we could open the svn repository for selected developers. Technically there is no problem with this, however we would need every developer to sign an agreement that everyting they commit to the svn repository would need to be copyright eZ systems as.
This is due to the dual licesing that we use, we need to have copyright of all the code.
If some developers would agree to that we could open the svn repository for write access.
Write access could also be for contributions/projects, which wouldn't need to be under the dual license.
How feasible is this given the past problems of svn? Even with the the new mirror Paul needs to perform sanity checks on his rsync'd copy before putting it onto pubsvn because of its instability.
Should we instead try to work with places like sourceforge for things like this?
Our current SVN server is not scaled for high traffic. I think an svn repository for extensions would be a good idèa. This could be a powertools package where we could have all contribution development.
However we would need to have a new server for this. If someone from the community would like to maintain such a server we could probably provide the hardware.
Sourceforge or similar could be good idea. On the other hand, pubsvn.ez.no has all on board to host one or more extra source trees (if you would like to collaborate on GPL'ed-only contributions/extensions with svn). It would not be the place for bugfixes I think. I could add an ez publish based collaboration portal too over time (basic project management/forums/bug system). This all should be first approved by the ez crew of course. Any comments?
-paul
PS: The sanity checks are necessary because of transient states of the underlying berkeley database: even a read while rsyncing makes it "corrupt" for svn. It happened also with Google bots on pubsvn.ez.no, which forced me to let apache go down briefly during rsyncing the sanitized copy.
We were planning to set up a similar server for our projects anyway. So unless there aren't any other takers ...
From what I can judge on current community development, that machine would not need to be a supercomputer cluster for the time being. Or do I underestimate the load/problems for a read-write svn server? The load on pubsvn.ez.no is very low, as well as the bandwidth consumed to the internet, so the same machine could be used?