"Tick-tock" release cadence?

[Peter Robinson]

On Thu, Dec 4, 2014 at 6:42 PM, Matthew Miller <mattdm at fedoraproject.org> wrote:
> On Thu, Dec 04, 2014 at 11:02:28AM -0600, Bruno Wolff III wrote:
>> >For us, that would mean alternating between concentrating on release
>> >features and on release engineering and QA process and tooling. During
>> >the "tick", we'd focus on new features and minimize unrelated rel-eng
>> >change. During the "tock", we'd focus on the tools, and minimize change
>> >that might affect that.
>> Presumably we wouldn't need to do this even up. We want say 2 to 1
>> or 3 to 1.
>
> A waltz beat, say. :)
>
>
>
>> >* prevent compounded delays caused by intersection of feature needs
>> >  and releng changes
>> There was a bit of that this time. But this was a really big change.
>> Are you thinking we will have this scale of change for releng on a
>> regular basis?
>
>
> So, frankly — and I think the rel eng team won't be offended here,
> because they know it more than anyone! — we're beyond what the current
> releng overall design can really scale to, and it needs an order of
> magnitude _more_ work in order to allow us to keep growing. (And that's
> not just with the Fedora.next stuff or new things like Atomic — the
> sheer _size_ means composes are going to take more than 24 hours in the
> forseeable future.)

No offence taken but it does depend on a number of things, some out of
our control. We're working to be able to parallelise a bunch of the
process but part of that isn't fixable with code alone. Parts are
being improved with more human resources for time (like me) and some
is due to things like IO on infrastructure.

Peter