On 4/3/08, Wolfgang Spraul <[EMAIL PROTECTED]> wrote: > Dear Hervé, > here is my perspective: > Most chip vendors see their business in selling chips. Documentation is just > a necessary evil to them, they are trying to get away with the minimum > amount of documentation that will still sell the chip. > Unless in very few cases, chip vendors do not see good documentation as a > strategic asset that will help sell their chips. Maybe down the road we are > lucky and Intel becomes a vendor that sees documentation like this, but I > will believe it when I see it. NXP also came around to us in a very nice > way. > We would like to publish documentation for the Toshiba ASIC in our LCM, very > hard with Toshiba (I'm not complaining, it's a big company and we are a > minuscule customer). > Samsung seems to be going closed, even though they joined the Open Handset > Alliance and are a big supporter of Android! > > Why that? Well, let's think from their perspective: Again - they are selling > chips, not books or PDF files. > In the case of Samsung, the legal department may look at a given PDF file > (say 1000 pages long) and see LOTS OF RISKS! When their lawyers read this > document (and they won't understand most of the technical stuff in there), > they are very concerned that the document will provide grounds for lawsuits > against Samsung later on. > If they just sell the chips as-is, those risks are reduced. > Plus they will say "why do we have to release THIS particular PDF?" Why not > a much shorter version, say a 2-page high-level overview, which the legal > department can carefully check word-by-word before release? And if it has to > be this specific PDF, why not release even much more? Samsung certainly has > another 100,000 pages documentation for each chip, internally. > > If you think about it from their perspective documentation is a very random > thing. You cannot easily convince them that if they release a 1000-page PDF > file about the say 6400 chip, they will sell this many more chips compared > to just releasing a 2-page PDF file. > > So we at Openmoko need to be smart, and accept realities out there: > > ---1 > The current model: We try to convince vendors to open up documentation to > the public, ideally allowing us (or even better everyone else) to > redistribute the documentation. Like Intel is doing with Creative Commons > now. > > ---2 > We can try to 'buy' chips+documentation, make the PDF file part of the > purchase. We would then put the PDF file behind a click-through license, > which says that the PDF behind the click-through license is just part of the > Neo product, and does not guarantee product behavior. The legal > effectiveness of such a click-through license is debatable, and we would > still need the vendor to like the idea and agree to give us documents under > these terms. > > ---3 > We can sign traditional NDAs and alert our vendors that we are legally > hiring respected FOSS engineers on a nominal basis (say 1 USD/month), in > order to give them access to the documents we have under NDA and allow them > to write FOSS software same as our traditional, fully-paid engineers can. > Again we could only do this with vendors who understand what we are doing, > trust us, and generally agree to the idea. We would not mass-hire thousands > of people this way, say having a form on the web where you can 'hire' > yourself, then download all docs. It all has to be reasonable and ideas and > intentions must not be ridiculed. But I could imagine that this is doable, > first with a few selected people, later maybe dozens or even hundreds of > people? The bigger we make this the more our own legal department will get > concerned :-) > > ---4 > We can become much more aggressive in documenting our source codes. Most > vendors would actually like that! Remember what I said above that the legal > departments see documentation THEY publish as a risk! But if we publish > something we wrote ourselves they usually don't care, in fact they like if > someone does free work for them. We can say whatever we want about a > vendor's chips, worst case we will get sued, not them :-) So maybe we should > just go ahead and EXTENSIVELY document source codes, to the point that you > basically have long lists of well-documented defines in header files, long > commented-out texts describing certain chip behavior, more or less based on > what we read in the documentation (just rewritten), etc. Same as always, we > would only do this with vendors that understand & agree to this, but as with > #2 and #3 there is actually a good chance many vendors would be supportive. > > There is no perfect solution to cover all cases. We need to work case by > case (vendor by vendor) to open up documentation, so that we Free The Phone, > as we have set out to do. > I see a big tendency to write 'pseudo' open source codes, where it's > nominally written say in C, but actually it's just long lists of writing > magic 32-bit values into magic memory registers. This is not much different > from having a binary driver outright, in fact a quick run of a binary driver > through IDA Pro might then result in the same type of C 'source code' that > you have otherwise. > In that case 'open source' would be meaningless. > This is what we all want to avoid. > Open Source must mean that every individual out there can UNDERSTAND how the > chip functions, and how to hack the chip in such a way to make interesting > new things happen! > Hope this helps, Best Regards, > Wolfgang
I really like this idea. I think Openmoko can trust a person hired from Trolltech, and I also think the chip vendor will. But it is only one way to figure it out. -- Please don't send me Word or PowerPoint attachments. See http://www.gnu.org/philosophy/no-word-attachments.html Join the FSF as an Associate Member at: <URL:http://www.fsf.org/register_form?referrer=5774> _______________________________________________ Openmoko community mailing list community@lists.openmoko.org http://lists.openmoko.org/mailman/listinfo/community