#qi-hardware IRC log for Tuesday, 2012-07-24

wpwrakhellekin: perhaps it's more on-topic on this channel, though00:06
wpwrakhellekin: one thing would be safety for sw/hw developers using the chips: they can search for bugs themselves (or find someone to do it), and in case the manufacturer folds, they have at least a chance to get someone else to make the chips for them00:08
hellekinI repost here...00:09
hellekinwpwrak: I think the best compromise would be some router stuff. They do have radio engineers, but mostly to produce RFID, so it's probably not the same as proper RF or WiFi00:09
hellekinbut in general, I'd like that they think about producing public infrastructure, so cheap mesh networking would be best00:09
hellekinas a fallback, PBX chips could be interesting. I don't see any of non-leaders building consumer nanoware anytime soon. A smartphone is very unlikely, although Brazil does have some local production (e.g., gradiente)00:09
wpwrakanother benefit would be that they may get people to help. e.g., by reviewing. or making substantial suggestions. depends a bit on how popular the chips in question would be and what segment they address.00:10
hellekinI think that having free designs is a great argument. Softcore implementation is another one.00:11
hellekinbut overall, there must be a large strategy. These people are manufacturing RFID for cattle, that's fucking boring00:11
wpwrakanything RF or smartphone in general is a patent minefield at the moment. if they can get suitable political protection, they could perhaps sell them nation-wide. but it's a big risk.00:11
hellekinbut there's a huge market00:11
wpwrak(RFID for cattle) hah ;-)00:12
hellekinI would go for public network infrastructure00:12
hellekinbut look at the perspective for the next few years: 2014 Soccer World Cup, 2016 Olympic Games00:12
hellekinall they can come up with is a U.S. imposed surveillance system with cameras all over rio de janeiro00:13
hellekinincredibly stupid00:13
wpwrak2014 is soon. you don't want to start designing chips for that now.00:14
hellekinbut we're dealing here with big geostrategical business: USA wants to keep control over latin america, and for that they need to pursue their non-conventional war program00:14
hellekinagreed wpwrak 00:14
hellekinit's a long term project. And probably one that will involve more than Brazil. Venezuela, Ecuador, Mercosur in general would be interested in having locally-based chip fabs00:15
wpwrakdoes CEITEC have a strong political orientation ? or are they just trying to do their business as good as they can ?00:15
hellekineven the USA has an interest in having a chip fab in latin america rather than China or Taiwan00:15
wpwrakhmm. whose project exactly ?00:15
hellekinCEITEC is a national company00:16
hellekina free project00:16
wpwraki think there are some chip fabs in mexico00:16
hellekinI'm going there to see whether there's an interest in funding free hardware, or helping in mass-producing it00:16
hellekinprobably some Intel ones, but AFAIK the CEITEC one is the first "Latin American" fab00:17
hellekinof importance00:17
wpwrakwhat do you mean with "free hardware" ? the sort of devices we make in qi-hw ? deeper down, like fpgas ? asics ?00:18
hellekinyes yes yes00:19
hellekinwhat's the relation between qi-hw and the freedombox foundation?00:20
hellekinI've met James Vasile and Jake Appelbaum, and Jake told me he wanted to have free hardware in the mix, but the first round would use ARM CPUs00:20
hellekinI think we cannot do free software without free hardware anymore. Within 5 years, it's a dead-end as every Internet company embraces hardware.00:21
hellekinwith the vendor lockins that come with it00:21
wpwrakno relation to freedombox. jon and i briefly talked with jake about whether milkymist could be a basis for freedombox last year (at fisl), but it didn't look like a close match00:22
wpwrak(lock-in) dunno. at the moment it's bad thanks to the smartphone and tablet rush. but that may change. e.g., google should see quite clearly by now that they can save themselves a lot of trouble by using a bit of their leverage to keep things a bit more open.00:24
hellekinwell I don't know about that. My take is that Microsoft pissed off a lot of people by releasing the Surface tablet and probably google can take advantage of it00:26
hellekinwhen I mentioned milkymist to jake he said that lekernel didn't want to hear about the freedombox because of the ARM processor. But I insisted to pursue that way both to Jake and Sebastien00:27
hellekinhopefully we can come up with something interesting. That's also why I want to talk to CEITEC. 00:28
wpwrakARM may have intellectual property issues00:28
hellekinI understand that a softcore is probably too slow for consumer apps, but for routing stuff it can be OK. And if we can turn it into an ASIC and have it mass-produced, bingo00:29
wpwrakfor softcore, you'd have to find an application that explicitly benefits from having an FPGA00:29
wpwrakthen work from there00:29
hellekinreconfigurable stuff I guess00:30
wpwrakcould be reconfiguration as part of operation (e.g., on a scale of seconds or minutes), could be reconfiguration as part of configuration (hours, days), or reconfiguration as part of in the field upgrades (months). e.g., the latter is why many big routers have fpgas.00:32
wpwrakand one point you run into tool limitations. fpgas are quite anti-open at that level.00:35
wpwrakthere are two ways out of this: 1) wait until wolfspraul figures out how to synthesize spartans, and overthrows the evil empire of secrecy. or 2) make your own fpga.00:36
hellekinremote reconfiguration of hardware can be problematic too. Depends on what and who updates00:36
wpwrakyou mean in terms of security ?00:36
wpwrakprobably the same issues as with software. i don't see a significant difference.00:36
hellekinI'm a bit short of arguments now, but I'd really like to be able to convince those guys to give it a try. 00:36
hellekinBrazil has great ambitions, and one of them is technological independence00:37
hellekinfor some reason rms makes a great difference between a circuit that's given and static, and one that can be flashed remotely00:38
wpwrakfirst you need to understand what they need :) if you just go there and tell them about your visions, they'll listen politely, maybe go "wow" from time to time, but unless they're really really terribly bored and unhappy with what they're doing right now, they'll ignore you00:38
wpwrakphew. i'd say rms is not the most useful guide in the realm of hardware :)00:38
wpwrakhis focal point is software and the gpl. flashable firmware can be seen as a way to circumvent the GPL. that's why he's against it.00:40
wpwrakof course, from a technical point of view, it's almost always the only approach that makes sense00:40
wpwrak(or, if you want, run-time loading into some RAM. but i don't think he'll like that any better.)00:41
wpwrakwhat are you looking for from CEITEC ? that they join the revolution, in one way or another ? that they take a leading role ? to give you a job ?00:43
hellekinthanks for the tip about flashable hw and GPL00:43
hellekinI think CEITEC is in an unique position: it's a national company. Their mission is different from a private corporation. There's an opportunity for them to embrace free hardware. They could be convinced by the free designs already, that's an awful lot of work saved.00:45
wpwrakwhen do you plan to meet them ?00:45
hellekinin a few days00:45
wpwrakno specific date set ?00:45
hellekinI want to make a first contact, see how it goes. No, no specific plan. I invited them to my talk at FISL. It's about "Transitioning to a cooperative civilization"00:46
wpwrak(already convinced) well, that depends on a lot of factors ;-) in general, brazil is "Free-friendly", but that doesn't mean that everything brazil does is00:46
wpwrakah, yes, i saw it. saturday. so you're planning to stay around for a few more days.00:47
hellekinI wouldn't miss the opportunity to go and meet them00:48
hellekinit's not like if I landed in Belem. It's just around the corner.00:48
wpwrakif your first contact will be on saturday, we'll probably have a few opportunities to talk about what to look for in a strategy.00:49
wpwrakbelem ?00:49
wpwrakyeah, they're just down the road from PUCR00:51
wpwrakthat road from hell ;-)00:51
kristianpaulwpwrak: yes in guadalajara00:56
kristianpaulmexico fabs^00:57
hellekinbelem is a city at the end of the Amazon river. No fab there.01:01
hellekinbut no Winter either01:01
hellekinwell, thank you for your interest, wpwrak. I hope I can be useful there. In any case I'll try to grunt at the EHSM next December :)01:02
whitequarkwpwrak: (rms and firmware) I don't quite get how loading fw into RAM/flash circumvents the GPL02:04
whitequarkif that does not involve tivoization02:04
wpwrakgrmbl. and now tiny.cc is unresponsive.07:03
wpwrakwolfspraul: see, that's precisely why such things aren't just "ugly"07:03
wpwrakby introducing a link shortener in the mix, we not only obfuscate the real location of things, but also add a dependency on a completely unrelated 3rd party and add state that needs to be maintained in parallel to what we already have07:04
DocScrutinizer05whitequark: wpwrak: I neither get RMS POV here. To me there's no obvious difference between configuring a chip by writing 32 register values to a known I2C addr and writing 250kBytes to some I2C addr for initializing that chip's RAM with "firmware"07:23
wpwrakfrom the RMS PoV, the (closed firmware) code has a close relationship with the (GPLed) driver. thus there's a conflict.07:25
DocScrutinizer05either way on my AP there's no rogue stuff running, as long as the uploader is documented and FOSS07:25
DocScrutinizer05hell, RMS then should consider the lito masks for chip's CPU core as similarly relevant for FOSS07:26
wpwrakperhaps wolfgang can elaborate on this. he's been - with sean - on that trip our guru ;-)07:27
DocScrutinizer05honestly RMS doesn't live in our universe anymore, and I'd dare to say he partially has NFC when it comes to hw07:28
DocScrutinizer05he mustn't use any computer at all as everything from GPS to printers to HDD has firmware nowadays07:29
wpwrakyou don't need to bark up my tree ;-)07:30
DocScrutinizer05for me the definition is: there's an AP that's the master of the system (usually defined by this master controlling a lot of subsystems and HID, and nothing controlling the master). As long as that master AP doesn't have non-GPL code running on it, the system is OK for me07:31
DocScrutinizer05given there's no tivoizations in boot process07:32
wpwrakthe AP may only provide an interface and the real logic is in the device07:40
DocScrutinizer05that's the usual way07:40
wpwrakthe device could even be a bus master, manipulating kernel data structures07:40
DocScrutinizer05no way07:40
wpwrakoh yes :)07:40
DocScrutinizer05I know about shared mem and bridge concepts, from my current job. It's just not matching requirement of "nothing controls my AP"07:41
DocScrutinizer05anyway there's no such thing like peripheral=busmaster, in I2c or SPI or HSI or CSI or CPI or ATA or.... you name it07:43
DocScrutinizer05not in the sense of busmuster you used 07:44
DocScrutinizer05for PCI(e) things are obviously a bit different07:44
DocScrutinizer05but it's up to APE to use MMU and stuff wisely, to not allow PCI bus masters to access anything they aren't supposed to07:45
DocScrutinizer05and btw *any* PCI card might do such rogue stuff. I.E. it's unrelated to whether there's any firmware to upload to the card, or the "virus" is mask programmed to the PCI card's controller chip07:47
DocScrutinizer05depending on how nifty they hidden the rogue code and "API" in that PCI controller, you'll never find out no matter how hard you investigate the card. Until it actually triggers on "0xDEADBEEF" written to R:MS07:49
DocScrutinizer05anyway in that respect embedded is rather sweet compared to a decent fullsize PC, as AP is controlling all the stuff like DMA, and usually there's not even any hw interface that *might* do this busmaster stuff (unless you have oneNAND shared mem for any coproc, like modem in bridge config)07:53
DocScrutinizer05so as long as your system is designed according to age old best practice to always expect worst case input on any interface, there is no threat to openness ala FOSS detectable at my side07:56
qi-bot[commit] Werner Almesberger: fisl2012/: first outline of FISL13 talk; abstract and biography (master) http://qi-hw.com/p/wernermisc/cd8064707:56
qi-bot[commit] Werner Almesberger: fisl2012/: editing (more to come) (master) http://qi-hw.com/p/wernermisc/83c089107:56
qi-bot[commit] Werner Almesberger: fisl2012/talk.tex: shorten URLs with tiny.cc (master) http://qi-hw.com/p/wernermisc/858ef9f07:56
qi-bot[commit] Werner Almesberger: fisl2012/: on-going editing (first drawings, more links, some small changes) (master) http://qi-hw.com/p/wernermisc/e342b7a07:56
qi-bot[commit] Werner Almesberger: Merge branch 'master' of projects.qi-hardware.com:wernermisc (master) http://qi-hw.com/p/wernermisc/85f5b5607:56
DocScrutinizer05of course ROM Bootloader of recent ARM systems is a PITA regarding that07:57
DocScrutinizer05as it's usually closed blob07:57
wpwrakwell, whatever :) i still have slides to prepare and to catch a plane in a few hours07:58
DocScrutinizer05and you never know if TI/whoever_built_the_chip maybe integrated a hw timer that jumps into "ring0" romcode which in turn checks for a certain trigger condition to go big brother and hijack your whole system07:59
DocScrutinizer05ooh, where you're traveling to?07:59
DocScrutinizer05deducing from CIA you're about to join something called FISL08:01
Action: DocScrutinizer05 waves08:01
wpwrakFISL13, in porto alegre. just a small hop :)08:02
wpwraki notice that we have too many tools. already at ~30 slides for a 1 hour talk. and that doesn't even include major screenshots.08:03
qi-bot[commit] Werner Almesberger: fisl2012/: on-going editing (more links, move libraries slide, add FISL12 items) (master) http://qi-hw.com/p/wernermisc/dc6128909:15
qi-bot[commit] Werner Almesberger: fisl2012/Makefile: fix printing (master) http://qi-hw.com/p/wernermisc/b81ee8a11:32
qi-bot[commit] Werner Almesberger: fisl2012/talk.tex: some more small changes (master) http://qi-hw.com/p/wernermisc/72b587f11:32
hellekinwpwrak: I'm flying from Berlin11:44
wpwrakhellekin: you have the longer trip then :) for me it's just a hop from buenos aires11:48
hellekinwpwrak: ah cool. I might go to Bs As after12:02
xiangfuwpwrak, there are  5 VCCINT pins under slx9. do I need connect 1.2v to all those 5 pins for power on the device?12:18
xiangfusame question on other GND, VCC_O_X, VCCAUX?12:19
wolfspraulin general you can probably assume the more pins you connect the more robust it is13:40
wolfspraulpins can help dissipate heat etc.13:40
wolfspraulthe current may be distributed better13:40
wolfspraulyou can start with one and see whether the chip runs at all, and if you already fried it with that test :-)13:40
wolfspraulmaybe the datasheet has more info on this?13:41
xiangfuI have tried connect one 1.2v(VCCINT) and one 2.5v(VSSAUX) and the GND.  and connect 4 jtag wires to slx9. 13:55
xiangfuwhen I run 'detect' it give me "warning: TDO seems to be stuck at 0"13:55
xiangfuI have to look into datsheet. 13:56
wolfspraulsounds like the real work starts now ;-)13:56
wolfspraulyou can try to ask azonenberg in #milkymist too, though he may be offline right now13:57
wolfspraulconnect a few more gnc and vcc is probably not a bad idea, though I'm not sure it's the source of your current problem...13:57
whitequarkDocScrutinizer06: well, there was a Motorola phone whose WiFi card had a backdoor forwarding all traffic to a node specified via a rogue ICMP packet14:21
whitequarkwith a slightly more complicated setup you could do DNS hijacking, at which point it actually becomes dangerous14:22
whitequarki.e. someone in a "trusted" network could intercept all your traffic without you noticing.14:22
whitequarkwell, arguably so14:22
Action: kristianpaul is getting ready a bom for digikey order in one click :)14:43
larschow to open a tcp connection from a c program? popen("nc.traditional ...") ...15:06
kyakseems more like a php program15:12
--- Wed Jul 25 201200:00

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