On Tue, Dec 21, 2004 at 11:04:05AM -0500, Colin Walters wrote: > A PostgreSQL database would be my suggestion. That way you can also > easily store in all the other metadata that you might want associated > with a version, such as: > > o which architectures it's built for > o what machine it was built on > o which test suites it passed > o which OS versions it was released in > o who maintains it Thing is, I truely dislike databases for such simple tasks. This are hard to setup/maintain beasts, that are not really needed for this task. Why not simply store this info in some files in the repository? Having a *simply* system to setup and work with is essential. Someone could simply get started by just checking out stuff, and go. With a database system required (but for most system not needed) it becomes a pain, and as a result user base gets smaller. That aside, I also don't like it at a conceptual level: we are splitting data that belongs together. I can't think of many examples where that was a good idea. You would have state now in two independent systems that you'd need to keep in synch, you'd need some sort of versioning in the DB, and before you know it you have to roll up your own little RCS in the DB. Enter subversion, see how well the DB idea worked for them :) So unless we'd have *big* problems without a RDBMS, I strongly suggest we stick to a simple to understand text file setup, keep everything together, and use the RCS to version stuff. -- Dimi.
