On Wed, Feb 19, 2014 at 8:16 AM, Barry Warsaw <barry@python.org> wrote:
On Feb 19, 2014, at 07:50 AM, Guido van Rossum wrote:

>That's why it's RC2 and not 3.4final, right? Once Larry says it's baked,
>everyone *will* have a chance to test it. What value is a preview of the
>preview really going to add? Give Larry some trust and freedom to do things
>in the way that makes him comfortable.

I totally agree that Larry should be given fairly wide discretion.  He's also
feeling out his first big release and deserves some slack.

Thanks for the support!

However, I think Matthias wants read access to the release repo because he's
*also* cherry picking patches into Ubuntu's archive.  We're already seeing
some problems, which we want to investigate, and Matthias has also performed
archive-wide test rebuilds to find Python 3.4 incompatibilities in 3rd party
libraries, most of which we'd like to fix (e.g. main packages, if its not
possible to get to everything in universe).

Again, this is what RC2 is for (and RC1, for that matter; apart from 20+ asyncio patches there really isn't much of a difference between RC1 and RC2). Larry may legitimately feel uncomfortable with what he's got on his local drive and prefer to tweak some things before telling people "go ahead test with this" -- the difference is that if he was working on new *code*, he could just not commit his work-in-progress, but since here he is assembling the final sequence of *revisions*, he prefers just not to push. (Georg alluded to the fact that you can undo changes in a public repo after they've been pushed, but I suspect he's referring to hg backout, which creates extra revisions, rather than a remote version of hg strip, which would go against the philosophy of DVCS. Either way, Larry's use of Hg is a totally legitimate workflow.)
 
Matthias just switched the default for Ubuntu 14.04 to Python 3.4 by default,
so this is a great test bed to find problems.

And that's great, of course. But what is really gained by giving Larry trouble over a few days' worth of delay, at most?

--
--Guido van Rossum (python.org/~guido)