On Tue, Jan 20, 2004 at 12:52:38AM +0100, Michael Banck wrote:
> > > I guess it's pretty clear what needs to be done in case Andrew's
> > > proposal passes, no? We've got the nonfree.org domain and we've got ten
> > > years of experience with hosting Debian packages.
> > What's this "we" ? Please speak for yourself,
> Fair enough, sorry.
There's no need to apologise, it's just that when you say "we", you
kinda obscure the fact that there isn't anyone who's come forward who's
actually willing to do this.
> One idea was to modify katie just so slightly that she installs non-free
> packages into /org/foo.org instead of /org/ftp.debian.org. That probably
> would't go far enough for people in favor of dropping non-free I
> suppose.
Personally, I think it'd be silly. Likewise having non-free.debian.org
or similar -- it doesn't save *any* time or effort or make maintaining
main any easier (which outright dropping non-free would, as, to a greater
extend, would dropping non-free and contrib), and it doesn't make the
separation between Debian's free distribution (main) and the non-free
work we do significantly clearer. It'd also make mirroring just that
little bit more awkward.
> > Personally, I'm finding it pretty hard to work out what I'd want to
> > work on should this GR pass -- can I put up with crappy, contrib-style,
> > third party non-free stuff well enough that I can avoid having to do
> > a whole lot of boring make work to reimplement various bits of Debian
> > infrastructure?
> I don't think you need to reimplement Debian infrastructure in order to
> duplicate it, just to adjust it. Of course, you're more expert than me
> to comment on this.
Re-roll-out? Whatever. You don't have to rewrite it, but you do have
to get new machines, and set it all up, and maintain it, and patch the
systems, and track upstream and all that other stuff. There's a lot of
effort there, and it's pretty boring, and given it's just for non-free
stuff, it's pretty low value -- certainly compared to doing the same
work for the main archive.
> > Would maintaining nonfree.org actually be significantly simpler than
> > just forking Debian entirely? Probably, but how much?
> What do you mean by 'forking Debian' anyway?
Reimplementing all of Debian, and maintaining/distributing free packages
as well; eg glibc, dpkg, etc.
Again, while there's not that much you have to rewrite, there's a lot
of work in setting up new infrastructure and so forth, but as well as
being able to keep legacy apps working, you'd get to experiment with
interesting new developments without having to worry about convincing
other maintainer's they're worthwhile, or having an established userbase
to support.
How's that compare to just setting up nonfree.org? It's certainly
more effort, but is it a lot more? I dunno. It'd certainly be more
interesting than just maintaining stuff for non-free software for me,
but enough more that it'd justify the extra effort? I don't know that
either. Would either of them justify the effort? Dunno.
> I wouldn't consider
> outsourcing less than 200 packages (forgot the exact number) a 'fork'.
> The requirements for infrastructure and maintenance are considerably
> lower than for a full-blown fork of Debian, IMHO.
I'm not really convinced. If you're going to have it work as well
as Debian, you need to have an archive and a bug tracking system and
probably some mailing lists. If you're going to have it be centralised,
as opposed to lots of independent apt sources, you need to have signed
uploads, and some way of verifying the people who send you keys are who
they say they are, and, ideally, aren't grossly incompetent. I don't
think any of the non-Debian apt repositories satisfy these requirements,
which, I think, means you're expecting debbugs, katie, lists.d.o, and
n-m to be duplicated. Once you've done that, you're most of the way to
having all the infrastructure you need to maintain your own distribution.
I mean, it's not everything Debian does by a long shot, but it's all the
core stuff, afaics.
> I don't imagine any of those options to actually happen, I'm just trying
> to figure out a transition plan in case the GR passes.
I don't understand. One of the options was "reimplement Debian's stuff
for nonfree.org". I thought you were claiming that was likely. Maybe
this is just a hangover from the "reimplement" confusion.
> IIRC, the biggest
> (factual) argument against Andrew's argument was exactly this: 'There's
> no transition plan. What will our users do?'.
>
> But I see that we're running in circles. The 'don't drop non-free'
> people just say 'd'oh, don't drop it and you don't need a plan', while
> the 'drop non-free' people don't care about it anyway.
Uh, dude, if you're arguing for a change, it's your responsibility to
rebutt the arguments against the change. If your rebuttal is "You're
right. There's no transition plan. Our users who rely on non-free
packages can switch to free software, switch to a different distribution,
find someone else to support them, or just live without it. Either way,
we're not going to help them." then, well, fair enough. If you're going
to say "Our users will be able to go to nonfree.org", then it's *your*
responsibility to set that up and make sure it works -- not the people who
are telling you that that's a stupid idea because it's way too much work.
Cheers,
aj
--
Anthony Towns <aj@humbug.org.au> <http://azure.humbug.org.au/~aj/>
I don't speak for anyone save myself. GPG signed mail preferred.
Linux.conf.au 2004 -- Because we can.
http://conf.linux.org.au/ -- Jan 12-17, 2004
Attachment:
signature.asc
Description: Digital signature