[Asterisk-Dev] Linux leaves Bitkeeper: quite a dustup

[Steven]

On Tue, 2005-04-12 at 20:28 -0700, Kevin P. Fleming wrote:
> Steven Critchfield wrote:
> 
> > For my office we have used subversion. We have been pretty happy with it
> > as well. But the reason I suggested we look at arch is that it would be
> > really nice to help merge in some of the foreign patch sets into an
> > easier to get and maintain branch. 
> 
> Exactly... one of my goals (I'll post more details in the other thread) 
> is to allow members of the 'team' to have a public area on a development 
> server to make their patches available for testing/comments/etc, in a 
> form that people can easily download and automatically update (without 
> requiring monitoring a Mantis bug and downloading/applying new patches). 
> While everyone can do this now by setting up their own CVS/SVN/etc. 
> public server, that's a lot to ask (and manage) for someone who just 
> wants to help out with development and bug fixing.

So that is another reason I think we may need arch. Instead of bogging
Digium down further with the administration of the various repositories,
let those of us who are willing to contribute back host our own
repositories. One of the big attractive features is the ability to pick
and choose from multiple repositories. Arch is supposed to break down
the commits into smaller section than we are used to with CVS or SVN so
you can even pick apart a commit and get only what you want from other
people's repositories. 

I look forward to a time where we can have someone tracking Digum's
HEAD, Steve Underwood's spandsp tree already merged with HEAD, and maybe
someone needs Kapejod(sp?)'s ZapBRI code again already merged.

In that example, the developers already are hosting their own
distribution sites. They could create their own extensions to the Digium
HEAD repository. From that, we as the users or other developers could
pick and choose what we need to get what we want done.

This type of management will splinter a few small portions out, but will
keep those splinters close to the core group. I would hope it would
allow people easier ways to explore new ideas and mix and merge all the
best of breed functions.

Hopefully we could remove some of the work required to organize the
development from Digium as sole gatekeeper to Digium as primary
gatekeeper. 

A concern I have about the distributed development model is in getting
the code back to Digium. In the current centralized development model, I
submit patches to Digium for inclusion. The submission along with the
disclaimer on file show an active interest in Digium using the code. In
a distributed model, I wouldn't need to be actively submitting code back
for it to be easily available. I don't know what new mechanism will need
to be adopted to keep all that clear.