Greetings monks. I'm in the early planning stages for a database application that will be written in Perl. One requirement that I know will be present is version control. Users will need to ability to do exclusive checkouts and checkins, as well as examine past versions of objects and revert the object to a past state. The objects are complicated enough that they will require several tables to store. Users will be doing the normal list, edit, view, save stuff with them aside from the versioning actions.
When I first encountered the problem my mind jumped to CVS. I'll just hack up a database-to-CVS bridge and handle it that way, thought I. But as I think about it further, I'm not sure that's a good idea. There's a large impedance mismatch between records in a database and files in CVS.
An alternative might be to have a version column in every table. The problem with this is that the primary key for each column is now (foo_id, version) rather than just (foo_id). I fear this will complicate the rest of the application enormously. Join tables in particular would become much less useful. It's so much nicer if a single integer is enough to identify a record.
Any bright ideas floating around in the monestary? Feel free to meditate on this one for a while - the application is at least a couple months away from implementation.