Re: Posted zsh 5.9

["Daniel Shahaf" <d.s@xxxxxxxxxxxxxxxxxx>]

Axel Beckert wrote on Sat, 14 May 2022 23:35 +00:00:
> On Sat, May 14, 2022 at 11:21:26PM +0000, Daniel Shahaf wrote:
>> Axel Beckert wrote on Sun, May 15, 2022 at 00:11:31 +0200:
>> > Hi,
>> > 
>> > On Sat, May 14, 2022 at 03:59:35PM -0500, dana wrote:
>> > > I've just finished posting zsh 5.9. It should be available in the usual
>> > > places:
>> > […]
>> > > creating-a-release instructs us to wait a day before posting to -announce,
>> > > which has always been followed a little loosely i think, at least by me...
>> > > but i'll do that this time, unless someone wants me to go ahead now.
>> > 
>> > I'll upload it to Debian unstable anyway, I as I assume there won't be
>> > any changes anymore until tomorrow. :-)
>> 
>> I'd rather you didn't.  Backwards compatibility promises don't kick in
>> until a release is announced, even if it has been tagged.
>
> IMHO having been uploaded is a bit more than just tagged.
>

Read "even if it has been tagged and uploaded", then.

>> If we pull the tarball for any reason, your users will get to keep
>> both pieces.
>
> If that happens, I'd expect that a 5.9.1 or 5.9a or something is
> released and 5.9 is never annouced. Like e.g. the ASF does with HTTPD
> releases. Not every 2.4.xy or 2.6.xy has been released (or at least
> not announced).
>

In such a case we might do the same and re-tag as 5.9.1, yes, but that's
not a reason to let users run 5.9 before it has been released (= announced).

>> Cf. https://dist.apache.org/repos/dist/dev/subversion/README.html.
>
> That talks about release candidates, not about already publicly
> downloadable tar balls with the correct version.

I'm aware of that.  I consider what that page describes to apply to
releases-to-be that have not been announced.

> Our -test releases are the release candidates from my point of view.
> And yes, of course I fully agree with that. That's why I uploaded all
> packages of release candidates to Debian Experimental, not to
> Unstable.

+1

> According to discusion in here, the reasons for the delay are also
> others. (I also remember from earlier discussions about the release
> process that SF's upload thingy is a bit fiddly and sometimes needs a
> few attempts to get their "newest upload" pointer right.)

Good night,

Daniel