motu-release

[Scott Kitterman]

On Saturday 28 June 2008 07:37, norsetto at alice.it wrote:
> On Saturday 28 June 2008 00:08:20 Stefan Potyra wrote:
> > Subquestions are: what do you (members of motu-release) would like to
> > add, what did you observer generally in regards to motu-release, what to
> > improve?
>
> One thing that irked me, and I would really do something about for
> Intrepid, was the fact that two positive votes are enough to approve any
> FFe, no matter if 3 out of 5 members are against it.
> I would definitively change this to a majority vote. The problem is that
> with the current number of members this would require 3/5 to pass which
> might not be attainable in a reasonable amount of time.
> Reducing the number of members to 3 (and therefore having 2/3 to pass)
> doesn't seem a good idea too.
> So, I'd propose a +2 in a (insert a reasonable amount of time here, 2 days
> since the date a _valid_ request was filed seems reasonable to me) ? The
> obvious drawback is that no FFe can be approved before the 2 days elapse,
> in my view a reasonable price to pay.
>
> An alternative would be to have a veto system, in which any member can stop
> the regular process by simply objecting (obviously with reasonable
> arguments) against the FFe. In this case the FFe will not be approved until
> the required majority is obtained.

I generally operated this way myself.  My view is that if somone on 
motu-release has expressed objections, the FFe should not be approved until 
those are resolved.    I think we did this in general (see the envy-ng FFe 
for an example).

I think +2 with no objection is enough.  I think that informally we mostly 
operated this way and it worked well enough.  

> I'd also discourage the practice of accepting an FFe on the base of a short
> IRC chat without apparently any research on the implications and background
> of the request.
> We have an FFe process so lets make the best use of it (accepting an FFe
> because your buddy is asking you to do it on IRC, or because somebody you
> trust is telling you that it will be good to have that package, are not, in
> my humble view, good reasons to accept an FFe).

I think it's up to each member of motu-release to use their best judgement.  I 
don't recall seeing any "becaue my buddy asked me to do it" situations.  I do 
think it's reasonable to consider the source of recommendations and accept 
advice from experts.  Along those lines, I think delegating specific 
packages/package types to specialists (e.g asac and mozilla related things) 
worked well and should be done again.

> Finally, I always found a nonsense that we have a rather strict system
> until few days before release and then exactly when we should really
> tighten the tap, we relax all requirements (its enough to have one IRC
> approval without sometime even filing a request).
>

Personally I'm more concerned with getting motu-release review and approval.  
I think there should definitely be a bug to keep track of what happened.  I'm 
less worried about what order things happen in.

In past releases we've had trouble with not all FFe (UVFe) being processed in 
a timely manner.  It was much better this time, but we did have one.  The 
first Universe upload to hardy-updates was a missed bug in the gfortran 
transition.  I think that we need to keep the process light weight and 
flexible.  I don't think in this release any FFe filers felt unresponded to 
even if they didn't always get the answer they wanted.

Scott K