Re: Proposal new source archive format
On Mon, May 01, 2000 at 10:14:44AM -0700, Joey Hess wrote:
> Ben Collins wrote:
> > We should get away from automatically diff'ing the original source to create
> > a debian .diff.gz file.
> Um, I seriouly hope you're kidding.
> Have you ever worked with a packaging system (rpm) that forces you to
> make your own diffs? It's a serious prioductivity killer. I am about 50%
> less productive with rpm than with deb's, and it's not due to lack of
> experience. Building those diffs yourself every time is a serious pain in
> the neck.
> If you _want_ to make your own patches in some circumstances, that's
> fine. However, many developers will not. Especially people who simply
> check the upstream source into cvs, and debianize it, which is a
> completly different way of looking at upstream source, and which tends
> to move you away from thinking about things in terms of patches and
> diffs entirely.
By automatic I meant that assuming that the diff between the foo-1.0/
directory and the foo-1.0.orig/ directory is what every package wants is
bad. The configure.in case is one such problem along with other generated
files (such as .y). So for the packaging system to assume that fact is
wrong and part of the reason a lot of the larger package maintaines have
moved on to non-standard systems like glibc and gcc have. It is restrictive
to a lot of things.
This does not mean that there can't be an easy way to create diffs, nor
even an automated tool for doing so, but the source-package program should
not assume this fact. The diffing of source should be interactive, even if
it is automated. How many developers do you think actually stop and check
.diff.gz before uploading anyway?
 I do, and I'm sure you do, but our current system does not even allow
for the maintainer to interact with the diff's much less decide how it is
/ Ben Collins -- ...on that fantastic voyage... -- Debian GNU/Linux \
` firstname.lastname@example.org -- email@example.com -- firstname.lastname@example.org '