSR bus, wich is good :-)02:59
kristianpauldammit i need edit every core that is slave of the CSR.. that means 16..03:34
kristianpaulwhat was tha coomand with ed to do massive substitutios..03:34
aw_kristianpaul, lekernel http://downloads.qi-hardware.com/hardware/milkymist_one/datasheet/VideoIn/Qi%20R49SSA-028636-F20-YYY-YQA.pdf03:43
aw_this doc is indeed wrote a 'fundamental' xtal, I wrote emails to vendor with how they are different and if it could have behavior with different oscillating voltage.03:46
aw_I'll let you know hopefully they reply me later. ;-)03:47
kristianpaul:-)03:47
roh:)03:52
kristianpaulaw_: tape & reel, they provide it already as a tape03:53
kristianpaulaw_: i remenber last run operator manually feed it03:53
kristianpaulWhy i'm thinking before crystal doc have a lot of copy and paste ;-)03:54
kristianpaulthis look better indeed03:54
kristianpaulor at least is from dec 201003:54
kristianpaulsed -i 's/csr_a\[13\:/csr_a\[14\:/g' */rtl/*v04:00
kristianpaul(for my record :-))04:00
aw_um..hopefully they were not copy & paste style. ;-) yes, last  time we used cylinder type bulk parts04:01
kristianpaul no no,you tell me :-) i have zero experience on this..04:04
kristianpaulok.. other 40~ minutes..04:41
lekernelwolfspraul: ethernet transfers work reliably enough in the BIOS, so the PCB is definitely OK09:29
wolfspraulok great, thx09:30
lekernel3MB FBIZ = after LZMA compression, before compression it's probably around 8-9MB09:32
lekernelthe problem is LZMA decompression is slow as hell and brings the boot time to around 45s, so it'd be good to debloat the code a bit (ideally, enough not to require LZMA anymore)09:50
lekernelif it's too difficult to debloat, leave it, it's not a central feature09:50
wolfspraul45 seconds, wow09:55
wolfspraulI thought it was 3-4 seconds so far09:55
wolfspraulwe definitely should try hard to keep it in that area, every second less is great09:56
wolfspraullekernel: let xiangfu_ know if he can help anywhere, he is looking for tasks :-)09:56
lekernelit's still a few seconds without PDF (and therefore without LZMA)09:56
wolfspraulsure, but in my mind a feature that increases boot time to 45 seconds simply doesn't exist09:57
lekernelbut since PDF isn't an important feature, let's not spend too much energy on that09:57
wolfspraulso either the boot time comes down, or we cannot include the feature09:57
lekernelit can be simply disabled at compile time already09:57
wolfspraulyes I know09:57
wolfspraulxiangfu_: you there?09:57
xiangfu_yes. reading ..09:57
lekernelhi xiangfu_09:58
wolfspraullekernel: as xiangfu_ is slowly coming up to speed on the platform, do you have any ideas of tasks for him?09:58
lekerneldark color theme for MTK?09:58
lekernelshouldn't be too hard :)09:58
wolfspraulmy big ones for xiangfu_ are easy updates via the gui ("update now" button, checking for newer versions via ethernet etc)09:59
lekernelyeah, that too09:59
wolfsprauland supporting adam on testing software, i.e. adding features adam needs, documenting the build process09:59
wolfspraulfixing rtems driver bugs as reported by others (or found by himself)09:59
lekernelthere's the flash subfolder FTP problem (I mailed him about that)10:00
xiangfu_lekernel: WITH_PDF. the flickernoise is 8.3M here. the boot time will increase a lot. haven't try to boot this 8.3M flickernoise.10:00
lekernelbut I suspect this has to do with a bug in the implementation of the RTEMS eval_path API, which is the most retarded programming idea I've discovered in a while10:00
xiangfu_lekernel: yes. I got that email.10:00
lekernelso... good luck :(10:00
wolfspraulxiangfu_: boot time 45 seconds is unacceptable anyway. so it either must come down, or if that's too hard we remove the feature (for now) and work on something else10:00
wolfspraulif that's an important bug, why not10:01
wolfspraulbugs need to be fixed :-)10:01
wolfspraulas long as we watch priorities, I don't care how hard it is to fix a bug10:01
lekernelxiangfu_: the flash partition is 4MB only, so you need to LZMA your 8.3M image10:02
xiangfu_wolfspraul: (boot time, pdf feature) ok. I just report my side :)10:02
wolfspraulxiangfu_: ok great, sounds like you have good practical tasks already, and you and lekernel are talking. great!10:02
lekernelthe scripts can do it10:02
lekernelbut LZMA decompression takes forever on the board10:02
wolfspraulxiangfu_: try to pick small tasks at the beginning, that will allow you to deliver real results to others in smaller incremental steps.10:02
wolfspraulxiangfu_: so what is your #1 priority right now on m1?10:03
xiangfu_wolfspraul: for me. I put the 'pdf' to #1. because the pdf task seems the easier one.10:04
xiangfu_lekernel: give me two task: http://en.qi-hardware.com/wiki/User:Xiangfu#2011-03-1310:04
xiangfu_sorry. wolfspraul ^10:04
lekernelxiangfu_: other things you can work on are 1. dark color theme (easy enough) - btw, we'll use this wallpaper: http://curlybracket.net/wp-content/uploads/2010/02/curlybracket-light-04.jpg10:05
xiangfu_lekernel: ok. got it10:06
lekernel2. different keymaps (atm it only supports the German keyboard layout)10:07
lekernel3. different languages... maybe selected at compile time10:07
wolfspraulthat curlybracket jpg is under a free license?10:08
wolfspraulthe language should be selectable at runtime, preferable. I cannot imagine that will take up a lot of space.10:08
lekernelno, but it's a bit harder to implement10:09
wolfspraulthat's what we have xiangfu_ for :-)10:09
wolfspraulxiangfu_: that's a great task too, imho10:09
xiangfu_:D10:09
lekernelwolfspraul: (jpg) no, it's cc-nc-nd, but I have the specific permission to include it on Milkymist, which is enough I think... I see it as a small extra that we distribute with the product10:10
Action: xiangfu_ have write down the task to http://en.qi-hardware.com/wiki/User:Xiangfu#2011-03-1310:13
xiangfu_dinner time. back online later10:13
wolfspraullekernel: -nc - argh. has he lifted the -nc -nd restriction for Milkymist?10:16
wolfspraulif not we should not include it10:16
lekernel-nc yes, I made it clear that we intend to sell devices with this file on the flash as default wallpaper10:19
lekernel-nd didn't ask10:19
wolfspraulok let's assume -nd is lifted as well, otherwise we remove it10:24
wolfspraulthis type of jpg is hard to modify anyway, but this -nd stuff is a sad thing. real creative people will laugh at the whole concept.10:24
lekernelit's just a wallpaper ok? :)10:24
wolfspraulnot ok. since the -nc -nd stuff exists we need to take it serious.10:25
wolfspraulI didn't write this crap.10:25
wolfspraulit's laughable to any real creative person10:25
wolfspraulthe good thing is it's easy to remove, and since he lifted -nc let's just assume -nd is lifted as well, for now.10:26
lekernelcertainly, but not everyone shares your point of view about licenses, and we have to respect those as well10:27
lekernela nc-nd good looking wallpaper is better than nothing10:27
wolfspraulyes, but I won't include it on devices I'm selling :-)10:28
lekernelplease don't do like rms bothering the BSD people with non-free software installable from the ports tree :)10:28
wolfspraulwe are on the same page that -nc -nd is an annoyance. it wastes everybody's time.10:28
wolfspraulabsolutely meaningless.10:29
Fallenouwolfspraul: how can you measure pre-orders ?10:29
wolfspraullekernel: should I email him to clarify the -nd part?10:30
wolfspraulpainful stuff10:30
wolfspraulFallenou: measure? don't understand10:30
wolfspraulI don't take pre-orders. if you want to order one you pay, then I ship :-)10:30
wolfspraullekernel: what is the other point of view about licenses you are suggesting? you are fine to include cc -nd -nd content?10:31
lekernelwolfspraul: no, I'll write her... but I don't think it'll be easy to make her totally change the license, so you can also suggest another picture :)10:32
wolfspraulI am following Wikimedia Commons guidelines, seems the easiest and best to me.10:32
wolfsprauland nobody would ever make a derivative of that jpg anyway.10:32
lekernelwolfspraul: as long as it's non-central content, I don't really care. this jpg is just a little extra that we'd distribute with the product, which isn't as free as we would like it to be, but what can we do10:32
wolfspraulhe. that sounds like no policy at all :-) "non-central content, I don't really care"10:33
wolfspraulI understand your point of course.10:34
wolfspraulI hate wasting time over -nc -nd clarifications.10:34
Fallenoulekernel: is lzma a "hardware" friendly algorithm ? can a core accelerate it ?10:41
lekernelno idea... i just copy and pasted the lzma decompression code from the linux kernel, I don't even know how it works10:42
Fallenouok10:42
Fallenoucause maybe a core could be our solution10:42
Action: Fallenou is just saying, doesn't know the algo 10:42
lekernelsounds like a lot of work for few things10:42
lekerneli'd simply leave pdf support out instead :) there are way more important things...10:42
Fallenouok fair enough10:43
Fallenou02:23 < wolfspraul> if some large pre-order customer shows up, I                 can increase it too, of course10:43
Fallenouthat's why I wonder how you can "count" pre-orders10:43
Fallenouyou have a website set-up for pre-orders ?10:43
Fallenouyou may want to have a look at http://getitmade.com/ wolfspraul10:44
FallenouI met the director of this website during oshug10:44
FallenouI am seeing it again on thursday10:45
Fallenous/it/him/10:45
Fallenoumaybe it can help you figure out how much customer we have / we can have before doing the runs10:45
wolfspraulFallenou: what exactly do you want me to do with getitmade.com ?10:46
wolfspraulno I don't have any specific website setup for preorders, and that is not necessary. when you have money, you know you can approach anybody about buying anything.10:46
wolfspraulwhat I will do is to contact specific people to explain Milkymist to them and see whether I can get them interested10:47
wolfspraulif one of them turns around and says "we are very interested, we want to get in big time and order 500 right away", that would be what I meant10:47
wolfspraulthe terms would be roughly 50% upfront, 50% before shipment10:47
wolfspraulthe normal OEM terms are 30% upfront, 70% before shipment10:48
wolfspraulsince this is a very risky/innovative product, I would ask for 50/5010:48
wolfspraulbut the main point is to find someone who really believes in the product and wants to jump in (i.e. take business risks)10:48
wolfspraulI definitely agree with lekernel about just leaving PDF out if it causes too many problems now.10:49
wolfspraulbut of course, if more people join we can work on more things in parallel ;-)10:49
Fallenouok I understand better now wolfspraul :) thanks10:50
Fallenouyou are trying to approach big buyers, not end users10:50
FallenouI didn't realize that10:50
wolfspraulend users can buy now10:51
Fallenouyes10:51
wolfspraulthe product is in stock10:51
wolfspraulso what is the problem you are trying to solve?10:51
Fallenounone I was just trying to understand10:51
Fallenouyou made it very clear now :)10:51
wolfspraulI just read about getitmade10:51
wolfspraulthey make it easy to collect pre-orders10:51
Fallenouyes10:51
wolfspraulthat sounds like a disaster strategy to me10:51
Fallenouoh really ?10:51
wolfspraulthen I believe much more in kickstarter.com10:51
Fallenouthe problem with kickstarter is that it ships only to US or something like that10:52
Fallenouyou need a US credit card or something10:52
wolfspraulI have 15+ years industry experience. I have never met one person with experience who says that pre-orders work, especially not for innovative products.10:52
wolfspraulin fact. if you see someone trying to finance anything innovative with pre-orders, you know it will all blow up.10:52
wolfspraulguaranteed. 100%.10:52
wolfspraulit has been tried many times10:52
Fallenouok I didn't know10:53
wolfspraulyou can have a few large pre-order customers, that works10:53
FallenouI have 0- experience in this domain :)10:53
wolfspraulyes OK, so I tell you mine10:53
wolfspraul(and others might disagree)10:53
wolfspraulthe problem is communication10:53
wolfspraulit's easy10:53
wolfspraulinnovative means something will not work as expected10:53
wolfspraultrue?10:53
wolfspraulsome features need to be removed10:53
wolfspraulprice need to be increased10:53
wolfspraulDELAYS!10:53
wolfspraul:-)10:53
wolfspraulthe more innovative thing you do, the higher the risk of all of those10:53
wolfspraulright?10:53
wolfspraulnow10:53
Fallenouyes10:53
wolfspraulimagine you are sitting on the money of 1000 individual pre-orders customers10:54
wolfspraulcan you imagine the complete impossibility of keeping the communication flowing with them?10:54
wolfspraulabout what strange bug showed up, what realization, what the options are, why option xyz was chosen, etc.10:54
wolfspraulthen some will cancel10:54
Fallenouwell it needs a dedicated man which maintains wiki/twitter/ and such10:54
wolfspraulsome will not agree with the judgments10:54
wolfspraulsome will only understand half10:54
Fallenoulike a community manager10:54
wolfsprauland so on10:54
Fallenoumanage the forum10:54
wolfspraulit's impossible10:54
FallenouBut I definitely see your point10:55
wolfspraulinnovative = things turn out different than originally thought10:55
wolfspraulthat is inherently incompatible with a high number of individual pre-orderers10:55
wolfspraulthe two are mutually exclusive10:55
Fallenouhum I guess you're right10:55
wolfspraulso if you do something low risk, ok fine, you can do it10:55
Fallenou11:47 < wolfspraul> innovative = things turn out different than  originally thought10:55
Fallenouthis is totally true10:55
wolfspraulbut you better be sure that nothing unexpected happens10:55
wolfspraulbecause you are sitting on this communication bomb10:55
Fallenoueven the creator of an innovative stuff does not see the real point at the begining10:55
wolfspraulyou cannot communicate effectively with those large number of pre-orderers10:55
wolfspraulif it's innovative, you can have a few large pre-orderers10:56
wolfspraulthat's fine10:56
wolfspraulsay at most 5-10 people10:56
wolfspraulevery one less will help10:56
wolfspraulthen if something unexpected happens, you can try emails, phone calls, phone conferences, etc.10:56
wolfspraulso don't know what to do with getitmade right now10:57
wolfspraulyou can register m1 there, and see how many pre-orders you can collect10:57
wolfspraultrue?10:57
Fallenoumaybe it's good when the product is totally finished and not evoluting10:57
wolfspraulyou can even take money!10:57
wolfspraul:-)10:57
wolfspraulyes of course10:57
wolfspraulbut then why pre-order at all? why not just sell?10:57
Fallenouit's to have bigger runs10:57
wolfspraulI think if you register m1 there, and start taking pre-orders (money), you will immediately see the problem.10:57
Fallenouwith less risks10:57
wolfspraulyou will be on my neck every day about the production date10:57
Fallenouahah ok :)10:58
wolfspraulbut that's not my problem then. I say "probably May", and then in May we see where we are.10:58
wolfspraulyou promised something to your pre-order customers, not me10:58
wolfspraulI know there are too many risks (unknowns) to promise a hard shipping date now.10:58
wolfspraulalright, hope this makes sense :-)10:58
wolfspraulbut thanks for pointing getitmade out to me10:59
wolfspraulvery appreciated10:59
Fallenoulet's be clear, I didn't want to upset you :)10:59
wolfspraulnot at all10:59
Fallenouok good10:59
wolfspraulI am sharing my thinking.10:59
wolfspraulI like those sites10:59
Fallenouand thanks for your point10:59
wolfspraulI just read their faq and scratching my head :-)10:59
FallenouI clearly understand your opinion now10:59
wolfspraulI don't want large number of pre-orders mixed with innovative products10:59
wolfspraulthat will blow up10:59
Fallenouok10:59
wolfspraulso I think kickstarter.com is better11:00
wolfspraulyou throw the money over to the other side, and get no hard promises back11:00
wolfspraulmore like a donation, and wish for the best11:00
Fallenouhehe11:00
wolfspraulthat's much more in line with what an innovative product needs11:00
FallenouI guess you're right11:00
Fallenouit's not for "common" end-users11:00
Fallenouit's more for "very-small" investers11:01
wolfspraulyes11:01
Fallenouwho wants to give a hand to promising project they find cool11:01
Fallenouand maybe get a product in return11:01
wolfspraulthey can develop a lot of power though, I do believe in that11:01
wolfspraulcorrect11:01
wolfspraulthat's a great model imho11:01
wolfsprauland whoever gets the money should work their ass off to deliver11:01
Fallenouyes but hard to advertise in this world11:01
wolfspraulbut if it's innovative, there are no guarantees11:02
Fallenouwhere you want something for your money11:02
wolfspraul'pre-order' is an illusion mostly11:02
wolfspraulwell, if you want to be sure, then wait until the product is in front of you11:02
wolfspraulin the store. try it there, only take the unit you hand-tested yourself.11:02
wolfspraul:-)11:02
wolfspraulsome people do that, and that's fine11:02
Fallenou=)11:03
wolfspraullekernel: this guy dizid has a few nice ones http://www.flickr.com/search/?l=commderiv&ct=6&mt=photos&adv=1&w=all&q=milkdrop&m=text11:14
wolfspraulanything you like?11:14
wolfspraulhttp://www.flickr.com/photos/dizid/61299977/11:15
lekernelnot too bad....11:17
wolfspraulok no worries we don't get stuck on this :-)11:17
wolfspraulactually, can the current m1 take screenshots?11:18
wolfspraulthat would be another nice feature to add, if it's not there yet, for xiangfu11:18
lekernelthe m1 no, but qemu can11:19
wolfspraulcool, I'll add it to my wishlist for xiangfu11:19
CIA-37flickernoise: Sebastien Bourdeauducq master * r79f88b0 / src/Makefile : Fix link order of libs (Xiangfu) - http://bit.ly/edjEX511:57
Fallenoulekernel: I think you should put the "FPGA proven" flag on your Navre AVR clone on opencores14:26
FallenouI noticed the flag is not set14:26
lekernelyeah, and document it too :)14:26
Fallenousince it works on fpga, just putting the flag can advertise it better14:27
Fallenoudocumentation take more time to just set the flag :p14:27
Fallenoulekernel: I was asked during the talk "which device class are supported by the usb host controller" ?14:28
FallenouI didn't know exactly14:28
Fallenouso I said for the moment it's just some devices, like keyboard and mice14:28
lekernelassuming I lost my #{|#~ opencores password, I'm not so sure about that14:28
Fallenoudidn't know the exact device classes14:28
lekernelah, found it14:29
Fallenou:14:30
Fallenou:)14:30
lekernelI guess I can advertise it as "stable" too, by opencores standards it probably deserves it14:30
Fallenouahah14:30
Fallenousome of your cores has been "elected/selected" as OpenCoresCertified ;)14:30
Fallenoulike HPDMC14:31
lekernelphew, it's also an old version in the opencores repository14:31
Fallenoudunno if it's a reward for you though :p14:31
Action: Fallenou puts a "OpenCoresCertified" sticker on lekernel 14:32
Fallenoutoo bad I won't be able to attend to OSHUG #9 about opencores14:32
Fallenoui could have grabbed some stickers =)14:32
lekernelI don't think opencores has stickers14:33
Fallenouoh :(14:33
Fallenouyou have to send your code to their svn ? you can't just give a git url ?14:33
Fallenouthat's annoying to have to maintain 2 repositories :o14:34
lekernel"opencores policies" ... :)14:34
Fallenouthey should implement a sort of daemon/cron that just git clones /svn co/cvs branch14:34
Fallenoufrom your main project repo14:34
lekernel"the code has to be checked in, opencores is not just a repository of links"14:35
Fallenouyeah I see14:35
Fallenouthey want to be like github :)14:35
Fallenouthey want content14:35
lekernel"we ask people to register to download, statistics are important to build credibility"14:35
Fallenouyeah that's strange too14:35
Fallenouyou have to register to access the content in read only14:35
lekernelat least github has an outstanding web design and version control system14:35
lekerneland no registration necessary :)14:35
Fallenouyep github is really cool14:36
Fallenounot opensource but cool14:36
FallenouI wonder if gitorious is of the same quality14:37
lekernelwell, everyone's got bills to pay14:37
lekerneland quite frankly I prefer Github's policy than Opencores'14:37
lekernelgithub makes money by selling a proprietary web service to companies14:37
lekerneland supports open source software with a free of charge and outstanding service14:38
lekernelopencores makes money by being control freaks and talking to advertisers and (probably) investors14:38
Fallenouwell anyway14:39
Fallenouthe service is clearly not the same quality14:39
lekerneland maybe by using opencores to heavily promote ORSoC's consulting services - though I doubt any serious ASIC company would hire them for HDL developments14:40
lekernelthey seem to do a decent job with embedded software though14:40
FallenouI really don't know what they do14:41
Fallenouam on their website14:41
lekerneloh, and Du talar och skriver Svenska och Engelska flytande :)14:42
Fallenouoh I didn't know they have a shop : http://opencores.com/shop,items14:42
Fallenouwhaaaat14:43
Fallenoudon't speak this language :p14:43
Fallenouwow their jtag debugguer is 100 EUR14:43
kristianpaulflash subfolder FTP problems ??15:20
kristianpaulproblem = crash ? :-)15:22
lekernelno, it just won't create the file at all15:47
kristianpaulhmm isee15:58
kristianpaulah, yes16:07
kristianpauli think i experienced that yday when trying to upload the pacthes to individuual folders16:08
kristianpaullekernel: do you have a milkymist slides for VJs?17:34
lekernelnyet17:34
kristianpaularg..17:35
kristianpaulok17:35
CIA-37antares: Sebastien Bourdeauducq master * r3254b68 / (6 files in 3 dirs): pack: transform skeleton - http://bit.ly/fAqYDR18:18
Fallenoulekernel: do you have comments about the change in ethernet driver19:22
Fallenou?19:22
Fallenouit seems not to be better :(19:23
Fallenoumaybe there si something terribly wrong in the new version that you could spot with having a quick look19:23
CIA-37antares: Sebastien Bourdeauducq master * r065e5fc / (3 files in 3 dirs): pack: replace IBUF and OBUF - http://bit.ly/fKPtye23:20
--- Tue Mar 15 201100:00

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