Normally I wait at least 3 weeks to upgrade, but there was a bug in 2.5 that was affecting a client critically (pagination on the Edit page wasn’t working), and I was told by EllisLab support that to solve it, I needed to upgrade to 2.5.1 (i.e. there was no simple code fix for this problem).
I think that patching, as opposed to upgrading, leads to more difficult support issues for yourself, and also to EllisLab, because it fractures the possible versions you could be running. Anything that makes EE more difficult for EL to support, I think we should avoid 😊
The idea, though, of having security-only upgrade releases—no other bug fixes—that makes some sense.
2.5—new release.
2.5.1—any security issues
2.5.2—2.5.1 plus any bug fixes and new features like IPv6. If there’s no 2.5.1, just go straight to 2.5.2.
2.5.3—any security issue
2.5.4—2.5.3 plus any security issues
Apple and Microsoft do this by having monthly security updates, separate from the regular OS updates. To be fair, EL rarely has a security issue in EE—I’m thankful for that often.
Or, should there be beta releases? I’d be much less upset if a beta release (2.5.1, say) was flawed but 2.5.2 was more solid and I knew I could upgrade to it right after it was released.
However, the original post raises the critical issue: why do upgrades seems to run into issues so often, and why does the testing of new releases seem to be so flawed? What can be done that hasn’t been promised and tried before?
@Sean C. Smith: If you put add-ons and themes in the same “third-party” directory, you’d lose the ability to put the PHP code outside the document root—and it wouldn’t be a good idea to make the add-ons folder be in a place that someone viewing source for your themes directory path would be able to guess. Though it would be more convenient, yep.
TTFN
Travis