#qi-hardware IRC log for Monday, 2013-10-14

wpwrakyeah, it got a bit longer than i thought00:00
wpwrakalready split off the mechanical part. that's for a later post then. gotta stretch the news :)00:00
DocScrutinizer05so what do you want me to do?00:04
wpwraknaw, just keeping you in the loop, in case you're interested00:04
DocScrutinizer05sure I'm interested00:04
DocScrutinizer05interested to help :-)00:04
wpwrakkewl :) so far, things seem under control. no weird surprises. well, nothing excessively weird :)00:05
DocScrutinizer05I for sure won't need such a password vault, but I could figure using one and what I would want it to work like and look like when I would do00:07
wpwrakah, so you have good memory :)00:07
DocScrutinizer05e.g. ideal size in my vision is credit card, but 2mm high00:08
DocScrutinizer05nah, I have very few RL occasions where password is needed00:08
DocScrutinizer05maybe only 10 or 15, and those few PWs I can memorize00:09
wpwrakcredit card gets tricky when you want to have USB and such :)00:09
DocScrutinizer05yep, of course00:09
wpwrakphew. 10-15 ... good memory :)00:09
DocScrutinizer05I don't consider that a usecase, since I only use my own PC and on that I got a sw vault00:10
DocScrutinizer05but I see the usecase, though I haven'z completely groked it how it works00:10
DocScrutinizer05I'm missing the user manual describing how to use that thingie00:11
DocScrutinizer05prlly I missed to read some ML threads00:11
wpwrakoh, you basically "dial" a PIN to unlock the device. then it shows you what "accounts" you have (account = short name plus additional data). dial the account, then press the button to show it. there, you see things like username and password. that's the basic operation.00:14
DocScrutinizer05generally I think the thing is missing some input interface(s), like optical to read barcodes (OCR? :-o) form PC screen, and a way to place it between kbd and PC like a keylogger to do exactly that: log keystrokes when you enter credentials00:14
wpwrakthen you enter your login information into whatever (browser, atm, secure door lock, ...)00:14
wpwrakyes, that's a more sophisticated use model. hence USB and RF ;-)00:15
wpwrakif you have a "keyboard out" interface, you can let the pw safe "type" things for you00:15
DocScrutinizer05yes, but how to train/teach/configure them?00:16
wpwrakyou can also use it as keyboard pass-through, also to make it easier to edit things in the pw safe. (the wheel should be nice for selection and reading, but not so great for writing a lot)00:16
DocScrutinizer05that's my point00:17
wpwrakoh, you could "dial" things. or type them in via an external USB keyboard. or let the PC send them over USB. or have some external entity provide the account records on the memory card.00:17
DocScrutinizer05how would that kbd pass-thru work?00:17
wpwrakso there are a number of possibilities00:17
DocScrutinizer05ideally the thing had a usb-f and usb-m00:18
wpwrakplug keyboard into USB, put USB dongle into PC, type on kbd. when you want the kbd to be "intercepted" by the pw safe, you'd tell it (via the wheel or maybe some hotkey)00:18
DocScrutinizer05using RF for one of the links always gets you into pwersupply hassle00:19
wpwrakyeah, i had that in the original design. gets a bit bulky, though. so now i'm down from USB A + USB B to USB AB00:19
wpwraknaw, we'd be externally powered in this case00:20
DocScrutinizer05HEY, use one of those crappy "augmented" USB plugs, that comes with 10pin and could go to a Y-cable that gives you A and B mode concurrently?00:20
wpwrakwell, it's called a y-box .. :)00:21
DocScrutinizer05while still keeping A and B compatibility for "normal" USB plugs00:21
wpwraknot the most elegant solution but something that can be made without too much hassle00:22
DocScrutinizer05mini(micro)-AB-extended-USB receptacle, OTG on the normal 5 pins, plus a discrete host on the extended 5 pins00:22
wpwraknaw, no bastard connectors :)00:24
DocScrutinizer05even smarter than my original idea to have discrete usb-A and usb-B in device, since you can unplug with only one plug. And when you use such a "box" you could evel place a (electronic?) switch into the box to bridge input and output when the keylogger is unplugged00:24
wpwraka more common use case would be simply editing records on the PC and sending them to the safe. if your PC has the same "security classification" as the data it's processing, there's no loss of security.00:25
wpwraki.e., if on a PC you use for the web and you just created a lolcats.com account, you'd create the account entry there, encrypt it, encrypt the key with the safe's public key and send it to the safe. the safe would then just add it to the database00:27
DocScrutinizer05let's think about e.g. my online bank account. When dongle in keylogger mode, it could detect my typing in my account number and then <TAB> and thus the dongle knows it now needs to provide the online banking password00:27
wpwraki see the box mainly as wiring. plus a lone power LED :)00:27
wpwrakthat could get dangerous. i'm more thinking of the browser on the PC telling the safe (via HID) that it now needs to log into X, and then the safe looks this up00:29
DocScrutinizer05hmm, not dongerous when you need to confirm release/injection of the password00:33
DocScrutinizer05btw HID, yes exactly, RF should be BT00:33
DocScrutinizer05BT HID00:34
wpwrakwell, 802.15.4 HID :)00:34
wpwrakBT is tricky00:34
DocScrutinizer05BT is common00:38
DocScrutinizer05and encrypted ;-)00:38
DocScrutinizer05and has drivers for HID on every PC00:38
wpwrakyes, BT would be very nice to have00:39
wpwraknow, do you know a small low-cost BT module with full documentation, available in small quantities, with SPI or similar ?00:39
DocScrutinizer05there should be a few of them00:40
wpwraktheir cheapest module is EUR 18 @ 10 units. and there's no technical info about it.00:46
wpwrakah, there's one at 16.60. that's better.00:46
wpwrakagain, no technical info00:47
wpwrakprobably fairly large, all of them00:48
DocScrutinizer05yeah, I also wondered why they are so large00:49
wpwrakthey have one small one. no price info. also, i'm not sure if the technical info really is enough to talk to it. beyond serial, i mean.00:49
DocScrutinizer05seems they often disclose the tech specs under "NDA" only00:50
DocScrutinizer05e.g. for the Option GW601 GSM module00:51
DocScrutinizer05known old annoying issue with all that stuff, with basically all manufs00:51
DocScrutinizer05It seems to me like WLAN+BT modules are generally smaller than the BT-only ones .oO(???)00:53
Action: DocScrutinizer05 looks at an arbitrary BT dongle, then recalls the BT-microdongles that hardly are noticeable when plugged into a USB-A receptacle since they and flush with the USB device's case00:55
wpwrakyes, NDA is one of my main reasons for staying away from BT. it's simply not compatible with the concept of open hardware.00:55
DocScrutinizer05sure, USB, bit hey...00:55
DocScrutinizer05wpwrak: it's always a question of disclosing the "ABI", you never get "full" specs since those even would need to consist of chip masks and litho process parameters etc00:56
DocScrutinizer05when you get enough of the interface specs and are allowed to quote them, you should be fine no matter what's the blackbox you're using - a FET, a LSI-chip, a module...00:58
wpwrakyes, but the NDA barrier is usually very low. basically anything you need to make it work or to understand how it's integrated is under NDA.00:59
wpwrakthat just doesn't work with open hardware01:00
DocScrutinizer05well, for some modules I know it seems you can use the NDAed docs to create your own "datasheet" that you're free to publish all the way you like01:02
wpwrakyou mean china-style ? www.china-shop.com/docs/foo-ds.pdf, for everyone to download, but with a big fast red CONFIDENTIAL on every page ? :)01:04
DocScrutinizer05for my personal "ethics model" the blackbox (and in the end you always and on blackboxes when you go "down" further and further) is sufficiently open when the specs and docs you the customer get with them (in this case it would be up to _you_ to provide those for your customers, prepared from your knowledge you got from the NDAed original datasheets) is specific enough to possobly replace the blackbox by something different01:06
DocScrutinizer05no, I mean in the way OM did (or planned to do) with glamo01:07
wpwrakphew. what was planned with the glamo is a LOT of work.01:08
wpwrakalso, don't forget the cost of negotiating such a deal01:08
DocScrutinizer05yes, for glame that's true. For a BT module it can't be that hard01:08
wpwrakcompletely inaccessible for a project without deep pockets01:08
wpwrakg-lame-o :)01:09
DocScrutinizer05basically, as long as there's a FOSS driver for a module, I think it's not nice but "open enough"01:09
DocScrutinizer05sure you also need the electrical specs of the interface, and the mechanical and thermical specs01:10
DocScrutinizer05my guidline is: could I replace it by another component, for repair or modding the device whatever01:11
wpwrakin this case, you also want to know a bit more about it. after all, this is a sensitive usage scenario01:12
DocScrutinizer05let's think of an ASIC. I don't need to know the chip design nor the verilog source, as long as the circuit datasheets are sufficient so I could design a replacement whenever the chip goes south01:12
qi-bot[commit] Werner Almesberger: poly2d/: add triangulation (untested) (master) http://qi-hw.com/p/cae-tools/f93e93c01:14
qi-bot[commit] Werner Almesberger: poly2d/f2d_tri_holes.cpp (tri_holes_append): rename to f2d_tri_holes_append (master) http://qi-hw.com/p/cae-tools/3d8a8d001:14
qi-bot[commit] Werner Almesberger: poly2d/p2d_hsort.c (p2d_hier_free): make freeing of polygons optional (master) http://qi-hw.com/p/cae-tools/d75835901:14
qi-bot[commit] Werner Almesberger: poly2d/f2d_tri_holes.cpp (find_point): don't loop forever on closed polygons (master) http://qi-hw.com/p/cae-tools/d0b569601:14
qi-bot[commit] Werner Almesberger: poly2d/f2d_tri.c (f2d_free_all): free list of faces (master) http://qi-hw.com/p/cae-tools/ca271dd01:14
qi-bot[commit] Werner Almesberger: cameo/: add "stl" command to generate STL slices (WIP) (master) http://qi-hw.com/p/cae-tools/8b6e41601:14
DocScrutinizer05yeah, sure for this special case you might want to know more about the BT internals. But I doubt you get to know sufficient internals to properly evaluate *any* hw on THAT level01:14
wpwrakwell, at least i'm at the chip level with 802.15.4. that's better than BT at the module level. there's a lot of information lost between chip and module.01:15
DocScrutinizer05the risk and threat evaluation needs to base on more general considerations (like RF power and range) and some assumptions form common knowledge about how BT works in general01:15
wpwraknaw, while i'd be more than happy to use a nice BT solution, that simply doesn't seem to be possible for now01:15
wpwrakieee 802.15.4, on the other hand, is possible, and i can reuse a lot of elements from a proven design01:16
wpwrakeven better, i can do ieee 802.15.4 at the chip level, not having to work with bulky modules01:16
wpwraki'd just keep BT for later. it's too tall a wall to climb now.01:17
DocScrutinizer05I don't see the usecase for that 802.15.4 stuff01:18
DocScrutinizer05you'd be honestly better off by providing a second wired interface in the volume you waste on that01:19
wpwrak2nd wired interface needs a different MCU that takes up a lot more space01:19
wpwrak(and has other inconvenient differences in the feature set)01:20
wpwrakwell, some of them are inconvenient. others would be very nice to have.01:20
wpwrakagain, something that may be worth considering later, for a number of reasons. but let's keep things simple for now.01:21
DocScrutinizer05I just fail to see the usecase for 6LowPAN01:23
DocScrutinizer05it's basically a poor cable replacement01:23
DocScrutinizer05and benefit from using it in this design is similar to the benefit of WLAN on a miditower PC01:25
wpwrakthere are tons of mice and keyboard with proprietary RF. few enough people seem to mind. so just forget that it's 802.15.4 and think of it as "SuperSecureLink+++" or whatever ;-)01:29
DocScrutinizer05all those mice and kbd are for stationary use without annoying cable01:30
wpwrakso, mobile use without annoying cable. a small enough difference :)01:31
DocScrutinizer05and they all have more or less severe problems with power supply resp batteries01:31
wpwrakdunno. some seem pretty decent. in any case, this one will have a very low duty cycle.01:32
DocScrutinizer05and in your case there is even more trouble with power supply than in any mouse or kbd, since you worst case need to power those from your device01:32
DocScrutinizer05honestly a USB host on a device with a CR2032 is a bad idea01:33
DocScrutinizer05you need to "borrow" that host - incl the power it provides - from somewhere else01:34
wpwrakif you want usb host, you need the y-box and an external supply01:35
DocScrutinizer05forwarding it may be feasible, but providing a USB to plug in a kbd on a device that has CR2032 is a nogo01:35
wpwrakthere's no USB power supply. and yes, the math is easy enough ;-)01:35
DocScrutinizer05that kinda defeats mobile use pretty much01:35
wpwrakwell, it's a fairly exotic scenario anyway.01:36
DocScrutinizer05I still miss the user manual that explains to me how to use this device in daily life01:36
wpwrakwith proper browser plugins, hardly anyone should want that mode01:36
DocScrutinizer05browser plugins?01:37
DocScrutinizer05where the heck I get those from, in an internet cafe?01:37
wpwrakif you really wish to manage your pws from a pc in an internet cafe ...01:38
DocScrutinizer05yes, that's the only usecase I can think of. I'm in holiday and want to access all the fora and maybe even register with new ones. Access my bank account, my email web frontend (no way to add browser plugins for that either) etc01:40
DocScrutinizer05another usecase I can think of is at ATM01:40
wpwrakbut you don't need them for that01:40
DocScrutinizer05to find out about my PIN01:40
wpwrakyes, of course01:40
wpwrakbut for that the baic device is suficient01:41
DocScrutinizer05like I already said, I don't basically need an external device at home for my own PC. There I got a sw solution that's working nicely01:42
DocScrutinizer05so unless I can read the user manual that gives me an idea about the usecases this device is intended for and how those are working, I probably can't contribute anything useful on the general design / productspec / usability level01:45
DocScrutinizer05no offense intended01:45
DocScrutinizer05I razther guess I'm simply too poorly informed thanks to myself not reading all the existing info about it01:46
DocScrutinizer05so my fault01:46
wpwrakyeah, this sort of information is widely scattered for now. and a good part of it is just in my head :)02:07
DocScrutinizer05I see 4 operation modes that make sense to me: (a: administrative) a1: attached to a PC via USB as peripheral like e.g mass storage, so user can edit stuff.  a2: same as a1 but via wireless, nice for frequent low-effort automated sync e.g. to a software pw vault on personal workstation.  (b: supportive) b1: same hw setup as a1 but acting as USB kbd "injecting passwords, works on virtually every PC.  b2: same as a2 and b1, via 02:07
DocScrutinizer05wireless (works only on PC that have a special driver installed for 6LowPAN HID, so this doesn't "make sense") b3: manual lookup of passwords using the device's MMI (scrollwheel, display)02:07
wpwrakbut it'll be easier to show things once the basic hardware works. and this should be relatively soon, if things progress as smoothly at they did so far02:07
wpwrakthe 802.15.4 dongle would just present itself as HID device. you can do quite a lot with that already.02:09
wpwrakbasically USB (cable) and 802.15.4 dongle would provide the same functionality towards the host02:09
DocScrutinizer05then that's just fine02:10
wpwrakan external keyboard may be less useful with anelok in the middle but just for massive standalone editing. e.g., if you need to enter a lot of accounts and you don't have any pc you trust for that nearby.02:11
wpwrakso it's a fairly narrowly limited scenario, too02:11
DocScrutinizer05basically an non-existent scenario02:11
wpwrakfor most uses, you'd just use HID. not quite sure how hard/easy browser integration for the fancier items would be, though02:11
wpwraknaw, not non-existent. depends on how you work.02:12
wpwrakan important concept is also that you can compartmentalize things. so your could manage things it has access to anyway, while it wouldn't be allowed to touch other things.02:14
wpwrakstill need to flesh out the details, though. plenty of crypto fun in this :)02:14
wpwrak"your" = your PC02:15
wpwrakwhat i mean is that you don't have to give every system that wants to maintain items in the pw database access to everything02:23
DocScrutinizer05I figure a scenario that I'd consider highly desirable and useful: you type e.g. "/msg nickserv identify". The moment you type the "y" of identify, on the screen appears "/msg nickserv identify 04<<ANELOK-PASSWORD y/n?>>01" and you hear a beep from the keylogger ANELOK in the cable from kbd to PC. For security sake you won't be allowed to enter master password to unlock ANELOK via 104-kbd, you need to use the ANELOK directly to 02:25
DocScrutinizer05unlock. So if you hadn't yet for this session, you unlock ANELOK and then type "y" on 104-kbd which results in replacing <<ANELOK-PASSWORD y/n?>> (via BS) by the real password02:25
wpwrakheh, could be fun :) with the currently planned feature set, you could certainly experiment with such things02:30
wpwrakif you're running just a HID, you probably wouldn't want to have that <<ANELOK-PASSWORD y/n>> show up. but anelok could then simply suppress further keystrokes until you answer whether you want to send a pw or not (with a timeout, in case it misfired badly and you're confused)02:32
DocScrutinizer05absolutely same applies e.g for my online banking, where the trigger regex would be my account number followed by <TAB>02:33
wpwrakif you have more than bare HID, e.g., some plugin, the IRC client could ask directly02:33
DocScrutinizer05of course the reaction and notification must be configurable for each single regex trigger02:34
wpwrakyes, you could apply this to a number of scenarios. it would basically compete with plugins. the latter would be cleaner but also harder to do.02:34
DocScrutinizer05since e.g for online banking the next thing after account number would be the 5digit PIN, evidently not enough for a lengthy message02:35
DocScrutinizer05I'm searching for hw- as well as sw-solutions doing basically exactly this, since ages. Basically a kbd-macro driver/dongle, just I didn't think of a master password yet02:37
DocScrutinizer05and while similar things already exist for windows on a sw kbd driver level, on linux the whole input stack is such a PITA that I gave up on finding as well as DIY coding such a thing02:38
DocScrutinizer05a dongle to plug between kbd usb and PC usb would be a dream02:39
DocScrutinizer05even more when it comes with own MMI02:39
DocScrutinizer05plugin for IRC client?02:45
DocScrutinizer05I prefer the hardware approach here02:45
DocScrutinizer05more universal, more "clean" (I differ here), and portable02:46
DocScrutinizer05even platform independent, as long as it's a USB kbd02:47
DocScrutinizer05related: http://www.thinkgeek.com/product/ae83/03:16
wpwrakwell, as i said, with the current design, you'll be able to have USB host and HID-over-RF. so you can have that sort of communication path, with whatever intercept you want to place there03:20
wpwrakgoing all wired would be considerably more difficult. you'd either need two MCUs, each providing one USB, or one of the MCUs that already come with two. in the latter category would be the STM32F2 (and the STM32F4, and that's about it), which i happen to know quite well.03:22
wpwrakcompared to KL25, they 1) draw more power, 2) come either in BGA or much larger packages, and 3) lack built-in USB power management. on the upside, they have03:22
wpwrak1) two OTG ports, 2) a hardware RNG (to complement the one in the RF transceiver), 3) more processing power.03:23
wpwraki'd like to look into using these chips in the future, but they're a considerably more expensive choice, not only in terms of chip cost but also in terms of design complexity03:24
wpwrak(phantom keystroker) yeah, you can of course use anelok like this, too: go your login screen, move to the user/account/whatever field, unlock anelok and select the account entry, then let anelok "type" the account info. repeat for the password.03:26
wpwrakthat's an open loop scenario: anelok doesn't get any feedback from the PC. but it's not necessarily weaker than what you're describing.03:27
wpwraka closed loop would be better, of course. e.g., the browser specifically asking for login data and anelok providing it in a way that doesn't depend on the input focus and all these little details.03:28
DocScrutinizer05I can't see this type of browser plugin go anywhere. Nobody has implemented such thing yet, and I guess it's terribly hard to implement something that goes seamlessly with random websites03:36
DocScrutinizer05not to think about random other browsers, PCs in cafes etc, and even OS03:37
DocScrutinizer05IOW when you already got a sw solution to deal with your passwords, so why do you need a hw dongle to complement this?03:37
DocScrutinizer05sw solution == the plugins needed03:38
DocScrutinizer05could as well use arbitrary existing SSO solutions03:38
DocScrutinizer05security audit a nightmare either way03:39
DocScrutinizer05another nice related site: http://www.irongeek.com/i.php?page=security/usb-hardware-keyloggers-1-keycarbon03:39
DocScrutinizer05btw thanks for pushing me on all that, it gave me an idea for a nice new app on N90003:42
wpwrak(difficulty) dunno. this guy seems fairly confident about the browser integration: http://ob-security.info/?p=63103:44
wpwrakhehe :)03:44
DocScrutinizer05N900 comes with: USB (both device and hostmode); drivers for HID (bluemaemo); hw KBD; touchscreen you can use as mouse replacement; BT to attach BT kbd and mouse while N900 is attached to PC emulating a kbd and mouse, and charging03:44
wpwrakyes, such a device has tons of interfaces. the drawback is that you'll want to use it for more things than just passwords. and that's where vulnerabilities come in.03:45
DocScrutinizer05and it easily can maintain any level of security and privacy and smartness you'd possibly need to keep your password vault safe03:45
wpwrakbut yes, as long as you find that risk acceptable, a smartphone makes a very nice password safe03:46
DocScrutinizer05now this porject just fails due to my schedule cluttered and me being too lazy ;-)03:47
DocScrutinizer05would be really nice particularly since I always carry a N900 with me, and even had a second one to sync with and keep the 2nd one plugged in to my PC all the time03:47
DocScrutinizer05doing plain HID emu via USB should be way more simple still03:49
DocScrutinizer05wpwrak: http://ubertooth.sourceforge.net/06:47
DocScrutinizer05wpwrak: designed in KiCAD ;-D06:59
mthapelete: -16 is -EBUSY, so it's probably like Paul said, something else claimed the range before07:56
mthdid you remove the UDC from the platform configuration?07:57
apeletemth: no I didn't. I thought not building the udc driver into the kernel was enough08:00
mththat should be enough08:00
mthmaybe double check that it is indeed not being built08:00
apeleteok, will check that08:01
apeletemth: from my kernel config file: # CONFIG_USB_JZ4740 is not set08:05
apeleteso it shouldn't be built08:05
larscyou can check /proc/iomem08:05
larscto see who requested the region08:05
larscjust `cat /proc/iomem`08:05
apelete'kay, will try that when I get back from work then08:06
Action: apelete wish he had more flexible work hours08:08
larscmine is more like 9-9 ;)08:31
apeleteha ha :-)08:32
apeleteI'm trying to stick to 9-5 in order to get some hacking time left08:32
apeletenot easy to do, especially in a big company08:34
larscwhat do you do for your dayjob?08:35
apeleteworking at intel as a contractor, on android devices08:35
apeletekernel debugging, and working on debug & profile tools08:37
larscany chance you'll be in edinburgh next week?08:38
apeletemostly trying to stick to low level free software stuff ;)08:38
apeletelarsc: don't think so, I'm working in Toulouse (France). why ?08:38
larscLinuxCon Europe08:40
larscand Embedded Linux Conference Europe08:40
larscand all the other Linux related conferences08:40
apeletedoesn't one need to get an invitation to attend those ?08:42
larscyou need to pay lots of money08:42
larscbut that's all08:42
larscI saw that some Intel employees are going to give talks, so I though you might be there as well08:43
apeletenever had the chance to attend those events yet08:45
larscIt's my first time as well, as it is close by this year08:46
wpwrakDocScrutinizer05: hmm, could be something worth looking into. the chip is already EOL, but there's a replacement08:46
apeletelarsc: ok, I see. I usually go to FOSDEM, not only close by but also take place during the week-end08:47
DocScrutinizer05I wonder if they eventually will get a complete bluetooth stack working on that thing08:48
larscapelete: me too.08:48
apeletetake place during the week-end -> no need to take the week off or ask for management agreement to attend ;)08:49
apeletelarsc: ah great, maybe we'll meet at Fosdem next year then :)08:50
DocScrutinizer05apelete: yeah, it's not that close to my place, so I felt a bit unhappy about the ticket prices08:55
DocScrutinizer05no matter if train or plane, it's always ballpark of 150EUR one-way and takes 5..8h08:56
DocScrutinizer05there are airplane tickets to Bruxelles for some 35EUR but not from any airport I could reach without a train ticket for 120EUR08:58
larsc6h train ride from here, I think08:59
larscor 8?08:59
apeleteDocScrutinizer05: yeah, I pay the same for train tickets to Bruxelles too, but since it's one of the few events I can actually attend, I try not to think about the price :)08:59
DocScrutinizer05well, at a certain expense all trying doesn't help ;-D09:00
DocScrutinizer05300 for tickets, 200 for hotel09:00
DocScrutinizer05another probably almost 100 for local transport09:00
DocScrutinizer05I could already get a week of holiday in the south for that09:02
larscIt's all in walking distance09:02
DocScrutinizer05except the hotel, and the airport09:02
wpwrakseems that the RF frontend chip isn't even required09:05
larscright, but last year taxi from the train station to the hotel was like 10 euros09:05
apeleteDocScrutinizer05: same problem here, but the positive side is there's not registration fee for FOSDEM09:05
apeleteonce you get there, it's all free ;)09:06
DocScrutinizer05:nod:  anyway I can't afford09:06
DocScrutinizer05always the same with such stuff (holiday as well) either you got all the time for it, or sufficient money. Hardly both at same time though09:07
apeleteyeah, that's a tricky problem to solve09:08
wpwrakDocScrutinizer05: being software-heavy does indeed seem to be the main limitation of this chip. well, and not supporting EDR. but that's probably okay. alas, it doesn't have an RNG, so it could only be used in a design that already has one.09:14
wpwrakDocScrutinizer05: a nice thing to consider in a v2 design, though. thanks !09:15
whitequarkwhich chip do you discuss?09:15
wpwrakTI CC250009:16
wpwrakthe successor of the one these folks use: http://ubertooth.sourceforge.net/09:17
DocScrutinizer05wpwrak: welcome09:17
paul_boddie"Inevitably many products have one or two components that fall into the general category of unobtainium, i.e. impossible to find, and that often causes delays. For the Spark Core, this component was the CC3000, our Wi-Fi module, which was an important part of the design and therefore could not be replaced."10:05
paul_boddieSorry: the CC2500 reference got me thinking about TI products and it all sounded familiar.10:06
paul_boddieAnother fairly interesting project I saw recently was one whose modules are described here: http://tessel.io/modules10:08
paul_boddieThat project seems to use a Bluegiga BT module. There are docs on bluegiga.com, but I imagine the firmware is off limits as usual.10:11
wpwraksourcing issues can pop up any time. that's just life :) chips are generally much safer than modules. as far as i now, ti don't have a bad reputation when it comes to lifecycle management.10:17
wpwrakof course, stay away from brand-new product categories. from anyone :)10:18
wpwrakit's also always nice to have a fallback solution. e.g., a cc2500 (new) could probably be replaced with a cc2400 (EOL)10:18
wpwrakmodules are high-risk because they depend on a lot of factors: availability of components, health of the chip maker, health of the module maker, volume of the module (generally much lower than that of the corresponding chips), and so on10:20
paul_boddieYes, TI should be OK. If you have a lot of unhappy Kickstarter donors (who think they are customers) then time is not on your side, I suppose.10:20
wpwrakif you're buying a million a month, you make making that module worthwhile on your own, so you don't have to worry about such things10:20
wpwrakbut if you're a kickstarter-grade project, you're completely at the mercy of others10:20
wpwraki like the title "why kickstarter projects are [...] late". something completely unheard of in IT ;-)10:21
wpwraki actually wonder how they deal with cost overruns. what if you find out your kickstarter project needs 50% more money than you planned ? is there a mechanism for getting people to add some, well into the project ?10:22
larscyou have to pay for it yourself I guess10:24
paul_boddieI was reading the comments on the GCW-Zero Kickstarter. Quite depressing with all the foot-stamping and "I want it now!" shenanigans.10:24
paul_boddiewpwrak: Like when a company has to issue more shares? Sorry, Kickstarters, but your share is diluted and you only get half a gadget unless you pony up!10:25
wpwrakwell, what kind of audience do you expect if you're making a game console ? ;-)10:26
DocScrutinizer05@ "pay yourself"10:26
wpwraklarsc: i suppose few projects would be able to do that ...10:26
wpwrakat least not after passing a certain size10:26
larscyou probably also need a full time position to just manage all the angry comments10:27
DocScrutinizer05ask pandora?10:27
wpwrakpaul_boddie: yeah, kinda. only that shared tend not to have a direct 1:1 connection with products. so you can hide it better.10:28
DocScrutinizer05afaik they should have experience in handling such a situation?10:28
wpwraklarsc: just recruit the angriest of the commenters :)10:28
DocScrutinizer05or wasn't it openpandora?10:28
wpwraknomen est omen ;-)10:28
DocScrutinizer05actually I wonder what happened really to them10:29
DocScrutinizer05I missed their undoubtly verbose and apologizing explanations what happened to the project 10:30
larscit looks like they shipped eventually10:30
DocScrutinizer05I just heard they had to ask for yet another full spin of the donation wheel10:30
DocScrutinizer05quite some time back10:30
wpwraki think they financed finishing their first batch with the money they made for the second batch10:31
wpwrakkinda pyramid scheme10:31
DocScrutinizer05yep, sth like that10:31
DocScrutinizer05but how did they burn the money for the first batch?10:31
paul_boddieThey dealt with a company that made bad PCBs, if I remember correctly.10:32
DocScrutinizer05bad luck10:32
paul_boddieOr maybe the company who was doing the board assembly dropped the ball, and then the PCBs weren't good.10:32
larsc"We received the last working PCBs from them July 2011. Over 1000 PCBs were broken... over 300K EUR lost."10:32
paul_boddieOMG, facts! ;-)10:33
DocScrutinizer05larsc: thanks10:33
paul_boddieI think Golden Delicious were talking to them about manufacturing partners at some point.10:34
DocScrutinizer05> >The bank froze Craigs bank account due to a sudden increase of funds there. Within a few days, over about half a million USD popped up there - and the credit card companies panicked (even though Craig explicitely told them before that this will happen) .Same with PayPal.. .Most of the payments had to be refunded and repaid using some other payment method.<<10:34
paul_boddieWell, dealing with the financial industry is perhaps even worse than random electronics manufacturers.10:37
DocScrutinizer05must've been in the USA. US Banks seem to be able to do whatever they like with _your_ money10:37
wpwrakin the US, chances are it's theirs anyway ;-)10:38
DocScrutinizer05see maemo Hildon Faoundation. The bank quit the account, and sent a check drawn on "Hildon Faoundation" that now nobody is willing to accept10:38
paul_boddieIt's the taxpayers' money, surely. ;-)10:38
larscI think it is common practice to freeze bank accounts if unusual transactions are detected10:39
paul_boddieDocScrutinizer05: And not for an amount that's worth changing your name to Hildon Foundation for?10:39
DocScrutinizer05I think they should go and die in a dark and dirty place10:39
DocScrutinizer05paul_boddie: a ridiculous 2k or sth10:40
paul_boddie$2M would be a different story. :-)10:40
wpwraklarsc: it just seems a bit unusual if that's for months. half a million EUR isn't *that* much after all. any bankster moves that in the morning even before entering the office10:42
DocScrutinizer05the idiot who had the approval on that account at "first niagara" was too dull to get it sorted10:42
DocScrutinizer05a lawyer ;-)10:42
larscpaul_boddie: or 20 billion: http://www.dailymail.co.uk/news/article-2437208/Is-Saddam-Husseins-fortune-warehouse-Moscow-Mystery-16-75bn-piles-cash-left-airport-years.html10:42
wpwrak"first nigeria" ?10:42
DocScrutinizer05yeah, my fav "abreviation"10:44
DocScrutinizer05it's obvious10:44
DocScrutinizer05I'd recommend to steer clear of that bank10:46
DocScrutinizer05and *paypal*... oh well.. we don't need to talk about them, eh?10:46
DocScrutinizer05they freeze your account when somebody in your street doesn't behave10:47
DocScrutinizer05they frozen the accounts of all those in a 10 story building that had a name slightly resembling the name of one user who also didn't do something really bad but somehow managed to get paypal's AI upset10:49
DocScrutinizer05anyway I need to finally check the policies/rules of kickstarter and that other similar thing10:51
wpwrakindiegogo. less US-centric than kickstarter10:51
DocScrutinizer05we (golDeliCo) eventually want to raise 20k EUR for the prototypes10:51
DocScrutinizer05and I for one would prefer that being a down payment to the final device cost as weel as some venture capital that doesn't need to get paid back when stuff goes terribly haywire10:53
paul_boddieI saw another one just recently: http://www.dragoninnovation.com/10:53
paul_boddieI even saw a Norwegian one recently, too. Everyone wants in on the action!10:53
DocScrutinizer05yeah easy money - for them10:54
rohDocScrutinizer05: afaik kickstarter is us only when it comes to payouts10:54
DocScrutinizer05roh: hi! please rephrase10:54
rohso maybe indygogo or its friends help you better10:54
rohDocScrutinizer05: you have to be a us citicen to be able fo get money from a successfull campaign at kickstarter10:55
rohatleast that was the case when i checked last.10:55
DocScrutinizer05Nikolaus disliked the whole kickstarter et al stuff particularly for the hassle and the fees, but he hopes for the "visibility" aka PR10:56
paul_boddieIsn't it US/CA/UK? Not that I really know.10:56
wpwraki think you need someone in the US to handle the finances (with kickstarter). you could still have the money go elsewhere, but yes, it's at least inconvenient10:56
rohDocScrutinizer05: kickstarter and friends is a very hipster-way of doing pr and funding.10:56
rohnot one which makes you look serious10:57
wpwrakwell, maybe US/CA now10:57
rohso if you dont need 'risk capital' but funding, i'd suuggest choosing a different concept. maybe also an open one, but different10:57
DocScrutinizer05I guess we'll go the jolla way of selling premium preorder vouchers10:57
wpwrakjust have the facebook and paypal logo on your campaign, then everyone will know you're serious :)10:58
DocScrutinizer05that give you eligibility for a certain service/goods10:58
rohDocScrutinizer05: i wouldnt give any cent for something as useless as that10:58
rohvouchers are a good sign of vaporvare10:58
DocScrutinizer05then what else?10:59
wpwrakif your project is a one trick pony, then there's no big difference between voucher and preorder10:59
rohsell goods.10:59
DocScrutinizer05just a sign "deposit money here please!" ?10:59
rohsimple as that. no futures, no bonds. simple preorder (if you need capital upfront)10:59
wpwrakindiegogo also has "flexible funding". there, you can keep the money also if you don't reach the funding goal. has higher fees, though.10:59
DocScrutinizer05sorry the last to posts don't help me at all10:59
DocScrutinizer05I dunno what's a "one trick pony"11:00
wpwrakif you're looking for kinda partial funding, that may be a choice11:00
DocScrutinizer05and "sell goods" is exactly what we can't do11:00
paul_boddieYes, as I recall Ubuntu Edge could have kept their supporters' money if they had chosen flexible funding, but Shuttleworth clearly wanted the full amount (or the publicity for specifying a big amount).11:01
wpwrak"one trick pony" = a pony (in a circus or such) that can do one trick, but only one trick. so it's totally single-purpose.11:01
DocScrutinizer05since there is a 0.3% risk we'll burn the money just like openpandora did11:01
wpwrakif your risk is really as low as 0.3%, you should go into the nuclear power business :)11:02
DocScrutinizer05wpwrak: fine, now I know what's a "one trick pony". Now what's a "one trick pony project"?11:03
wpwrakpaul_boddie: i think he must have been glad that he didn't reach the funding goal :)11:03
wpwrakDocScrutinizer05: a project that only makes one thing. so a voucher would still amount to buying that one thing.11:03
DocScrutinizer05no, since the voucher isn't allowed to trade in and get your money back11:04
wpwrakyes, see "flexible funding". in the end, the result would be pretty similar.11:05
DocScrutinizer05while with a preorder with down payment yo are liable to refund11:05
rohDocScrutinizer05: you're liable anyhow.11:05
wpwrakexcept that you could make the voucher for < 100% of the value. but that makes the final orders messy because you have two payment transactions.11:05
DocScrutinizer05roh: no11:06
rohthat doesnt change. the difference is that one thing sounds serious, one sounds fishy as hell11:06
DocScrutinizer05you're liable to provide what you sell11:06
wpwrakso sell what you can provide :)11:06
rohDocScrutinizer05: yes and no. if you can't deliver you are liable to money-back11:06
rohDocScrutinizer05: depends on who you accept as customer and where from11:06
DocScrutinizer05well, we can make stickers "Neo900" and attach them to Nokia311011:07
rohinternational sales law is a bitch. and you WANT to keep is simple and trustworthy11:07
wpwrakuif you think you have a significant risk of total failure, you could simply explain this in the funding "brochure"11:07
DocScrutinizer05roh: so what do you suggest we should do to get 20k venture capital?11:08
rohDocScrutinizer05: 2 possibilities. one: run it as a 'funding' campaign, give people nothing in the end but their pride11:08
wpwraksay that this is for a first small run with high risk. say that you examined all the visible risks ("known unknowns") but that there can be invisible risks ("unknown unknowns") that you won't be able to see before you actually try making some units.11:09
rohDocScrutinizer05: well.. the other option is make it more expensive, put in real goods.11:09
larscDocScrutinizer05: what's the expected ROI?11:09
rohDocScrutinizer05: or: of course:.. go to a bank. loan money.11:09
DocScrutinizer05everybody gonna save their money for the final product, since it's already expensive enough11:09
rohbasically its just a) raise money from people or b) raise credit from bank (or somebody rich)11:10
wpwrakif you have a large enough community, you could ask if there would be people who'd together cough up 20 kEUR, if it's really so little you need11:10
DocScrutinizer05wpwrak: all that doesn't change a bit on the "contract"11:10
rohraising from people could be easier, if the risk seems low and one does get more than a thanks. so.. like 'buy a expensive phone which gets shipped in a year'11:10
wpwrakit changes the contract substantially, since you're no longer promising the product. you're promising to make a honest effort at TRYING to make the product.11:10
DocScrutinizer05roh: thanks! both won't fly11:11
rohDocScrutinizer05: why not? raising money from people via 'here is your voucher' will get you nothing but laughter and some beers, but not 20keuro11:11
wpwrak20 kEUR is not a lot. some people spend more that that every year on pretty cars.11:12
DocScrutinizer05so what?11:12
DocScrutinizer05unless you're one of them11:12
rohatleast i would never give money to a maybe which asks for more to get a real product. only to a 'we try hard or go broke'.11:12
wpwrakit means that you have a good chance that your community may contain individuals who could give you 20 kEUR without too much fuss11:12
wpwrakin a role as investors11:13
DocScrutinizer05roh: your problem11:13
wpwrakif the thing flies, you sell the units you made and repay them11:13
rohi could imagine giving some money to a good 'we try, but you are only investor', but that would be much less than when promising a real device.11:13
DocScrutinizer05wpwrak: nope11:14
rohDocScrutinizer05: if 20keuro invest are your problem on that project, it looks grim anyhow.11:14
paul_boddieWell, Kickstarter is almost the same as the "you are only investor". It's just that most people haven't been burned, and so everyone thinks of it as a cheap gadget store.11:14
rohthats the kind of money even i could borrow from a bank if in dire need.11:14
wpwrakplan in depth: try first am investor search. if that doesn't work, you can always fall back to crowdfunding11:14
DocScrutinizer05roh: then by all means DO IT!11:15
rohpaul_boddie: thats why so many kickstarters fail which nobody knows about11:15
wpwrakpaul_boddie: indeed. alas, DocScrutinizer05 is too honest to just use that ignorance for the greater good :)11:15
rohDocScrutinizer05: for such a project? nope. that needs to be done by somebody really involved.11:15
paul_boddieroh: Kickstarter failure cover-up shocker!11:15
wpwrakwhere's the NSA when you need them ?11:16
rohDocScrutinizer05: its not my project, and thus it shouldnt be my risk.11:16
DocScrutinizer05you're psting extremely helpful suggestions11:16
rohDocScrutinizer05: dont forget.. you are talking to somebody who still doesnt have and use a smartphone with a reason ;)11:17
wpwrakDocScrutinizer05: I suppose you have a means to communicate with people interested in neo900 ? like a mailing list or such ?11:17
paul_boddieThe EOMA68 people seem to be getting by, but their funding isn't completely transparent.11:17
DocScrutinizer05wpwrak: we do 11:17
paul_boddieI think they have various companies involved who see the whole thing as a means to an end.11:17
rohpaul_boddie: it it transparent at all? i dont have a clue where the money comes from at all11:17
wpwrakDocScrutinizer05: then you can float the idea. see if you, say, can find 4-5 people who together invest the 20 kEUR.11:18
rohDocScrutinizer05: my point is not that 20k you are searching, its that if the 20k are a real problem, i fear for the project in total.11:18
wpwrakkeep the number small, so you don't have an angry mob to deal with on the first sight of trouble11:18
larscDocScrutinizer05: write a business plan11:18
paul_boddieIn other words, some people want what they're doing but don't want to design it themselves.11:19
DocScrutinizer05we got 77 people who are willing to invest the 20k when they get a device accordingly cheaper once we get some devices to ship11:19
rohDocScrutinizer05: there are much more expensive things coming your way when going into production, some small intermediate financing shouldnt be an issue11:19
DocScrutinizer05larsc: for whom?11:19
wpwraklarsc: at least the part that goes to the point where the 20 kEUR are paid back in some way (if things work out)11:19
paul_boddieBut unlike the usual cheap company behaviour where they wait for everything to be done for free by volunteers, they are at least paying something towards getting things made.11:19
DocScrutinizer05roh: thanks, cptn obvious11:20
rohDocScrutinizer05: anything else will just keep you and the others working on it busy with boring stuff, hindering you solving real problems11:20
rohand i like the idea of you succeeding too much to like that idea; )11:21
wpwrakDocScrutinizer05: roh is right, though. you're worrying about the wrong problems. all you need is to convince a very small number of people. don't approach this as if you were selling millions of products on the mass market.11:21
rohafter all, its a 'unlicensed vertical product experiment' :) to use business terms11:21
DocScrutinizer05ok, lesson learned. Don't discuss business plans with public11:23
wpwrakand the biz plan would be for your investors. show them that you know what you're doing. if you know that 20 kEUR will get you there, you should have done these calculations already anywhere. if you haven't done them, you don't know whether 20 kEUR will be enough.11:23
DocScrutinizer05there ARE NO investors11:23
rohDocScrutinizer05: sorry that we are'nt able to help you in a way you'd like better, but atleats we arent bullshitting you like a bank would11:24
wpwrakhave you asked the people on the list whether some of them would be willing to put money into the project in such a role ?11:24
rohDocScrutinizer05: somebody always is the investor. thats the one carrying the risk.11:24
rohanf if its not you, its the bank or the customers11:24
larscDocScrutinizer05: well you won't find investors without a business plan11:25
DocScrutinizer05it's the customers!11:25
DocScrutinizer05I said that11:25
DocScrutinizer05if you had foolowed what I wrote11:25
DocScrutinizer05followed even11:25
DocScrutinizer05larsc: we don't even look for investors11:25
wpwrakwhat strikes me as odd is that this is a trivial amount. for most people with a regular income, this would be a tiny fraction of their disposable savings. so unless your community consists entirely of starving artists, there ought to be a few members who'd do it.11:26
rohDocScrutinizer05: so you want 260E from 77 persons upfront for 'risk capital' and the rest after?11:26
wpwrakDocScrutinizer05: write "investor" if you prefer ;) it's not as if you'd ask for a large amount of money11:27
DocScrutinizer05actually 100EUR from 20011:27
rohwell.. if you got the people ready do it. but call it what it is: risk capital. and you need to pay it back or give out shares or so.11:27
rohanything else will get less people to sign up11:27
paul_boddieIn the Kickstarter era, it isn't "investor": it's "premium backer". ;-)11:28
rohpaul_boddie: well. thats why so many of those fail.11:28
wpwrakif you have a plan that goes like this: 1) "investors" provide 20 kEUR. 2) you make a small run. 3) each investor gets his/her unit. 4) you sell the rest into your community. 5) from the revenue, you repay the investors.11:28
DocScrutinizer05roh: I called it vouchers and you nashed it. Now you call it shares and suggest it11:28
wpwraksorry, the "investors"11:28
DocScrutinizer05bashed even11:28
paul_boddieOr "unicorn rider". You get to choose!11:28
rohDocScrutinizer05: its the 'worse' option from my pov. either way11:29
wpwrakmake it clear that you expect them to be at least somewhat idealistic. so you don't promise 30% ROI p.a. or whatever.11:29
rohDocScrutinizer05: and shares are 'owned parts of a company'11:29
DocScrutinizer05well, seems Nikolaus has some experience that is generally lacking in this channel11:30
wpwrakyeah, you should involve him. he's got business experience.11:30
roh'vouchers' are pieces of paper which allow you first dibs onto a rare product.. which is a stupid idea... because its only rare if you plan to make too few of it... so i wouldn buy any voucher at all.11:30
paul_boddieSeriously, though, "vouchers" is OK if you aren't offering any share in the success of the endeavour, but are just offering preferential terms on any delivered product.11:30
rohi'd buy shares if i believe in a project and 'dont need that money' for now11:30
DocScrutinizer05who says vouchers are about "rare"?11:31
paul_boddieBut as roh says, if someone believes enough in something to back it like that, they probably don't need a discount.11:31
DocScrutinizer05and we don't plan to make too few but exactly the right amount of those devices11:31
wpwrakvouchers may be okay, but they're basically a detour. you could just call it a purchase with a pre-agreed risk of default.11:32
DocScrutinizer05that's the same in my book11:32
wpwrakif you fail, you'll have burned the money. so you have nothing to back the voucher with. so it's just the same as declaring defeat on a production run of preordered units.11:33
DocScrutinizer05roh: we don't want money from shareholders who are not interested in the product but only in the shares11:33
rohthen extend your book. its the essential difference for a lot of people11:33
rohDocScrutinizer05: thats not the point.11:33
rohDocScrutinizer05: the point is that vouchers arent worth shit11:33
larscmaybe you can find a hedgefund that is willing to invest ;)11:33
wpwrakDocScrutinizer05: if the investment is 20 kEUR, you will only attract those who like your project (for whatever reason)11:34
rohits nothing. shares are worth your investment.11:34
DocScrutinizer05maybe I can find a better way to spend my time. Bye. Have to run doing some RL stuff11:34
larscroh: if the company is bankrupt your shares are worth nothing11:34
wpwraklarsc: step 1) raise funding goal from 10 k to 10 billion :)11:34
rohlarsc: true. thats risk11:34
larscsame with the vouchers11:35
rohlarsc: for vouchers to be worth nothing, nothing needs to happen. they are worthless from the start11:35
rohits just a promise, no contract. shares are a contract.11:35
wpwrakshares may be tricky - you may run into regulations with these ...11:35
rohwpwrak: true. but thats true for all these investment forms11:35
wpwraknot sure about vouchers11:35
paul_boddieYes, shares bring about lots of financial regulations.11:35
larscthe vouchers are basically a product you sell11:36
rohpaul_boddie: both do, atleast when selling to people around here.11:36
wpwraki'd keep it simple: first, try to find a small number of people who can put the money. there, it's just a gentleman's agreement. easy and clean.11:36
DocScrutinizer05paul_boddie: the voucher costs 100EUR and is worth a 150EUR discount on the final product sales price11:36
paul_boddieroh: Well, everything is regulated, but shares are things that you cannot even offer people in some places, depending on where you are.11:37
wpwrakif that doesn't work, you can still try the "small crowd" approach. but it's messier because you have to deal with a lot more people.11:37
wpwrakpaul_boddie: and then the mandatory filings, ... :)11:38
rohpaul_boddie: thats why i'd start with neither. and simply presale a defined number of units for the 1st run.11:38
paul_boddieDocScrutinizer05: Right. So the purchaser risks 100EUR to gain 150EUR.11:38
DocScrutinizer05to gain 5011:38
rohpaul_boddie: fixed price, fixed gain, finance the 2nd run from that.11:38
DocScrutinizer05which is virtual anyway, since we're not planning to seel any more than got preordered11:38
rohin the end you are 'liable' for the money you collect anyhow.11:39
paul_boddieI'm no expert in shares, but I see it every time I get a mailing about some worthless shares I have. Stern voiced warnings about people in the US not even being allowed to look at the offer, and so on.11:39
DocScrutinizer05roh: LOL, sure. you're thinking 80% win margin, eh?11:39
DocScrutinizer05we're thinking 0% to 20% right now11:40
wpwrakDocScrutinizer05: i think you may underestimate how enthusiastic people can be. unless they really distrust your business acumen (and that's where you need that biz plan), it's probably just as easy to get the full amount you need from than than having them pay almost everything now and then a little bit more later.11:40
paul_boddieDocScrutinizer05: Does Nikolaus have anything to say about the different donation things he was doing?11:40
DocScrutinizer05I didn't ask yet11:41
wpwrakpaul_boddie: that's what you get for investing into cuban tobacco farms ;-)11:41
DocScrutinizer05he suggested to get 50% from preorders, 50% from kickstarter for the PR effect11:41
paul_boddiewpwrak: I didn't know energy companies invested in Cuban cigar manufacturers, although stranger things have happened. ;-)11:41
DocScrutinizer05he also suggested a few donation options like "5 k buy you one of the prototypes"  "500 and we etch your name on all PCB"11:42
wpwrakpaul_boddie: they probably do, if something is left after they've through with investing into arms, drugs, politicians, and all that other shady stuff11:42
rohDocScrutinizer05: 80% sure not.11:42
paul_boddieI think there's a lot of goodwill for people who are serious about pursuing open hardware projects, and I wonder whether a donation model would work for some.11:42
DocScrutinizer05wpwrak: eh?11:43
paul_boddiewpwrak: But in terms of the US-based warnings, I think that it's purely financial: they don't want foreigners offering goodies to Americans when they have people like Bernie Madoff to do that domestically.11:43
DocScrutinizer05what do you think this device will cost?11:43
DocScrutinizer05wpwrak: 100EUR are like 15% down payment11:44
wpwrakpaul_boddie: precisely. and you can dress it up as investment. they may still write off the money but it'll motivate you a bit more and if something does come back, they'll be happy. (or may just tell you to keep it because their real goal wasn't to make money but see the project prosper)11:44
paul_boddieBut for the Neo900, yes, you can't really do it as an "everybody pitches in and it's work out" project.11:45
wpwrakDocScrutinizer05: you said you'd sell it for EUR 150 ? now you're saying it'll cost EUR 666. i'm confused.11:45
wpwrakoh, i see where i misunderstood11:46
paul_boddiewpwrak: I do wonder about the potential. I mean: you have Kickstarters that exceed their blatantly understated $10000 goal by a factor of 50, so there must be loose cash floating around.11:46
wpwrakthe voucher, bought for EUR 100, counts as EUR 150 against the EUR 666.11:46
DocScrutinizer05though it more likely will be 750 or 80011:46
wpwrakso the 20 kEUR won't really yield any devices for the waiting masses. now the "5k and you get a prototype" makes sense11:48
DocScrutinizer05no, the 20k will build the prototypes11:48
DocScrutinizer05see http://talk.maemo.org/showthread.php?p=137471611:49
wpwrakpaul_boddie: well, people generally do have a bit of money. unless they're trapped in a "have no cash, can only buy on credit (card)" situation. which is of course what the big players like.11:50
paul_boddieI guess it's a common problem. The EOMA68 people have taken orders, but I think the prototypes have all been funded by other (mysterious) means.11:50
paul_boddiewpwrak: There was a big argument about pay-day loan companies in the UK which touches this nerve.11:51
paul_boddieThe head of the Church of England got involved. Some trash talking, too. You could get a lot of people on firmer footing if they didn't have to borrow to make the cashflow work.11:53
wpwrakDocScrutinizer05: okay, so the voucher has two risks: 1) you fail in the prototyping stage, 2) you fail getting enough real orders. people may still drop out even if they've bought a voucher. "i just lost my job / got divorced / lost interest / whatever and can't afford the extra EUR 550 now".11:53
wpwrakDocScrutinizer05: no idea what percentage of "defection" you'd have to expect, but the number may be significant.11:54
DocScrutinizer05I consider that risk low enough to get handled by late orders11:54
wpwrakyes, they can make up for that. just don't call them "late" :)11:54
wpwrakpaul_boddie: on which side was the church ? traditionally, through should be all in favour of keeping the masses on a tight leash11:55
DocScrutinizer05they are going to pay the 50+11:55
wpwraka hardly noticeable difference :)11:55
DocScrutinizer05once we start the MP, no more orders taken for that batch obviously11:56
paul_boddiewpwrak: Actually, I shouldn't have said "the head of the Church" as that's the Queen. ;-)11:56
wpwrakDocScrutinizer05: you can always take orders for the next :)11:56
DocScrutinizer05sure, but that is a completely different topic11:56
paul_boddiewpwrak: But it's not really an instrument of control any more, at least not politically.11:56
wpwrakpaul_boddie: she's got her fingers everywhere, doesn't she ?11:56
paul_boddiewpwrak: Maybe we should ask her if she has some spare cash. ;-)11:57
DocScrutinizer05of course we hope for another 500 preorders coming in as soon as we show and ship real product11:57
DocScrutinizer05and Nikolaus said the MP bargain starts somewhere @ 1000/months11:58
DocScrutinizer05at least on the production line he's using11:58
wpwrakwhere does he plan to produce ?11:59
DocScrutinizer05factory takes risk :-D11:59
wpwrakgood :)11:59
wpwrakas long as the assembly cost is low, china isn't all that attractive if you factor in the "surprises"12:00
DocScrutinizer05I wouldn't like to hear a "sorry we fscked it up"12:00
wpwrakwell, at least not if you have a decent domestic industry. which you do of course.12:00
paul_boddieIt surprised me that the Fairphone people didn't try the German/European route given their volume.12:01
wpwrakwell, if you fab lies to you and then just burns your pcb/components stock, it's more like them doing the honors. but yes, in the end, you'll be perceived as the one having failed.12:01
DocScrutinizer05wpwrak: who's been those insane suckers who basically wanted to produce their own design and only sell it to us? FoxCon?12:02
wpwrakDocScrutinizer05: hmm ? don't get the reference12:02
DocScrutinizer05OM GTA0412:03
wpwrakpaul_boddie: perhaps they didn't listen to the right myths :)12:03
DocScrutinizer05anyway I guess that's what fairphone did when they decided toproduce in China12:04
wpwrakDocScrutinizer05: not sure i heard of such a plan. i remember some initial confusion about GTA04 but that didn't really go very far. and i thought the impulse for outsourcing came more from openmoko than from anywhere else.12:04
DocScrutinizer05yes, but those outsourcing plans went too far on the flipchart of FoxCon12:05
DocScrutinizer05they simply offered a custom build product to us which we could by from them on a per-device agreement12:05
DocScrutinizer05for an insane price12:05
DocScrutinizer05built and designed by _them_12:06
DocScrutinizer05sth along that line12:06
wpwrakheh ;-) yeah, i remember something like that. didn't know it was foxcom. though it was a small design house that wanted to reuse a previous project.12:06
DocScrutinizer05they didn't even want to have a look at our schematics or anything12:06
wpwrakthat may have been a wise choice ;-)12:07
DocScrutinizer05in the end yes12:07
DocScrutinizer05also a wise choice to not hand them the schematics12:07
wpwrakhmm yes, that could have been considered an act of sabotage12:08
DocScrutinizer05and I'm not sure if it was FoxCon, but I seem to recall the FC acronym flying around and everybody going "SHHHHH!" whenever somebody used those two chars 12:09
DocScrutinizer05wpwrak: at hindsight indeed12:09
DocScrutinizer05FC been damn right to not consider perpetuating life of that S3Cxxxx based design12:10
DocScrutinizer05I still wonder why OM picked that SoC12:10
DocScrutinizer05it can't have been the smartest design around12:11
DocScrutinizer05maybe the only SoC available to OM12:11
DocScrutinizer05which also had semi-public TRM12:11
DocScrutinizer05though even that got reverted at that point in time or even before12:12
DocScrutinizer05Samsung had a publicly available TRM of S3C2410 iirc, then it turned out they never will disclose the TRM for S3C24442, then they tokk down the TRM for S3C2410 as well12:13
DocScrutinizer05weird guys12:14
DocScrutinizer05btw, anybody heard sth of Rasterman?12:15
DocScrutinizer05is he still working for Samsung?12:16
wpwrakmay have been because of good contacts, plus access to data sheets (which was very good at that time)12:16
wpwrak(trm) yes, exactly12:17
wpwrak(raster) dunno. haven't heard from him in years12:17
rohDocScrutinizer05: it wasnt picked. its was there already.12:19
DocScrutinizer05aah, makes sense12:19
rohDocScrutinizer05: was as in 'legacy' stuff.12:20
DocScrutinizer05~lart FIC12:20
rohalso.. not the worst choice. samsung makes nice soc.12:20
DocScrutinizer05the S3C2xxx were EEEW12:20
rohalso cheaper than freescale and easier to use that anything from ti. also easy to source in small volumes12:20
rohDocScrutinizer05: anything else you know now is more recent or even bigger crap12:20
DocScrutinizer05OMAP3 not been arounf at that time12:21
DocScrutinizer05prolly not even OMAP212:21
rohalso ti is always shitty to source, bad documentations with lots of nda issues back then, expensive AND complicated to integrate12:21
rohthe latter seemingly didnt change12:22
DocScrutinizer05SoC evolved fairly nice during last 6 years12:22
DocScrutinizer05hmm, the integration doesn't look that cimbersome on OMAP312:22
DocScrutinizer05NDA changed a bit I guess12:23
rohomap3 is hell compared to others12:23
DocScrutinizer05how's that?12:23
rohdifficult ram design/layout12:23
DocScrutinizer05hm, yeah the PoP isn't really nice12:23
rohi dont know anybody who did a successfull design without help from ti12:23
DocScrutinizer05regarding integration12:23
rohafaik even nikolaus did need their help12:24
DocScrutinizer05never heard of any such help12:24
rohbut it know loads of people who used samsung and freescale without help12:24
DocScrutinizer05at GoldeliCo12:24
rohDocScrutinizer05: check the ml from gta03 times12:24
rohalso the beaglebones/boards had issue with the external parallel bus interfaces12:25
DocScrutinizer05it's just a PITA to find approved OneNAND/RAM PoP12:25
rohonenand is shit anyhow12:25
DocScrutinizer05and the really funny stuff I'm dealing with right now is the unobtainability of a 1GB RAM PoP12:26
rohatleast i dont like it. nonstandard stuff. expensive. not easy to source. not many vendors who make it.12:26
DocScrutinizer05seems like all chip manufs decided that 512/512 is enough12:27
rohyou mean for pop only?12:27
DocScrutinizer05do you know if DDR2 or DDR3 is backward compatible to DDR1?12:27
DocScrutinizer05yep, for PoP12:27
roh(backward compat)12:28
DocScrutinizer05nope compatible or nope know12:28
DocScrutinizer05yeah, that's what I thought as well12:28
rohthats why so many use ddr3 now. lower voltage and better availability12:28
rohddr is hard to get and ddr2 is getting more expensive12:28
rohstill ddr3 is more hassle routing properly.. improved clocks mean 'improved' requirements when it comes to layout12:29
DocScrutinizer05LPDDR1 168p PoP is seemingly impossible to get at 1GB12:29
rohwhy should that be avail at all?12:29
rohddr1 is like 10 years ago12:29
DocScrutinizer05well, fun bit is that Nokia got a custom chip that did it12:30
DocScrutinizer05in N912:30
rohdoesnt matter. they arent part of normal sourcing processes12:30
DocScrutinizer05a phantom chip, nonexistent12:30
rohso dont try to compare it12:30
DocScrutinizer05usually chip manufs later on commercialize any custom chip12:31
DocScrutinizer05like they did with LP5523 for example, which originally been custom built and designed to specs of Nokia12:32
rohi dont even know what that should be12:32
rohbut i find that its an urban myth that 'chips are made specifically for company XY'12:33
rohits always just a 'special variant' of socs the same vendor builds in generic purpose versions... like samsung does for apple..12:34
rohDocScrutinizer05: jup lp5523?12:34
DocScrutinizer05then why is the 110J1H0K not even known by google?12:38
rohwhat should it be?12:38
DocScrutinizer05(the OneNAND/RAM PoP used in N9)12:38
rohbecause its NOT AVAIL12:38
rohit isnt avail as part-> nobody gives a fuck12:39
DocScrutinizer05you'd think when it been available 2 or 3 years ago then google would know about it12:40
rohwell.. it was never 'available'12:40
rohjust because its sold special order from vendor-a to intregrator-b .. doesnt mean anything.12:40
DocScrutinizer05so you'd call that "custom built"12:40
rohbasically you can handle it as 'doesnt exist' .. makes things easier12:41
rohafter all.. its not our problem, and it will not be as long as we do not order >6 digit volumes of such circuits12:41
DocScrutinizer05yeah, which puts you take of "its an urban myth that 'chips are made specifically for company XY'" into the correct context which will make it self-evident ;-)12:42
DocScrutinizer05since those chips "don't exist"12:42
rohDocScrutinizer05: my point being.. those chips are 'made specifically'12:42
rohthey are 'produced and labeled specifically'12:43
rohin the end its the same features, maybe in a different 'setting' as the general purpose socs12:43
DocScrutinizer05that's not a soc after all12:43
rohyou maybe have one bank more ddr controller.. or trade a serial for a spi lane12:43
DocScrutinizer05it's a 512NAND/1GBRAM PoP chip12:44
rohso its more a 'different combination of soc components' but in the end the 'ip cores' are the same as in the general purpose chips12:44
rohheh. sounds expensive12:44
DocScrutinizer05why is DDR2 not working on a DDR1 bus?12:45
DocScrutinizer05(since there is actually a 1GB RAM-only 168p PoP chip)12:46
rohhttp://en.wikipedia.org/wiki/DDR2_SDRAM ?12:46
DocScrutinizer05but it has DDR212:46
DocScrutinizer05shrug, seen that12:46
wpwrakdon't they have different voltages ?12:46
DocScrutinizer05it seems the difference is longer burst mode 12:46
DocScrutinizer05yes, the volatge is a problem12:46
rohseems to be mostly clocks12:47
DocScrutinizer05clock is specified and identical12:47
rohdifferent voltages, sizes, clocks... enough to be incompatible12:47
DocScrutinizer05DDR2 prefetches one word (iirc) DDR3 4 or even 812:47
rohhuh? clock RATE may be identical12:48
rohdatarate is double12:48
DocScrutinizer05you always can run a RAM at lower clock speed btw12:48
rohthe physical interface is simply different12:48
DocScrutinizer05yes DDR means Double Data Rate12:48
rohread the text on ddr2 again12:48
roh" The key difference between DDR and DDR2 is that for DDR2 the memory cells are clocked at 1 quarter (rather than half) the rate of the bus. This requires a 4-bit-deep prefetch queue, but, without changing the memory cells themselves, DDR2 can effectively operate at twice the bus speed of DDR."12:49
DocScrutinizer05so what?12:50
DocScrutinizer05this is about internal operation, not about the bus12:51
rohread again12:51
rohits also about timing12:51
DocScrutinizer05I read it the 57th time now12:51
rohDDR2's bus frequency is boosted by electrical interface improvements, on-die termination, prefetch buffers and off-chip drivers. However, latency is greatly increased as a trade-off. The DDR2 prefetch buffer is 4 bits deep, whereas it is two bits deep for DDR and eight bits deep for DDR3. While DDR SDRAM has typical read latencies of between 2 and 3 bus cycles, DDR2 may have read latencies between 4 and 6 cycles. Thus, DDR2 memory must be operated at twice t12:51
rohwhat do you think that means for the guy implementing a memory controller12:52
DocScrutinizer05again, about internal timing, unless the 4 prefetch stages change the bus logic12:52
rohwell.. calling it bus is too much12:52
rohafter all its only a connection between the one memorycontroler and the one memory usually12:53
rohddr3 does loop back the clock afaik.. but i haven't read up about the details12:54
DocScrutinizer05on PoP it's usually 2 separate banks of RAM12:54
DocScrutinizer05I still wonder why DDR2 wouldn't work on a DDR1 controller12:55
DocScrutinizer05ok, it has longer latency, thanks to 4 bit prefetch12:55
rohfrom what i read in there its mostly timings and latency handling12:55
rohso.. yes, maybe the phy works, but the upper layers will not12:56
DocScrutinizer05probably you can't set the DDR1 controller to obey 4 cycles latency when not doing sequential reads12:57
wpwrakbig-bang the bastard ! ;-)12:57
DocScrutinizer05big, yeah, correct12:58
DocScrutinizer05this sucks so terribly12:58
DocScrutinizer05you got a SoC that supports 1GB addr space but you can't find RAM for that12:59
rohDocScrutinizer05: i guess you need to talk to sebastian about ddr timing details ;)12:59
DocScrutinizer05who's sebastian?13:00
DocScrutinizer05well, I'm convinced now that DDR2 won't fly on DM373013:00
DocScrutinizer05and after all that's what the docs say13:01
rohwell.. use a new soc13:01
DocScrutinizer05this defeats the whole purpose of the project13:01
DocScrutinizer05we're not designing a state of the art leete new product13:02
wpwrakthat's the problem when you get into these higher-end chips. the zone between bleeding edge and obsolescence is relatively narrow13:03
DocScrutinizer05binary compatibility to N900 is one design goal. Reusing the "proven" GTA04 design is another13:03
DocScrutinizer05wpwrak: ack13:03
rohm-/ binary compatibility13:03
wpwrakbtw, if shopping for cpus, you may want to look at the various linux boards olimex make. they have quite a collection of chips now.13:03
rohi can understand using prooven designs, tho13:04
DocScrutinizer05yeah, that's the problem that most people don't get it what's the concept behind Neo90013:05
wpwrakmany (most ?) of their designs are open13:05
DocScrutinizer05without binary compatibility the whole project goes poof13:06
DocScrutinizer05since the major selling point is moot13:06
DocScrutinizer05which is "use maemo5, it works and has existing apps"13:06
rohDocScrutinizer05: whats the issue with rebuilding?13:06
wpwrakand more recent arms are incompatible ?13:06
DocScrutinizer05roh: closed bobs13:07
DocScrutinizer05wpwrak: I've never checked but i'd guess the SoC is too different13:08
DocScrutinizer05not the ARM13:08
DocScrutinizer05ARM binaries work mostly fine still I guess, but the kernel API will differ too much13:09
DocScrutinizer05different GPU, different DSP, different peripherals and whatnot13:09
wpwrakthose bobs seems to go rather deep13:10
DocScrutinizer05the dual core ARM is probably our least problem13:10
DocScrutinizer05well, the 3D lib is just one of those closed blobs13:11
wpwrakhmm ...13:11
DocScrutinizer05again, I never looked into it indeep. But I consider it way too risky to try and port Maemo5 aka fremantle to a OMAP4 or OMAP513:12
DocScrutinizer05and OMAP5 isn't available yet, OMAP4 I know has nasty SiErr13:12
wpwraki'd plan for having some way to do things without such lose-lose dependencies. even if you succeed, you'll have a very fragile construct that can blow up anytime.13:12
wpwrakbesides, you rest easier with a plan B under your pillow :)13:13
DocScrutinizer05yeah, you'd plan for a 150 man project, porting debian to the thing and creating your own ecosystem13:13
DocScrutinizer05sorry, I realy have to run now, I'm 90min late meanwhile13:14
wpwrakwell, there it typically a path of reasonable compromise somewhere. that is, unless your system is basically closed and the open source between the blobs mere window dressing.13:16
wpwrakbut i hope that's not the case here.13:16
DocScrutinizer05regarding avoiding of lose-lose dependencies I probably should stop worrying about embedded and do something more rewarding, like growing crop13:16
DocScrutinizer05wpwrak: there's no feasible plan-b in vicinity13:18
paul_boddieThe backwards compatibility route makes sense for those wanting to prolong their N900 experience.13:18
paul_boddieBut a roadmap beyond the blobs would be cool.13:18
wpwrakhmm, that's bad. well, you already have a 77 man community. recruit 73 more and you have the 150 man team :)13:19
DocScrutinizer053 month ago I would've bashed newbieswhen they ask how to port maemo to another platform13:19
DocScrutinizer05paul_boddie: there is a roadmap for that. Slow migration to a set of completely RE#d blobs13:19
DocScrutinizer05we're already starting, since there is quite a bit that needs to get done nevertheless even for Neo90013:20
DocScrutinizer05I don't even dare to think how we should accomplish porting fremantle to an entirely new hw platform in <1 year13:21
DocScrutinizer05in 18 months though, things might have changed a lot and we *might* consider a OMAP5 design by then13:21
paul_boddieI thought so. Still, getting the wider community onboard is perhaps part of why Nikolaus is involved and helps to free up manpower for the eventual blob removal.13:22
DocScrutinizer05starting with that though is a sure and painful death13:22
DocScrutinizer05paul_boddie: of course13:22
paul_boddieWhere "wider community" means everyone not using Maemo but needing Linux to run on the device (which it does, of course).13:22
paul_boddieIt's a shame the NanoNote wasn't taken in a cellular direction somehow.13:23
DocScrutinizer05cya l8r13:23
paul_boddieAll the awkward stuff like charging and power management seems to be a solved issue, but then I wasn't really paying attention while people struggled with those issues. ;-)13:24
paul_boddieSee you, doc!13:24
wpwrakcellular alone is more messy than the whole nanonote :)13:24
paul_boddieYes, I lurk on various lists, and the modem control seems quite hateful.13:27
wpwrakwhat, no INVITATION ?15:12
larscI didn't get an invitation either15:28
qi-bot[commit] Werner Almesberger: poly2d/: store actual values in struct f2d, not pointers (master) http://qi-hw.com/p/cae-tools/626dbe516:35
qi-bot[commit] Werner Almesberger: cameo/stl.c (stl): update for poly2d API change (master) http://qi-hw.com/p/cae-tools/ae9ad0316:35
qi-bot[commit] Werner Almesberger: Revert "poly2d/p2d_hsort.c (p2d_hier_free): make freeing of polygons optional" (master) http://qi-hw.com/p/cae-tools/ebb7d0216:35
qi-bot[commit] Werner Almesberger: cameo/stl.c: generate real STL (WIP) (master) http://qi-hw.com/p/cae-tools/c5f54df16:35
qi-bot[commit] Werner Almesberger: cameo/fped2stl.pl: fped 2D stack to STL mesh converter (experimental) (master) http://qi-hw.com/p/cae-tools/96063cc16:35
larscwpwrak: I think there are test suites that can tell you how random your randomness is18:40
wpwrakyeah, there should be several statistical tests. my plan is to dump a few Gbits of randomness via USB (once i have that up) and then upload it somewhere. that way, others can play, too :)18:51
sanderrNeed a place to dump^whost it?19:14
wpwrakmaybe downloads.qi-hw will be fine. it has 18 GB free. i'll make the test data set maybe 1 GB in total. just some basic key material to play with. anyone who wants to run an in-depth analysis could (and should *) get their own device :)19:19
wpwrak(*) also to make sure we get some spread over the population of deployed devices19:19
sanderrHehe, indeed.19:22
larscwpwrak: have you already collected venture capital for the production? ;)19:33
wpwraki wish :) naw, first make sure the basic tech is right, then think about how to make a small "public" run. (maybe 100 units or so. incomplete software, crude case. etc.)19:42
whitequarkwpwrak: google "diehard"20:10
whitequarkit will get you both a decent test and a flurry of relevant better tests.20:10
ysionneau15:21 < DocScrutinizer05> in 18 months though, things might have changed a lot and we *might* consider a OMAP5 design by then < OMAP is a dangerous way20:20
ysionneauTI is now moving away from OMAPs20:20
ysionneauthey fired basically their OMAP teams20:20
ysionneauthey don't want to do such chips anymore20:20
ysionneauso support will be terrible20:20
wpwrakmsp430 fro, now on ?20:21
ysionneauthey want to move away from tablet and phone business20:21
ysionneauwpwrak: I think they will focus on analog stuff, radio stuff20:23
ysionneauand stop doing big socs20:23
ysionneaumaybe keep msp I don't know20:23
ysionneaubut OMAP is definetely dead20:23
ysionneauI just typed "omap + dead" in google just to see if I was not imaginating things20:24
ysionneaua lot of articles seem to say it20:24
ysionneauanyway, I've heard it from TI employees when I was at sophia antipolis20:25
ysionneaujust before they got fired with the rest20:25
wpwrakwhitequark: thanks ! and dieharder is even in ububtu. couldn't get easier :)20:25
wpwrakysionneau: would that extend to AMxxxx as well ?20:27
larscI think they are still doing automotive grade omaps20:30
ysionneauwpwrak: AM is kind of a copy of the core of OMAPs20:31
ysionneaubut it does not target phones or tablet I think20:31
ysionneauso maybe these are safe from the big cleaning20:31
ysionneaularsc: we are still buying automotives omap3 20:32
wpwrakysionneau: yeah, it looks like omap without the megalomania. a bit like freescale's i.mx chips20:32
ysionneauusually TI makes an omap chip, they target smartphones, then they target set-top-boxes and they end up targetting automotive (when they are fully qualified etc)20:33
ysionneauautomotive is like the "end of life" of those chips20:33
ysionneauas I've been told20:33
ysionneauand maybe that's where they sell most of them :p20:33
larscthe margin is bigger20:33
ysionneaua lot bigger20:34
ysionneauyou must know something about that :)20:34
ysionneauas I see it, for TI big digital chips is a nightmare20:35
ysionneauit cost a lot of money20:35
ysionneaubecause not only you need to design it20:36
ysionneaubut then you need to do software...20:36
ysionneaudo have BSPs20:36
ysionneauto give software support to your customers...20:36
ysionneauthey pay other companies for the support , but I bet it's expensive20:36
ysionneauwhen you sell analog chips, I think it's really less expensive and you almost don't need to provide any software20:37
ysionneausometimes a linux driver :p20:37
ysionneaubut it's not like porting linux to the SoC, with management of the whole clock gating tree, PLLs and all sillicons errata20:37
ysionneauand giving support about it :)20:37
apeletelarsc mth pcercuei: you were right this morning, about the UDC driver claiming the mmeory range for the usb device -> http://paste.debian.net/57578/20:39
apeletesince the driver is not being built in the kernel already, I guess I also need to remove the udc platform data code from platform.c20:40
wpwrakysionneau: one would think that all that should be a relatively minor task, considering all the other work that goes into making such a chip20:51
ysionneaubut for a chip company20:51
ysionneauwith strong hardware engineers20:51
ysionneautest engineers20:51
ysionneauvalidation engineers etc20:51
ysionneausoftware is just the nightmare20:52
ysionneauand the thing they just don't want to do20:52
ysionneauand they don't care about it20:52
ysionneauit's just a cost they don't sell it20:52
ysionneauit's really trouble, it's bugs, support to give etc20:53
ysionneauwhen there is no cpu core in your asic, no software, you're so happy :)20:53
ysionneauone big less trouble20:53
wpwrakbah, outsource it if you must. make sure the latest linux kernels runs on the critter and politely encourage all customers who want something else to reconsider20:54
ysionneauyour customer wont just need "linux to run"20:54
ysionneauthey will have very precise needs20:54
ysionneauand since they paid quite a lot for the asics20:54
ysionneauand have serious business to deal with20:54
wpwrakyes, let someone else fill in the blanks20:54
ysionneauand other customers20:54
ysionneauthey will ask you for support to fix the vicious bugs in the linux port20:54
wpwrakif they need extra handholding, have a list of consultants20:54
ysionneauthe one they trigger20:55
ysionneauand is not triggered by a trivial use20:55
wpwrakthe consultants in turn will want a nice clean generic kernel. when they have that, they can take care of the rest20:55
wpwrakyou run into all these problems when you stark forking. then suddenly all the world's combined misery becomes you very personal problem.20:56
ysionneaumy company kept opening tickets in TI customer service webservice20:56
wpwrakand of course, chip makers traditionally love to fork. and never merge things back. it's inevitable that this explodes. some dodge the bullet by simply no offering any support beyond the BSP. (they may still make helpful noises, but they're not going to really help)20:57
ysionneaubecause somehow you always find issues even in nice upstream TI BSP20:57
ysionneauand you don't email LKML for the fixes20:57
ysionneaubecause you need the fix soon and you need to be able to kick their asses20:57
wpwrakthe BSP should be the linus tree20:57
ysionneauso you call  TI20:57
ysionneauit's trouble for them20:58
ysionneauyou cost them money20:58
wpwrakmaybe make a branch for patches that haven't been merged yet. so your customers can take that. but never fork.20:58
wpwrakand yes, of course it costs them money. so they should make sure they're efficient.20:58
ysionneauand you have the countless cores in the chip, which makes countless drivers20:58
wpwrakmost of them usually don't change all that much from chip to chip20:59
ysionneauusb host, DSP, mmc controller, nand controller, ddr controller20:59
ysionneaucountless opportunities for driver issues20:59
wpwrakif you really have to redo everything from scratch each time, you probably have other issues :)20:59
wpwrakGPU is a good example. are TI's still closed source ?21:00
ysionneauon omap3 it's imagination stuff21:00
ysionneauit's powerVR SGX21:00
wpwrakah yes, as closed as possible21:00
ysionneauthe firmware yes21:00
ysionneauthe driver we have the source21:00
ysionneauI don't know if it's public21:00
wpwrakwell, that's one of the things you want to avoid. because if it's closed, you're the one people will go to for fixes. simply because you're the only one who can fix it.21:01
ysionneaubut at least my company have them21:01
wpwrakwell, that's the evil driver-under-NDA scheme. creates all the wrong incentives.21:01
wpwrakyou could fix it but you'd basically be doing TI's work. nothing better to motivate you at the beginnign of a workday21:01
wpwrakand you can't share it with anyone else21:02
ysionneauyeah indeed21:02
ysionneauusually we just add printf() to debug some issue21:02
ysionneauthen fill a bug report21:02
ysionneauto ask for a proper fix21:03
wpwrakwhich may or may not come ...21:03
ysionneauanyway I'm certainly talking too much :)21:03
ysionneauI probably shouldn't say all of this21:03
ysionneautime to go to bed21:03
wpwrak.. and then you need to explain why their stupi^H^H^H^H^Hfix doesn't solve the problem. and so on. been there, done that, it's hopeless. just stay away from closed source :)21:04
wpwrakheh, sweet dreams regardless :)21:04
ysionneauyou too, in your timezone shift21:04
wpwrakthanks :) the sun is setting, time to get up ...21:10
apeletelarsc mth: much better, now it's only complaining about not having DMA interrupt line -> http://paste.debian.net/57591/:21:23
apelete[    1.230000] musb-jz4740: glue kzalloc OK!21:23
apelete[    1.240000] musb-jz4740: platform_device_alloc OK!21:23
apelete[    1.250000] musb-jz4740: platform_set_drvdata OK!21:23
apelete[    1.260000] musb-jz4740: platform_device_add_resources OK!21:23
apelete[    1.270000] musb-jz4740: platform_device_add_data OK!21:23
apelete[    1.280000] musb-jz4740: init OK!21:23
apelete[    1.290000] musb-hdrc musb-hdrc.0.auto: No DMA interrupt line!21:23
apelete[    1.300000] musb-jz4740: exit OK!21:23
apelete[    1.310000] musb-hdrc musb-hdrc.0.auto: musb_init_controller failed with status -2221:23
apelete[    1.320000] musb-hdrc: probe of musb-hdrc.0.auto failed with error -2221:23
apelete[    1.330000] musb-jz4740: platform_device_add OK!21:24
apelete[    1.340000] musb-jz4740 registered21:24
apeletethat message comes from dma_controller_create() in drivers/usb/musb/musbhsdma.c:22:07
apeleteint irq = platform_get_irq_byname(pdev, "dma");22:07
apelete 22:07
apeleteif (irq <= 0) {22:07
apeletedev_err(dev, "No DMA interrupt line!\n");22:07
apeletereturn NULL;22:07
apelete 22:07
apeletelarsc mth: I chose to go with Mentor Graphic's Inventra DMA engine, but I don't know if that's a problem or not22:09
mththat's what the JZ4770 uses as well, iirc22:10
mthwhether it is applicable for the 4740 I don't know22:10
apeleteyeah, but there modifications done in JZ4770 code to use Inventra DMA engine22:11
apelete#if defined(CONFIG_USB_INVENTRA_DMA)22:11
apeleteif (musb->b_dma_share_usb_irq)22:11
apeleterv_dma = musb_call_dma_controller_irq(irq, musb);22:11
apeleteb_dma_share_usb_irq and musb_call_dma_controller_irq were added to take advantage of Inventra dma engine22:13
apeletemth: don't know how I should handle this one, any advice ?22:14
mthno idea, sorry22:16
apeleteokay, I'm going to sleep on this one, hoping to come up with some brilliant idea :)22:21
apeletemth: thanks for being there anyway, always help to have someone to bounce ideas with :)22:22
--- Tue Oct 15 201300:00

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