[Biopython-dev] Moving from CVS to git

[Bartek Wilczynski]

Hi guys,
On Tue, Aug 18, 2009 at 6:39 PM, Peter<biopython at maubp.freeserve.co.uk> wrote:
> On Tue, Aug 18, 2009 at 5:09 PM, Hilmar Lapp<hlapp at gmx.net> wrote:
>>
>> On Aug 17, 2009, at 11:52 AM, Peter wrote:
>>
>>> My impression from talking to OBF guys is if we really want to we
>>> can do this, but it require us (Biopython) to take care of installing
>>> and running git on an OBF machine.
>>
>> That's how I would put it too. Moreover, if you as people who want this and
>> know more about it already than anyone else among root-l can't be bothered
>> to take the initiative to spearhead this on OBF servers, the argument that
>> OBF "sysadmins" (which in essence is all of us who know how to do this)
>> should do the work is a lot less strong than it might have to be. I.e., if
>> you don't feel this would be time well invested for you, it is probably even
>> less well invested for other OBFers.
>
> Sure. Right now I don't think anyone at Biopython knows exactly
> what would be involved in running a gitserver, and it would take
> some investment of time to get to that point.
>

I think there is some grave misunderstanding here.

There is nothing magical or difficult in installing git on OBF
servers. It's  just a package.
There is no effort to be spearheaded by anyone. The command "yum install git"
needs to be run by someone with root privileges. That's it. It's
absolutely enough
to allow people with obf developer accounts to use git for development.

As for running a git-protocol-server, this is a bit more complicated
and can be done in many more
ways than with CVS. I don't think that anyone is expecting OBF to
provide git repository
hosting in a standardized way (currently only BioRuby uses git and
they seem to be fine
with github, similar for biopython)

The importance of having git installed on OBF machines comes from the
fact that it can
be useful for many things even if we don't host the repository on OBF servers.

Most importantly, for doing regular backups of git branch from github
to OBF servers we
need a machine with git installed. Currently it's my work machine, but
I think it would be a
 much better setup if we could do it directly from an OBF machine.

> In the long term I think running git on an OBF machine would be a
> good idea, but I don't personally want to spend time learning how to
> do that right now. By using github, we don't have to invest a lot of
> upfront effort in configuring a git server right away.
>
> I think it makes sense to just move Biopython to github in the short term,
> in the medium term we can (expertise permitting) get a git mirror running
> on an OBF machine, and then other tools like the git equivalent of
> ViewCVS (and if need be then abandon github - we won't be locked
> into anything permanent).
>

I don't quite understand what do you mean by "running git". Once we
have git installed,
you can use push and pull over ssh to a branch sitting on OBF machine.
We can also
make the mirror available for people (read-only) through http (just
place the repo in a
directory published with apache, no extra software required), But I
don't think this makes
much sense if we actually want to use collaborative features of
github. In my opinion this
would only bring confusion: either we make the github branch official or not.

The most difficult part is the "viewCVS" replacement. There is the
gitweb.cgi script, which
is (in my opinion) inferior to github interface. Installing it
wouldn't be difficult (it's CGI) so
we could do it, but is it better than github here? I'm not sure.. (you
can see how it would look
on a slightly out-of-date biopython branch on my machine:
http://83.243.39.60/cgi-bin/gitweb.cgi?p=biopython.git;a=summary )

To summarize, I think that the only thing we really need from OBF is
to have git installed
(Hilmar, can you help with this? I tried to even compile it on
dev.open-bio.org but there it depends
on multiple libraries and I gave up...)

best regards
Bartek