On 07.12.2014 01:41, Vlad wrote:

> While I agree with the "no closed-source binary blobs" philosophy on
> principal there is the real fact that completely free and open-source
> software written by enthusiasts just might not cut it in some
> situations, the blade server example is just one of those.

IMHO, the correct solution to this problem is to put these
manufacturers on a big-fat blacklist. Give everybody out there
a loud and clear warning, not to buy suh crap hardware.

Yes, it's political issue, not a technical one.

> Furthermore there will be obscure equipment with very limited usage
> outside of a few niches,

Do we need to support them in our primary repos ?

> and we must not forget that the developers working for the manufacturer
> will have both earlier access to the equipment, better knowledge of
> the operating specifics, and that they will be working on the
> drivers and such full-time.

They should instead publish proper specs an driver source code.
Otherwise the manufacturer should be blacklisted,

Binary-only drivers, IMHO, are an absolute no-go. The usually miserable
quality isn't the only reason. Also, the source they usually publish
(eg kernel patches) is of miserable quality.

I'm currently dealing with imx53. The kernel patches from freescale and
amd are horrible crap. And, of course, no source for the gpu driver,
just some xorg driver .so's (for fat glibc, even worse: softfloat).
In other words: GPU is completely unusable.

> And while it would be great if everyone could share their code, again on
> principal, in the real world there are such things like the desire for
> return on investment, protection of intellectual property and other
> trade secrets and sheer stubbornness on the part of the manufacturer.

Such manufacturers, which consider interface specs, needed for driver
development, an "intellectual property" aren't worth a single penny
of customer's money - let evolution eradicate them.

--
Enrico Weigelt,
metux IT consulting
+49-151-27565287