[1834] in athena10
Re: Separate packaging [Re: Athena delete]
daemon@ATHENA.MIT.EDU (Greg Hudson)
Tue Mar 24 12:01:48 2009
From: Greg Hudson <ghudson@MIT.EDU>
To: Anders Kaseorg <andersk@mit.edu>
Cc: debathena@mit.edu
In-Reply-To: <alpine.DEB.2.00.0903221400470.4457@vinegar-pot.mit.edu>
Content-Type: text/plain
Date: Tue, 24 Mar 2009 12:00:52 -0400
Message-Id: <1237910452.6246.252.camel@ray>
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
On Mon, 2009-03-23 at 15:00 -0400, Anders Kaseorg wrote:
> When multiple packaging systems are involved, or worse, multiple packaging
> systems maintained by different people, there is the additional problem
> that the overhead of keeping them all in sync becomes a severe impediment
> to fast development.
For the debian directories, we have to evaluate these concerns in the
context of our own work. Maintenance of the sources in athena/ is, in
pretty much every case, tied very closely to the release of a site-local
Debian package. Almost every commit is intended to correspond to a new
package release. The maintainers of the sources and packages are always
the same. The process is "edit sources, add changelog entry, commit
with no SVN log message, deploy to proposed". It's difficult to imagine
a dual-repository approach making the process any more streamlined--in
fact, no matter how good the version control tool is, there would almost
certainly be one or two extra steps involved in every change.
In the original context of jik's request to add RPM packaging materials,
however: I think we should probably say no, on the grounds that it's
dogfood we have no interest in eating, and thus it would just bitrot in
our repository.
(On a general level, Subversion maintains packaging materials for a
bunch operating systems in its packaging/ directory and that hasn't been
nearly as much of a problem as, say, maintaining in-tree scripting
language bindings. So I think the Debian concerns are a bit overblown,
but I recognize there are valid arguments on both sides.)
