[Arm-netbook] Debian GNU/Linux on tablet hardware

[Luke Kenneth Casson Leighton]

On Fri, Oct 28, 2011 at 7:34 PM, Rob van der Hoeven
<robvanderhoeven at ziggo.nl> wrote:
> On Fri, 2011-10-28 at 17:01 +0100, Luke Kenneth Casson Leighton wrote:
>> On Fri, Oct 28, 2011 at 3:51 PM, Bill Gatliff <bgat at billgatliff.com> wrote:
>>
>> >> > I think it must be possible to buy an android motherboard for just a
>> >> > fraction of the price that i paid for my tablet.
>>
>>  only if you're prepared to place a cash up-front order for a large
>> number of units, because you're dealing "direct" with the factory.
>> the exceptions to that, that i know of, will be companies like the one
>> i'm dealing with (because they trust me), and franson who runs
>> quickembed would almost certainly be happy to sell you an
>> off-the-shelf motherboard from something he's designed (or licensed)
>> already.
>
> What is needed for the FreedomBox is hardware that is already used by a
> major brand. This hardware can be more trusted with regard of quality
> and availability. It is not needed to place orders before testing. We
> can just get the motherboard from a device that we buy. If things go
> well, then we can contact the manufacturer.

 rob - you're not aware of the realities of the situation behind the
factories.  the factories DO NOT HAVE the source code.

 i repeat.

 the factories do NOT have the source code.

 you're therefore asking too much, and expecting too much, on several counts.

 not only do the factories not have the source code, but they don't
actually do the full designs themselves.  they *modify* designs (just
like the guys at quickembed modified the beagleboard "Open Source" PCB
designs to create the DEVKIT8000 product).

 this is why certain companies can supply you with samples at very low
prices, whilst others want $150,000 up-front before you even start.

> I don't think we should try to specify our own hardware. Too risky for
> me. We should liberate existing hardware from the android :-)

 rob, feel free.  you will then end up with a device that is not only
completely unsuited to the task, because it's in the wrong
form-factor, but by the time you've finished negotiating (or
reverse-engineering) it will be end-of-lifed.

 or, worse, superceded by a cheaper chipset, which *causes* everyone
to run like hell towards it, abandoning everything else that's even
vaguely comparable.

 however: the point about needing testing is valid (in some ways) and
not valid in others.

 the ways in which it is valid is if the SoC vendor themselves makes a
mistake.  this has actually happened.  three times that i know of.
the first was with the OMAP3530 (beagleboard).  the second was with
the sPEAR1310.  the third was with the OMAP4440 (pandaboard).  ask
around: those people who got early pandaboards had to be provided with
replacements.  in each case, there were errors *in the CPU*.

 now, this is easily "solved", by picking a CPU that is in
mass-production that has, clearly by virtue of it being in
mass-production, has no errors.  if it does, you can sue the SoC
vendor.  so that's not a problem.

 another way is that the factory makes a hardware mistake.  this is
"their problem" and "their responsibility to fix".  given the amounts
of money that they are likely to make, if you're "in early" and are
therefore helping them out to do software development, they will go
out of their way to get you a replacement board (revision N+1),
because you, in return, are helping them [remember: normally they have
to pay cash up-front for the ODM ready-resigned PCB as well as the GPL
software _and_ get themselves into GPL Hell over the Linux Kernel
_and_ the OS!]

 so _that's_ not a problem.

 however that does leave the case where there are people who would
just like to buy very-early-revision hardware, which may or may not
work, but who do not want to actually work on the software.  such
people i am *not* interested in.  at all.  if you are one such person,
who is *solely* interested in obtaining early-revision-hardware and
who does not want to help get the product working, you need not read
any further.

 so leaving out that small group of people, we can discard them and
move on to the next group.

 we're now into the scenario where not only are we using a "reliable"
SoC, but also that small group of dedicated people have worked on the
software enough to be able to "prove" that the hardware works
reliably, and we can now proceed to the scenario of offering "beta"
versions to a wider group.

 this is all absolutely standard practice, btw - goldelico are
following this procedure to the letter, for the GTA04 development.

 so let me rephrase the question.

 if there was a "heavily-beta-tested" PCB (as illustratively defined
above) which, in mass-volume would be around the $15 mark, was made
available with at least one working linux kernel and at least one
GNU/Linux OS, at an approximate guide-line price of $70 to $80 for the
first 100 and potentially dropping even further as sales picked up,
would anyone be interested enough to commit publicly to saying "yeah i
could go for that", such that i can then show the factory that there
is interest, such that it is not necessary to pay them up-front, cash,
for the PCB development?  oh, proviso, of course: only if there are
enough *other* people interested.

 also please bear in mind that the design i have in mind will be
EOMA/PCMCIA compliant.  thus, far from being "problematic" if the CPU
card is quotes wrong quotes, actually it can just be... replaced.
and, precisely because it is EOMA/PCMCIA compliant, the exact same CPU
card (which is where most of the risk and hard work is) can be plugged
into a mass-produced tablet, or a FreedomBox chassis, or a NetTop,
or... whatever.

so, if you are in that group - who would like a heavily-beta-tested
PCB - please do say so, publicly, so that i can point the factory at
the responses.  if you know of anyone or any group who would also be
interested, please do also contact them, asking them to either send me
a message with permission to pass it on to the factory, or ask them to
post to the debian-arm list, or subscribe and post to arm-netbooks.

remember that this is about gaining traction.  traction cannot be
gained if there is nobody else willing to commit publicly in order to
gain traction.

i can, if you prefer, use a site like "pledgebank.com" - that might
help people to understand that i'm not asking people for a
"committment to buy".  i'm asking people if they would be *willing* to
commit to buying, if a certain threshold of other people also publicly
made the exact same committment.

btw if anybody else would like to take over these negotiations, please
feel free to make yourself known and i will pass responsibility on to
you.

l.

(EOMA/PCMCIA specification:
http://elinux.org/Embedded_Open_Modular_Architecture/PCMCIA )