On Tue, 2009-08-18 at 15:02 -0700, Matt Graham wrote:
> Ubuntu's problem AFAICT is that they're trying to be New! Shiny! and
> Awesome!.  This is a worthy goal, but it can lead to the system
> being as stable as a stegosaurus on rocket-powered roller skates.
> Debian seems to be much better at the whole upgrade thing because
> they're so conservative in moving forward in stable.

Which is basically the difference between a standard release and an LTS
release.  The LTS releases are conservative, while the more standard
releases look at user features as well as risk.  Sometimes the benefit
long term is worth a little risk short term.

At the end of the day, release management is a real pain.  If it's for
your own project where you're the only developer, it's a pain because
you have to stop developing to do it.  If there's more than one, you
have to start agreeing on when to freeze, etc.  When there's thousands
of developers on thousands of projects -- I'm glad that I'm not a
release manager for a distro.

Just to add some information to the discussion.  All Ubuntu releases are
tested as upgrades from the previous release and from the last LTS (e.g.
Hardy -> Karmic).  Of course that can't cover every package everywhere,
but some testing is done in that regard.  Also, to cover settings that
require some user interaction in Jaunty the "Computer Janitor" has been
added to try and clean up the stuff that "may be on purpose."

Personally, I only clean install when I get new hard drives.

		--Ted