Re; 4KSTACKS again.

[Andrew Farris]

On Wed, 2004-04-14 at 11:36 +1000, Dan wrote:

> <cough> I'm sorry - no-one says nVidia or ATI or anyone else *have* to do
> anything. If I've spent my hard-earned development dollars on a driver
> that - admitedly being closed source - actually WORKS, and the kernel goes
> and changes such a fundamental thing as to completely break things, with
> absolutely no way to config-option things to a working state again, then
> I'm affraid I would say "stuff you too". What motivation is there for
> nVidia to go BACK to the drawing board and re-do all their drivers?

They provided, although in some part helped by Zander (www.minion.de),
working 2.6 support when the time came.  It is in their best interests
to continue the development -- I am fully confident that they will
provide support for major changes in the kernel when the time comes..
but it would be foolish to expect them to have made this change prior to
the kernel being released with the change (or too quickly afterward).

> I just think it's a bit rude to go breaking compatibility, with NO
> OPTION to back out of the compatibility change, and then have the attitude
> of "well, too bad. if you wanna run with working 3D drivers, go use
> Windows".

Noone suggested that you use Windows as a solution to this simple
problem, and jumping to that conclusion (although probably used for
exaggeration...) indicates that you should not be involved in the
testing process.  The simple solution is to use the kernel that works
when 3D is required, until the driver is fixed to work with the new
kernel technology.  This is precisely what was required during the
entire kernel 2.5.x development cycle and early into 2.6.x (run 2.4, use
3D, test 2.5.x, go back to 2.4 when required).

This entire argument over whether or not the downstream kernel
maintainers should have left the OPTION rather than adopt it 2 or 3
weeks early is ridiculous.  This short timeframe does nothing to fix the
real problem, that you will not be able to use the driver when the
upstream kernel merges the stack change.  Rather than belabor this
point, testers should simply remain on the last kernel to work until the
driver is fixed -- and get back to testing non-kernel related issues.
Obviously those who do this can no longer test the latest kernel in
combination with the driver.. but they can still test the kernel, then
use the old one for their 3D acceleration needs.  I played Americas Army
Ops for hours this afternoon on my test system, using the nVIDIA driver.

It comes down to this: if you cannot live without the 'closed source'
drivers for just long enough to test the new kernel, then you are not
testing.. you are using.  nVIDIA has nearly a whole month left to fix
this problem before FC2 is released... and it is actually a bit rude to
expect the driver to work by then.  Give the engineers over at nVIDIA
some slack, the drivers work very well on a great number of linux
systems with a great number of differences.
(personally, I think open sourced they would work better, but not our
decision to make)

And lets give the Red Hat engineers equal slack, the kernel is
changing... they change with it because that is ultimately what this
project is about (if in doubt, read up at www.fedora.redhat.com).

-- 
Andrew Farris, CPE senior (California Polytechnic University, SLO)
fedora at andrewfarris.com :: lmorgul on freenode
"The only thing neccessary for the triumph of evil is for good men
to do nothing." (Edmond Burke)