We've had this discussion before, and the reasons we chose SVN are: (a) SourceForge supports it, so we can keep the repository on our SF page, (b) it was a system we all knew how to use, and (c) it's a system that most people in the OSS community know how to use as well. GIT is rather esoteric, and not many people know how to use it. I would personally be very lost if we switched to that system and don't really have the time to learn something new at this point. I use CVS actively on my other projects and have used SCCS in the past for collaborative development. SVN is miles better than either of those. Every system has its problems, but it doesn't make much sense to spend the time migrating unless the problems occur very frequently. Realistically, how often is anyone going to develop a version of Virtual Network Computing without an Internet connection? I find the SVN response to be perfectly adequate. It's a lot faster than CVS.
As far as regression hunting, that's what tags are for. I easily do regression hunting in CVS by creating a tag each time I release something to the general public. Then I can simply compare between the tags. This is even easier with SVN, which has atomic commits. Adam Tkac wrote: > Hi all, > > I would like to propose change of our default version control system - > from SVN to Git. Let me try to explain main advantages of such change. > > 1. Easier development of new major features > > Development of major features is quite ugly in SVN. Even if we create > a branch and then merge it, it usually consists of huge number of > commits in random order, many of them revert or modify earlier > patches. It obviously makes further review of patches harder. > > In Git it is pretty easy to do development in local branch and then > send patches when they are really ready. Patches could be very easily > restructured so development history looks nicer and patch review is > very straight-forward because changes can be logically separated and > each change is already done and you don't have to look if some next > commit doesn't modify same part of code. > > 2. "Regression hunting" is far more easier > > Sometimes we get report that "something worked in previous version but > now doesn't". When I would like to check which commit introduced > regression I have to use git-svn utility and then use git-bisect. > Without Git it's quite hard to find which commit caused problems. > > 3. No Internet, no development > > Every operation, like print of development history or print diff > between specific revisions need Internet connectivity. Without > access to central SVN repository development is quite hard. > Additionaly I hate when I have to wait for every operation with > repository. It is simply annoying. > > 4. Authorship of patches > > There is no possibility to keep authorship of patch in SVN. If someone > send us a patch then commit author is someone of us. We can only > mention in commit message that patch author is person X. With Git you > can easily preserve original authorship. > > I believe there are far more arguments for Git which I didn't mention > above - in my opinion it is far more better version control system for > projects than SVN. > > Comments are welcomed :) > > Regards, Adam > ------------------------------------------------------------------------------ _______________________________________________ Tigervnc-devel mailing list Tigervnc-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/tigervnc-devel
