[Incubator] OpenLayers graduation time

[Christopher Schmidt]

On Thu, Oct 18, 2007 at 10:49:30AM -0700, Jody Garnett wrote:
> Daniel Morissette wrote:
> >I don't know if other graduated projects are in the same situation, 
> >but if yes then I think it would be smart to have daily backups from 
> >those projects not using OSGeo infrastructure as well.

There are three aspects of backups that you expressed concern about:

 * SVN 
 * Trac
 * Website

I'll state, to begin with, that I would love to move all this to OSGeo's
infrastructure. The machine that is currently hosting OpenLayers is
ill-suited to the task, and if I could move maintainership of the
OpenLayers materials to OSGeo's infrastructure, I think it would be a
great benefit to the project. (This is my personal opinion: I don't
speak directly for the PSC, but I think that the PSC would agree on this
point).

Unfortunately, there is currently no way to migrate the OpenLayers SVN
to OSGeo's infrastructure without abandoning existing functionality.
Specifically, OpenLayers has a policy of maintaining an open 'sandbox'
where users can commit changes. 'trunk', however, can only be committed
to by core committers -- this is the equivilant of the committers list
in most other projects.

There is no way to pass through LDAP group information to SVN in a way
that it can interpret, so there is no way to maintain these two distinct
groups -- even if one were to be open to all users with an OSGeo ID --
on the OSGeo infrastructure.

There is a desire to maintain these distinct groups because it allows
active development from non-core committers, while at the same time
ensuring that committers do not have to review each commit to ensure
that nothing in trunk has accidentally been committed. (Given that we
have several times had a core committer accidentally commit in trunk
instead of a sandbox, even *with* the limited access, this  is a real
concern.)

So, although I would love to move to OSGeo's infrastructure -- it would
make management of the project much easier, it would make reporting bugs
much easier for users, etc. -- there are technical limitations to moving
trac/svn. 

My hope is that the core website *can* move to OSGeo infrastructure -- I
will have to, of course, confirm that that is the desire of the PSC.
However, this is relatively minor -- the trac aspect of the OpenLayers
website is where all the real content is.  However, there are some minor
technical challenges to that -- things like our regular svn checkouts to
build documentation and example directories may not be appropriate for
OSGeo machines. I will work with SAC and the OpenLayers community to
attempt to figure out what we can move, what we can ditch, and what we
can leave in place for now until we come up with a solution to the
SVN/Trac problem.

The website is in SVN. 

Backups of OpenLayers trac/svn are now publicly dumped nightly to
http://dev.openlayers.org/backup/ . They should appear around midnight
(depending on the state of the machine -- the nightly backup last night
took 723 minutes, for example, which is one of the reason I'd love to be
off of our infrastructure and onto OSGeo's.) Backups of the mailman
archives are now there as well. I will work with SAC to ensure that
these backups are mirrored to OSGeo's backups, so that if I go rogue and
blow up the server room after getting one too many reports of IE bugs,
there will be a record of the things that drove me there :) 

Daniel -- does this resolve some of your questions? 

Regards,
-- 
Christopher Schmidt
MetaCarta