On Sun, Aug 27, 2006 at 07:50:46PM +0200, Michel Dänzer wrote:
> On Sun, 2006-08-27 at 13:06 +0000, David Nusinow wrote:
> > On Sun, Aug 27, 2006 at 12:27:26PM +0200, Michel Dänzer wrote:
> > > You're saying we would have stayed with an older and clearly inferior
> > > upstream version of free software if we hadn't found a way to deal with
> > > a flaw in proprietary software?
> >
> > I don't know, you'll have to ask Steve, but that was certaintly a real
> > concern. You know as well as I do that people are going to use the binary
> > drivers whatever we do. Either way, it was a minor concession. We already
> > give people plenty of rope to hang themselves with X, adding another switch
> > isn't going to cause any more issues.
>
> I take it you're volunteering to shoulder some of the potential
> additional upstream bug triage burden then.
I really don't like binary drivers, but this is a bit harsh. It logs --
so we can look for it in logs and smack people down if so -- and there's
already a -ignoreABI command line option.
If David was adding this functionality for the first time, you might
have a point, but he's not, so I don't think it's fair to blame him for
his existence.
> > My priority is to support our users as best we can, and if it was a choice
> > between letting nvidia users hang themselves when they screw up or not
> > letting i965 owners be unable to run Debian without backports then it was
> > an incredibly easy choice.
>
> Are you seriously arguing that's the only significant upstream
> improvement in 7.1 vs. 7.0?
No, but in terms of driver support, it's the most visible, is it not?
Everything else pales in comparison to 'my chipset is totally
unsupported in release y.z, but works in this release'.
> > > Indeed, without any prior discussion.
> >
> > I didn't realize I needed to ask for permission for such things. If the
> > -ignoreABI option didn't already exist, I wouldn't have done so. I didn't
> > really add anything new, I just provided a new way to get at it.
>
> Once the option is in xorg.conf and things work, people will tend to
> forget about the option even when it's not really needed anymore. Then
> when the ABI breaks again, that gets ignored, potentially causing all
> kinds of weird behaviour resulting in spurious bug reports that will
> waste the time of and potentially confuse bug triagers, unless they add
> the option to the already too long list of gotchas to watch out for.
I don't believe that it's going to be a big problem. Users with
proprietary drivers are screwed anyway, as far as obtaining support now.
Hell, if you're _really_ bored, you could grep for this in all new
attachments, and add an abi-version-ignored tag for the bug. Adding
this to BZ wouldn't be immensely difficult.
> > Also, if you want to remove it from upstream, go ahead, but I'd like
> > to discuss it first.
>
> Ah, so removing it requires discussion, but adding it didn't?
Well, David didn't know it would be so controversial, I'm tipping, given
how inclusive he's been on -x in the past. Now that everyone knows it's
a point of difference (to say the least), it would be rude to get rid of
it without list discussion. Plus, general etiquette says that it's rude
to revert without discussion, surely?
> > I don't really love the option myself, but it'll very likely be a
> > patch that we have to carry around for a while in Debian if it doesn't
> > go upstream.
>
> I don't understand why, given that there is a compatible third party
> release now. Even without that, the Debian packaging of the older
> releases should have been able to arrange for -ignoreABI getting passed
> to the X server when necessary.
Right, that would be quite an easy change when it came down to it.
Indeed, you'd just need to add five characters ('0 && ') in the most
extreme case.
I personally have extreme disdain for proprietary drivers, but I don't
think it's right to jump on David for this. -ignoreABI was already
there, and the release team apparently explicitly requested IgnoreABI
as a prerequisite for supporting 7.1, which I think we can all agree is
a very worthwhile end. If you want to take the release team to task
for implicitly supporting binary drivers, I don't think anyone would
begrudge you.
'Can't we all just get along?',
Daniel (reformed Ubuntu nvidia-glx/fglrx maintainer)
Attachment:
signature.asc
Description: Digital signature