Hi Derek, On Fri, Jan 24, 2003 at 12:24:07PM +0100, Derek FEICHTINGER wrote: > > For our non-central infrastructure at CERN we would like to make the location > of the CVS repository configurable on a per group basis (as of now it is set > via the group type). > > We do not want to add new columns to the data base without really needing to > (and we would only want to do it in a way that the Savannah developers agree > - it would hurt the project if everyone starts up his own code base). > > Now my first question: > Would it be worth to make an effort together to add fields to the group > table?
Sure. I think we should do that together. I was planning to due some similar changes in the next few weeks, because we are going to let projects decide whether they want to use CVS, arch or aegis. > The basic philosophy would then be: > The 'group type' table provides the default values, but the 'group' table > allows project admins to override these. Yes. In the original SF code those values used to be read from the group table. Mathieu did some changes to avoid the overhead of reading that table, by guessing the default value from the project type. I'm not convinced that there is a significant overhead (the group type table has to be accessed anyway) but even if there was, it is worth the gain project configuration flexibility. If we start to allow different Revision Control systems, we need that flexibility. > I know that these changes will probably not improve Savannah much for use at > savannah.gnu.org, but they will make it more useful for other site layouts. > What do you think? I agree. And I also see advantages for Savannah itself. > A second, more specific question: > In the 'group' table there is the cvs_box entry which now always gets set to > "subversions". This existing field could be used for the CVSROOT locations, > if it is not reserved for another purpose (was not able to find anything in > the gnuscripts). Are there any plans to use this entry? If I remember correctly it was used in the original code before Mathieu introduced the group type table but now that the information on the CVS box is inferred from the type of project, it became redundant. I'm pretty much in favor of your proposal. I think that the cvs_box field should contain the actual URL; sure, we have a lot of redundant information, but it will allow more flexibility. > This would still leave us short of one field that would be needed to store > the > URL for the CVS web viewer (viewcvs, cvsweb - however that project would > like > to manage their CVS web access). We'll have to see if there is some unused field that we can use. Apparently the actual structure of our tables in Savannah do not match the one that would be created in a newly installed copy, so we will have to go through tables structure and it could be a good time to settle on a new structure that we all agree with. Regards, Jaime
