On Mon, Aug 24, 2009 at 07:02:18PM +0200, Milan Jurik wrote:
> >Another question that may arise is the architectural status of /contrib
> >(e.g., can the ARC bless integrations into /contrib in some manner that
> >confers, say, protection to filesystem namespace camping in /contrib?
> >do interfaces in /contrib have to advertise stability attributes?
> >etcetera).
> >
> >IMO: Integrate it all, and let users sort it out (a paraphrase of a
> > Spanish inquisitor quote) (e.g., with a voting scheme). If an
> > i-team can't commit to interface stability, then make all public
> > interfaces Volatile, and mark the package as "toxic interface
> > stability", see if users still want it :)
> >
> >
>
> And who will sustain such thing for the next 10+ years? ARC is looking
> at effectivness Sun Engineering resources. Pushing something to
> repository does not mean end of issue, it is the cheapiest thing.
IMO: We can commit to something better than Volatile if we have frequent
releases that allow us to break backwards compatibility for such
interfaces. Arguably OpenSolaris is just that (though, of course,
that's yet to be seen for sure).
Interface stability statements set expectactions, but making
everything Volatile is kinda useless. Breaking random things every
couple of weeks is equally bad. But posting notice of impending
breakage, say, 6 months' ahead of it, would be a lot better than
either. "FOSS Xyz is Uncommitted, except that if the upstream
community changes their mind and breaks backwards compat we'll
follow suit in a subsequent OpenSolaris release, with N months'
warning posted on xxxx-discuss at opensolaris.org."
Banishing would-be Volatile stuff (e.g., because we have no i-team
committed to supporting it) to /contrib won't absolve us when the
upstream community breaks compat and we blindly ship the broken
version. With infinite resources we could just make everything
Committed. But that's not going to happen.
Nico
--
