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

Re: Some proposals

Jaldhar H. Vyas wrote:
  > Proposal:  Stop releasing altogether
  > sid would be the only Debian distribution.  Other
  > companies/interested groups of people would be responsible for
  > putting out finished products on whatever schedule suited them.  The
  > debian projects role would be to produce the packages, and provide
  > infrastructure (BTS, policy, installer.)

What about this, providing Debian is really considered too unwieldy at
this point:

Debian as a project just releases core functionality:

* installer
* kernels
* glibc
* gcc
* key development library packages
* X (or maybe not)
* some important shells
* some important editors (maybe)
* some basic window managers (very maybe)
* some key interpreters (Perl; let the flamefest begin)

*Everything* else is subprojects, which actually would just be (sets
of?) unofficial debs that track official releases.

Perhaps this would speed up releases and take stress off the core
developers.  In other words, why the hell should key people in Debian
stress themselves in any way shape or form with any aspect of QT3 or
mplayer packages?

(Why should *anyone* worry about this QT3 debacle?  If someone wants
Debian packages and thinks Martin was doing a bad job, just fork the
packages, and the folks that wanted the fork could get it.  I'm not sure
forking is bad if we're talking about non-core projects).

Also, the people who want to work on packages would then not necessarily
need to be DDs (which might knock out this problem with long wait times
for prospective DDs).   Yes, that sounds dangerous.   But if I download
and use, say, Postgresql, it really relates to my trust of the
Postgresql developers and Oliver Elphick a lot more than the Debian
project as a whole.

And if I trust Oliver Elphick to be able to package the latest
Postgresql against the latest stable release, why in the world should
*he* be waiting on some glitch in unstable glibc relating to mips
architecture?  [1]

I am not denigrating the importance of integration testing by any means,
but suggesting that it be moved down to the level of a package, its
maintainer and *his* users.  I think the core Debian project people
sometimes get trapped in the mindset that these non-core packages are
still somehow theirs, but that's not really a user's view.

Yeah, security, yada, yada.  How many times do security fixes come from
Debian non-maintainers of the package in question, versus upstream and
the responsible package maintainer?  (I actually don't know).

This plan might even encourage upstream developers to start making good
Debian packages on their own, in some cases.  Taking out the middleman
is a *good* thing, if the quality does not suffer.  It might well
improve.  See QT3.  Absolutely *key* here would be making good Debian
packaging so easy that many upstream developers will not find it
onerous.  (Though I am glad he exists, should there really need to *be*
an Oliver Elphick?  Actually this is probably a bad example, since
Postgresql upstream doesn't seem to do the work of seamless database

Then, if this turns out to not be sufficiently lean, start shedding the
ports.  :-)  But that's a joke: I believe ports are important, for the
quality and prestige of the project; it's *vertical* scope that maybe
y'all should think about.
  > Proposal:  Make all the people who talk but aren't going to do
  > anything shut up.
Well, that would be me, since I'm a non-DD.  But I can dream, can't I?

[1] This, that I just saw on debian-user, is the kind of thing that
should *never* happen, in an ideal world:

Jaldhar H. Vyas wrote:

  > On Tue, 25 Feb 2003, Keith O'Connell wrote:
  >>> Can someone please tell me if Webadmin is available in Debian
  >>> (woody).
  >> If you are referring to webmin, then yes (ver 0.94.7)
  > Oh Btw, don't use the packages in woody.  They are broken in various
  > ways and plus upstream says to use 1.070 which is in unstable.  but
  > it should install fine on woody.  (And if not send me bug reports.  I
  >  feel very bad about the woody situation but unfortunately it isn't
  > easy to update stable.)

Why should Jaldhar be waiting on anything to fix this shit?  And, I
mean, there's some obviously broken stuff in there.
If you are going to run a rinky-dink distro made by a couple of
volunteers, why not run a rinky-dink distro made by a lot of volunteers?
           -- Jaldhar H. Vyas on debian-devel

Hey, why not run a timely and high-quality distro made by fewer
volunteers, plus timely and high-quality packages made by a few
volunteers?  :-)

Reply to: