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

Re: [Fwd: Fedora & openSUSE meeting / cooperation ?]



On 2/6/07, Christopher Blizzard <blizzard redhat com> wrote:
Luis Villa wrote:
> Both of those are equally dreamy. The whole point of what opensuse is
> trying to do is make it so that people can do it once, distribute it
> over a large number of platforms. Having the maintainer of the package
> also be a co-maintainer in every distribution they want to distribute
> something through is insane, and just the kind of totally bullshit
> hoops that we must eliminate if we actually want linux to become a
> first-class platform.

Agreed.  Part of my underlying thinking of bundles is that they can
represent the lowest common denominator for _desktop_ functionality.
i.e. they support the gtk2 API/ABI, service activation via dbus, self
describing mime files anything else they carry around themselves.  Also
adding support to debian/suse/fedora desktop and then we can actually
create a pretty portable package.  We need better build tools to make
this happen, and I need to prove the idea out through OLPC, but it's a
good start for a real roadmap to success.

I wish you good luck and godspeed. ;)

> [This is only true if your vision of the future of Linux includes
> multiple distros; if your vision is that there is only the One True
> Distro, then yes, having upstream also be co-maintainers in the One
> True Distro makes sense. This is basically Canonical's vision. In some
> ways it makes sense- we gain a lot of efficiencies by having One True
> Kernel which everyone else forks from; why not have One True Distro
> that everyone else forks from. Of course, putting it in the hands of a
> proprietary tool makes me get violent.]

I haven't been particular shy about the fact that I want Fedora to
become that base.

To be perfectly honest, I don't think Red Hat/Fedora will ever earn
enough trust to be that base, or at least, will never unless it works
to actively piss off Red Hat and comes out the other side relatively
unscathed. And I *like* Red Hat- like you I wouldn't work for them
otherwise.

> It was the only way to do what we did, so we didn't have much choice
> in the matter. Implementation details aren't all that important, but
> basically it had a sort of meta-spec file and worked its way down from
> there roughly as you describe. I'd imagine opensuse's build farm does
> something similar, though AFAIK the two tools do not share any code or
> even common developers.
>
> It was of course wildly popular, because it solved exactly the problem
> we're discussing here, which is a very real problem for users.

Right.  But doesn't that seem to you like it's wedging into a system
that's fundamentally broken for your use case?

It is *very* broken, but not *fundamentally* broken, otherwise we
wouldn't have been able to do it. Would solving it another, less
broken way be better? Absolutely.

But Ximian was the proof that it could be done the horrible, broken
way; opensuse is trying to resurrect that (and I think has a
reasonable chance of doing it), and as of yet no one has actually
proven that it can be meaningfully done in the non-horrible, beautiful
way. Until then, working code (no matter how ugly and godawful) tends
to trumps pipe dreams.

Luis


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