30 Nov
2004
30 Nov
'04
8:50 p.m.
On Tue, 2004-11-30 at 15:09, Fons Adriaensen wrote:
On Tue, Nov 30, 2004 at 05:32:30PM -0500, Dave
Robillard wrote:
It depends on how Y and Z interact with the kernel itself. There has to
be some sort of driver that will handle the card. If that driver is kept
open then I would use Y or Z, no problem (I'm already doing that with
any card that has firmware in it, whether I'm aware of it or not). If
that interaction is closed source I would try hard to use neither
because the kernel internals will change over time, and at some point I
will end up with hardware that has a non-working driver and no way to
fix it.
Nvidia tries to do that with their closed source drivers which have a
pure binary and some glue source code you have to build for each kernel.
Even then the approach sometimes fail, as when Fedora started shipping
kernels using 4K internal stacks and the driver stopped working.
-- Fernando
I'd rather not have people with this ignorant
'closed drivers = good'
opinion turning Linux into yet another Windows. If you want a sh***y
proprietary OS, there's plenty to choose from already. Do the rest of
us a favour and don't advocate turning Linux into yet another one.
I will follow you reasoning up to a point.
Suppose we have a card X with a closed source driver. You will not
use it. Now we make a new card Y, which is actually an X with the
functionality of the driver pushed into the (closed source) firmware.
We make a third card Z with the same functionality pushed inside the
hardware.
Will you use Y ?
Will you use Z ?
If the answer to either is yes, what is the essential difference ?