Releasing Alphas and Betas without "freezing"

Scott Moser smoser at
Mon Jun 18 14:30:29 UTC 2012

On Mon, 18 Jun 2012, Thierry Carrez wrote:

> > With our dailies, I've found that the milestones are most useful for
> > planning bug fix landings and feature deliverables. I'd be +1 on
> > dropping alphas all together. [...]
> I'd certainly agree that the main value of milestones is as target dates
> and to create an internal cadence within a cycle. That said, I still
> think there is value in singling out a given build once in a while to
> serve (1) as a reference point ("this bug wasn't present at alpha2"),

I will +1 that.
I do not think it justifies alpha in itself, but it does justify keeping
*some* historic builds around for a longer period of time.

In cloud-images, anything we mark as milestone (which includes alphas) are
kept indefinitely.  The cost of a build is currently storage of ~2.1G.

The value is that people can easily reproduce that "this bug wasn't
present in alpha2", as compared to "I don't think this bug was present in
alpha2" (which is almost entirely useless).

I realize, that, because we do not keep historical content in the archive,
"alpha2" really only means "packages included in the image (or ISO) in
alpha2".  That said, I do think it is useful to keep around some
generally-functional builds to reference and bisect from.  (And I surely
wouldn't complain if someone said we should also then implement something

More information about the ubuntu-devel mailing list