I think the discussion is moving in a wrong direction here. We should maybe start to rethink what the intent of version numbering is to start with: · A newer version should be recognizable by comparing it to an older one and this must be doable by both computers and humans · Newer versions might introduce interface changes, but these are described in the package history -- the version number may not reflect this change (e.g. Python's API level will change even though it's only a new patch level that's being released) So basically all that's needed is to be able to compare the version numbers of packages within a release sequence. It doesn't matter whether they use numbers and dots, words or dates. Trying to automate compatibility checks by only using version numbers isn't really a feasable approach, IMHO. This will always remain a manual and errorprone process... -- Marc-Andre Lemburg Y2000: 386 days left --------------------------------------------------------------------- : Python Pages >>> http://starship.skyport.net/~lemburg/ : ---------------------------------------------------------