Re: Release process thoughts

["James K. Lowden" <jklowden schemamania org>]

On Mon, 20 Oct 2003 06:07:55 -0500, Lars Clausen <lrclause@cs.uiuc.edu>
wrote:
> We have the "release early" thing down, but we
> also need to "release often".  However, we should certainly avoid the
> disasters of hasty releases in 0.86, 0.88.1, 0.90.  We should probably
> in the future make a CVS branch when we start a release cycle, so
> development can continue while a release is being tested. 

As the maintainer for a smaller project, I can say that branching releases
definitely *saved* us work.  Patches frequently come in for the released
version.  They might or might not directly apply to the Head, but they'll
surely apply to the release branch if they're any use.  

I don't know about you, but once I start a release cycle, I find there's a
compendium of tedium to deal with.  Errors in the documentation (like
references to the wrong relase, or to some now-missing feature or
long-fixed bug).  Writing the README, the announcement, the new feature
list and the how-to-upgrade procedure.  Come to think of it, from here it
sounds like I'm talking about Dia.  ;-)

Not to mention minor bugs.  We issued two point releases this year fixing
big-endian problems that didn't crop up in normal (read: i386) testing.  

For us, branching ended the "I don't want to put in your patch until after
the release" debate.  

I would think Dia would benefit more than FreeTDS did from using release
branches.  

--jkl