Re: Anyone here made a "TV computer"?
On Thu, Nov 3, 2011 at 6:53 PM, Wookey <wookey@wookware.org> wrote:
> +++ Phil Endecott [2011-04-10 23:02 +0000]:
>> Dear All,
>>
>> For a while I have been planning to make some sort of "TV computer" i.e. DVB
>> recorder + DVD player + web video device etc.
>
> Did you make any progress on this Phil?
>
> There was a session as linaro connect on set-top-boxes with quite a
> few interested people, but no-one present who'd actually made it work
> usefully.
it's pretty hellish/specialist to get stuff compiled/running: last
time i tackled this it required several full-time weeks [for those
people familiar with how i work, you know that is one hell of a long
time].
about this time last year i was working on a very special type of
ARM9 (400mhz) that had 800mhz DDR2 RAM and a 1080p30 video decode
engine (turned out to be MIPS-based). the project was "python-tv",
and it was for the brazilian govt (indirectly).
the reason why python was chosen was because of the high bang-per-buck
ratio when downloading python scripts: most people in brazil are still
running off of pay-as-you-go GPRS so the last bloody thing they want
is to have to download a random pile of crap^H^H^H^Hjavascript _and_
pay for it. the plan was to do something like kde-look.org combined
with palm's webos but using python instead.
so, we compiled up webkit, and it turned out to be too slow or use far
too much RAM or both, for qt4 _and_ gtk2 _and_ enlightenment _and_ qt4
with directfb _and_ gtk2 with directfb... so please don't laugh or
look too shocked: we had to get webkit working *directly* with
directfb. it turned out that denis had just been sponsored to get
webkit working with directfb, so i helped him out.
then once that was working, i then added python bindings. then when
_that_ was working, i fired up pyjamas-desktop. when _that_ was
working it was possible to run, on a 400mhz embedded ARM9 (no, really
ARM9 not Cortex A9) an entire python-controlled HTML5 web browser with
ffmplayer directly built-in onto a framebuffer, no threads to get in
the way and bitch-smack that poor ARM9 between the eyes due to its
lack of 1st level cache coherency across thread-switching, and...
yeah, it all kinda worked.
i used openembedded because nothing else in the free software world
comes even _remotely_ close to providing the amount of
cross-compilable recipes in order to do the research and testing that
was needed. (if you recall, this was around the time that people were
deciding that i was trying to tell debian developers how to suck eggs
and that it was interpreted that i was telling everyone that they
should completely abandon everything that makes up debian and convert
entirely to openembedded. i really wish people wouldn't deliberately
assume the worst, it's getting really stale. *sigh* anyway....)
so added to the list of technology required you could put:
* directfb
* liblite-directfb (0.8)
* webkit-with-directfb
* angstrom-linux
* openembedded
* python
* pyjamas-desktop
the recipes are on assembla.com.
> Anyone interested should say so, or add themselves to that doc. It's
> not entirely clear what the most appropriate forum is for discussion
> of this subject. Suggestions welcome.
yep count me in because i have to keep an eye on this stuff. i've
been thinking whilst writing the above: as it's "cross-strategy" i
don't believe you'll find an appropriate forum. if you've read "the
strategy-focussed organisation" you're mad, but if you've read only
the first two chapters you'll appreciate in a very real way that it's
simply not the job of any of these projects to understand where and
how they fit into the "bigger picture". their job - their role - is
to get their heads down into the code in front of them, and "make it
work".
the person who recommended that book to me said that it's a very very
important question to be asking, "why should anybody care?". linaro
is focussed on strategy, wookey. you therefore need to be asking
yourself "why should anybody care about what it is that you're trying
to achieve?" _especially_ because linaro is tasked with cross-distro,
cross-platform, cross-pretty-much-everything optimisation.
i sympathise if you can't find an answer (or a suitable forum!) - as a
strategist myself i get pretty frustrated by not being able to think
of a good reason why people should care about cross-project
optimisations that i can see damn clearly! why do you think i've been
barred from so many free software projects?? people don't _like_
"strangers" wandering in, becoming familar in weeks with code that
took them years to understand and coming up with something that
expands the reach and horizon of quotes their quotes code, it pisses
them off and they accuse _me_ of having a humongous ego. mmmm :)
anyway, yes, please: do keep me informed, esp. if you find a suitable
forum: i deal with this stuff on an irregular basis.
thanks wookey.
l.
Reply to: