Re: Maintenance of the Hurd parts in glibc

[Pierre THIERRY]

Scribit Barry deFreese dies 24/07/2007 hora 14:00:
> Didn't we try this already with the AMS branch or whatever it was?
> How is it much different to have patches in an experimental branch if
> no-one reviews and commits them upstream then it is to have them in
> Debian?

Sorry, but a branch only provides an infinitesimal part of a DVCS
benefits. IIRC, like in many other projects this size, some active
contributors had to wait quite some time to obtain write access to the
CVS.

With a DVCS, you don't have to provide patches, and people won't have to
deal with applying them on the source. Any contributor can provide a
first-class repository with its own modifications, preserving metadata
like commit logs and modifications dependency, because they are a tree
of linked changesets.

And changesets from any set of those repositories can be merged
together. If they don't conflict, it can be done with a negligible
effort, and repeated merges will continue to as easy.

And of course, you also get all the performance enhancements: all
operations are local and quick (no network used) and you can work
offline (no network needed).

That said, I have a very strong bias towards DVCS, because the one I use
daily, Mercurial, made my every day work a lot more easier, and not only
for my own projects. A repository is created in a directory and
populated with the directory's content in a matter of seconds, which
makes it possible to trivially track even small modifications to any
kind of content in a directory. You even get a quilt-like tool
integrated with the VCS (the idea was even ported to git).

Quickly,
Pierre
-- 
nowhere.man@levallois.eu.org
OpenPGP 0xD9D50D8A

signature.asc
Description: Digital signature