Re; 4KSTACKS again.

[Andy Ross]

Well, I got your attention, so I'm going to post one last time
(really, this is the last one, I promise) in the hope that you might
change their mind.

Dave Jones wrote:
> The config option in 2.6.6 will be going away.
> Read the comments from the commit message...
>
> : At present 4k stacks are selectable in config.  When the feature has
> : settled in we should remove the 8k option.  This will break the
> : nvidia modules.  But Fedora uses 4k stacks so a new nvidia driver is
> : expected soon.

Well, just to poke a little bit at your logic: this is circular.  You
can't cite future kernel configuration defaults as justficiation for
making an early change when the kernel is citing Fedora's change as
the reason for the new defaults.  And similarly, I'm not sure if the
text "When the feature has settled in" is quite synonymous with "in
2.6.6".

Honestly, we're on the same side.  My fear is that if NVidia doesn't
come through with working drivers in a timely manner after the FC2
release* then Fedora will end up with a black eye, with expert users
unable to recommend it and a "Fedora doesn't support 3D" meme
spreading in the community.  I remember how many projects refused to
accept bug reports from Red Hat 7.0 users because of the custom gcc
version, and this sounds like a similar situation in the making.

Now, maybe the idea here is activism: using the FC2 kernel as a
mechanism to encourage NVidia to open their drivers and/or change
their architecture to make the kernel module GPLable.  And I'm not
sure I would even disagree with that.  But until the mainline kernel
and/or other distributions start pushing this change too, it seems
awfully dangerous to me.

Andy

* Remember they have a brand new GPU architecture to support in their
  next release.  And auditing a big (!) source tree (which is itself,
  I gather, mostly a port of their WinXP driver) for stack overflows
  is not a trivial task; I get the feeling that some of the non-redhat
  posters are under the impression that all NVIDIA needs to do is
  recompile the thing.**

** Some people will no doubt be tempted to post more flames railing on
   about the stupidity of this choice of driver architecture.  Please
   (please!) don't bother, I agree with you.  I'm just trapped with a
   different set of priorities.