[Arm-netbook] Anyone here made a "TV computer"?

[Luke Kenneth Casson Leighton]

On Mon, Nov 7, 2011 at 2:16 AM, Bari Ari <bari at onelabs.com> wrote:
> What exactly is the reference design? A tablet or??

 don't know - usually they have "everything including the kitchen
sink".  i know that the little factory in china have obtained the
reference design, but they are incredibly busy.  also it doesn't help
that the "Great Firewall of China" is in the way (looovely).

> Why don't we get the hardware specs and provide better CAD files to the
> companies making boards with them.

 because they're only available under NDA from the SoC manufacturers,
who are at first somewhat shocked at the factory's complete lack of
software expertise and then "resigned" to dealing with the situation
by entirely developing a complete design - all on their own and
*including* the complete software package.

 this is the situation i'm working to break them free from, by putting
them in touch with Free Software Developers.

 this is the situation i'm working to break them free from, by putting
them in touch with Free Software Developers.



 exactly why, when this situation only *prevents* the SoC
manufacturers from selling their own SoCs, isn't clear.  i _have_ been
explaining this to them, but even the CEOs of these SoC companies in
China aren't actually allowed to make .. y'know... something called
"decisions"!

 they are effectively puppets, answering to their "superiors" (the
money people), who flatly refuse to let them do anything other than
what they have been dictated and authorised to do.

 so it is a cultural thing.

 we just have to work with the system as it is, for now, and show them
a better way, later.


> Part of the problem is inexperienced
> hardware 'engineers' generating some pretty awful boards that almost
> work.

 yes.  that's usually the smaller factories, and they usually succeed
only with things under 400mhz.  that's why that Skytone Alpha 400 took
off so well: Ingenic sold something like 25 million jz4740 CPUs before
MIPs caught up with them and went "oi! license! naoooww, sunshine" :)


  luckily, these guys in the little factory i have access to aren't
inexperienced [they just don't have any software expertise].  my
friend adam however can tell you some interesting stories about the
continuous cycle of experimentation he's witnessed :)


> Why put all this software effort into buggy hardware when we could
> also provide board stuffers with solid design files?

 there is more than one way to skin a cat, bari.  there is
"experienced design" and then there is "rapid incremental design".
there's a beautiful description in "A Young Lady's Illustrated
Primer", aka "The Diamond Age" by Neal Stephenson, in which
nanotechnology is used to seek out the missing copy of the Primer
(ultra-high-grade nanotechnology book).

 Dr X's nanotes are varied, often fail, but are so numerous that it is
obvious that they are designed by genetic algorithms and by
trial-and-error.

 the Neo-Victorians nanotes are clearly "engineered".  clinical,
calculated - and yet also fixed and rigid.

 we _know_ that evolution works, and we know that when evolutionary
algorithms are put into a fast spin cycle, the results are staggering.

read Ian Macleod's sci-fi books for references to the "Fast Folk",
where people are run inside "nanotech soup computers" that run 1000s
to 10s of 1000s of times faster than "real time".  eventually, they
evolve into a civilisation that discovers how to manipulate the
universe so that they can break out of the box.  the scientists,
knowing when roughly this will happen, nuke the box each time
destroying the nanotech computer "soup" before the civilisation within
it can reach that point (singularity).  civilisation genocide,
described casually "in passing" in this way, in later books, because
an entire book was dedicated to the subject some years ago (when they
_didn't_ nuke the nanotech civilisation and it broke free), earlier in
the series.

 and, in china, the factory PCB costs are cheap.  they can operate on
a fast spin-cycle.

 it's messy but _one_ of them - like brownian motion - occasionally
pops to the surface.  the point of keeping an eye on all these
factories is to catch the one that actually does a decent job,
snapshot their designs and put them into mass-production *before*
brownian motion sucks them back down, randomly, into the genetic soup.

 l.