[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: 4KSTACKS again



On Tue, 13 Apr 2004 16:08:00 -0300, Ben Steeves <bcs metacon ca> wrote:
> NVidia is, of course, at fault here by not a) releasing their drivers as
> source and b) not fixing this problem sooner.  But just 'cos NVidia is
> dragging their heels is no reason to penalize their users AND hurt the
> testing process by forcing those users to compile their own kernels
> (thereby invalidating their systems for kernel testing).

Its not hurting the testing process at all. Its a matter of priorities
for the developers. The 4kstacks option NEEDS testing, its a
trade-off.  nvidia breakage has happened before in rhl betas and it
will happen again. Since the kernel developers can't fix the nvidia
driver, its a trade-off in terms of other development priorities and
continued nvidia compatibility. It makes absolutely no sense to
continue to hold back on performance enhancing kernel features because
nvidia's kernel code is buggy. It sucks, but thats the reality. Kernel
development must move forward, if the smart people think 4kstacks is
worth turning on and worth testing for its own particular
benefits..then its worth testing. We can't hold up progress because
nvidia's driver is buggy and nvidia isn't moving as fast as kernel
development.
 
I think testers who are requesting this kernel patch be reverted need
to understand the point of signing up and using a test release. It is
NOT to have a fully functional and stable system. The point is to test
the features the developers think need testing. Downgrading kernel
enhancements becuase of a buggy closed source driver...sort of defeats
the point. Testers should expect to install a test release and watch
it instantly meltdown their computer, and should consider anything
better than that a blessing.
Test releases eat babies. The nvidia driver, is just one of the many
juicy suculant babies the test process will devour. This and worse is
to be expected.

And i think testers new to the testing process with Fedora Core who
have not been a part of several previous rhl betas need to understand
that the nvidia driver has in the past had problems on rhl beta
releases because of new kernel changes, this is not something out of
the blue new. I absolutely expect the nvidia driver to break at some
point in every testing phase. And nvidia has in the past specifically
said..they do not track beta distros. Nvidia will fix this on their
own schedule, which if the rhl history is to be a guide, is about 3
weeks after fc2 is released. It sucks...its life..and its completely
appropriate for fedora core, which has an objective of developing an
open source OS. Holding back on ANY useful kernel level feature
becuase a closed source add-on kernel driver drags its feet is just
not appropriate. Especially if you consider that the move to 2.6
kernel from a 2.4 kernel results in hardware support regressions for
hardware that was supported with open source drivers under 2.4. If we
can't even prevent open source hardware regressions...even pretending
to worry about closed source hardware drivers is absurd.

-jef"owns nothing but nvidia hardware in his modern desktops...and has
no problem with losing nvidia driver support during a testing
release"spaleta



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]