Supporting distro re-packaging (Was: [sclug] Linux Apprentice Wanted !)

[Roland Turner SCLUG]

On Tue, 2006-11-07 at 07:22 +0100, pieter claassen wrote:

> BTW. a little off-topic but has anybody read this regarding the problems with 
> supporting custom apps in distros?
> 
> http://www.regdeveloper.co.uk/2006/11/04/apache_packages_support_vacuum/

Hmm. I've not heard of this guy before, but he appears to have a bit of
an attitude problem; the distros don't make the changes that they make
arbitrarily, they make them to satisfy norms and interfaces that apply
across thousands of packages within the distro. In the particular case
of Apache (and other web servers) on Debian, the evolution to the
current situation was slow and deliberate. The essential problem that it
solves is to be able, reliably, to turn modules on and off in installer
scripts, rather than by hand; the other changes are a straightforward
consequence of those changes.

The solution to his problem is either of:

- learn to say "no" and mean it (and deal with the fact that people will
get cross about that)

or

- provide a Debian-ised build and instructions for sources.list and
apt.conf to cause Apache's build to be used in preference to Debian's
(and still learn to say "no", but only to those who are seeking support
from apache.org for debian.org's build)

To demand that the distros undermine their own raison d'etre to save him
having to learn to say "no" is more than a little unreasonable.

> How interesting and a real problem. The solution is obviously to extend and 
> implement the LSB (http://www.freestandards.org/en/Specifications) rigorously 
> that would remove all differences between the distros and therefore turn 
> everybody into Debian/RH which would lead to another Microsoft ;-)
> 
> It seems to me that the LSB is inadequately defined and the scope too narrow 
> and it  should be extended to cover the minimal requirements for groups like 
> apache to provide flexibility and supportability?

Umm, is that a question?

LSB's problems are numerous, but it is only succeeding if its stuff is
getting implemented in most distros, which limits its pace. (If it gets
ahead of the distros then, rather than providing a unifying force, all
that it will have done is to add yet-another-platform.) To get to the
level of detail required to make the Debian-specific packaging for
web-servers uniform across most/all distros is going to take a _very_
long time.

Accelerating this is possible in principle, but infeasibly expensive and
may do more harm than good. (You'd need to hire hundreds of developers
and, unless you were willing to take and train people who aren't
currently active open source developers, you'd stall much other
development. You'd also have the problem of interfering with the
motivations of some existing developers who aren't paid to do what they
do, although this is difficult to quantify.)

- Raz