[Bioperl-l] Bioperl on GitHub

[Chris Fields]

On Apr 14, 2010, at 3:46 AM, Peter wrote:

> On Wed, Apr 14, 2010 at 7:16 AM, Jason Stajich <jason at bioperl.org> wrote:
>> I think we should migrate as well. The primary can also be at github if you
>> like - the lack of code.open-bio system is a real problem that doesn't seem
>> to be getting fixed....
>> We could also make the primary github if need be?
>> 
>> -jason
> 
> We (Biopython) and BioRuby certainly seem happy with github.
> 
> You've probably thought of the following issues, but just in case I've
> written some thoughts from our migration.
> 
> If you want to go down this route, I would strongly urge you delete the
> current github repositories and instead do a proper import from SVN
> to git with the FULL history. I'd do this ASAP to disrupt as few users as
> possible.

We'll have to announce this on the list first, in order to make sure there is a proper migration path.  We did this with the CVS->SNV migration a few years back without any real pain.  My guess is this wil be much easier, particularly if they have SVN support (though I would like to really test that out first).

> For BioPerl you have the four sub-repositories (bioperl-live etc) which
> adds a later of complexity we didn't have to worry about.

The current setup just looks at main trunk for those repos and pulls every 15 minutes.  

> I'd then suggest you continue with SVN on dev.open-bio.org as the
> primary in the short to medium term, but declare github.com/bioperl
> as the official mirror for anonymous read only access (via git or SVN).
> 
> You can then encourage the BioPerl devs to setup their own forks on
> github, and experiment there with branching and merging - but for now
> keep github.com/bioperl as read only, updated only via the SVN import
> from dev.open-bio.org
> 
> We had a couple of months like that (all commits via CVS, pushed to
> github), and I personally found it a very useful training period.
> 
> (Not wishing to dictate - these are just my suggestions for the outside
> having been though something similar)
> 
> Regards,
> 
> Peter

My thought is we'll test the complete migration to git locally on dev first (mirroring to github as ro, which can be forked/etc).  Not sure what we would need to do that; my local setup uses gitosis, but we could probably set up a /home/git-repositories/bioperl.git.  

This is another critical point:  the easiest path to migration is to move the entire svn repo over.  However, all bioperl is contained as one large repo.  Should we take this opportunity to try making them truly separate?  Or leave them as is?

Theoretically we shouldn't have to worry about which repo is the 'primary' as git is supposed to be truly non-distributed, but in reality most dists tend to make one location the dominant one for releases, so we can designate the github or dev the primary one that at that juncture.  

chris