Re: Odd release numbers for development versions?

Lamar Owen <lamar(dot)owen(at)wgcr(dot)org> writes:
> If the Linux kernel used CVS like a reasonable Free Software effort,
> then the current odd/even split wouldn't even be necessary. We use CVS
> -- if you want development trees to play with, you fetch the tree by
> anon CVS and update as often as you need to. There is absolutely no
> need for a Linux-style release system with CVS.

Actually, another way to look at it is that we do have two release
tracks. There is the bleeding edge (CVS sources, or the nightly
snapshot if you don't want to be bothered with setting up CVS), and
there is the prior stable release (RELm_n CVS branch, which we update
with critical patches and re-release as needed). Seems like the main
practical difference from the Linux release model is that we don't
bother to make formal labeled/numbered tarballs of the development track
until we are in beta-test cycle. You want development track at other
times, you just grab the latest sources.

So far, the alternating development and betatest/bugfix cycle has worked
really well for the needs of the Postgres project, so I don't think
anyone is eager to change that approach.

regards, tom lane