What next?

[Toshio Kuratomi]

On Sun, 2005-06-05 at 14:12 -0400, seth vidal wrote:
> >   about the technical work which need to be done and which need to change
> > the Fedora cycle durations. What, who, when, how long, and why this
> > can't be done in parrallel ? All I'm getting is "we need longuer cycles"
> > and still not understanding.
> > 
> > BTW I'm not represnting Red Hat position, just an engineer trying to
> >     understand some engineering plumbing and not getting it.
> > 
> 
> Who: fedora community people along with some red hat release engineers
> What: Fedora extras steering committee
> Why: We have to re-engineer EVERYTHING that y'all already have in order
> to: 1. support systems which are not inside red hat's network
>     2. have a buildsystem at all
>     3. have an errata release mechanism at all
>     4. have a db to coordinate any of these things.
>  We're not allowed to use or even see anything already written at rh b/c
> either: 1. it sucks or 2. it's part of the 'secret sauce' that some vp
> at rh thinks is valuable, somehow.
> 
> So we have to start over from scratch on everything and only a handful
> of those systems are in place:
>   1. cvs - only took a year to get it happening and another 6 months to
> use it
>   2. account system - only 18 months
>   3. build system - once extras got moving it took a mere 4 months to
> get time to take care of it - and we still have to throw one away.
>   4. errata system - not there yet - probably AT LEAST another 6 months
>   5. db for coordination - just starting
> 
> 
> oh and we need all of those things YESTERDAY.
> 
> So there you go. - that's the list - now do you understand why we would
> like a little more time before the next release?

We don't need all of those things yesterday.  If we did, we wouldn't
have FC/FE4 rolling out soon.  OTOH, I think you, Jesse, and Miloslav do
a good job of stating the primary goal here: infrastructure has been
taking too long to appear.  We need more developer hours to make
everything come together for FC5.

Will changing the amount of time in a release cycle get us to that goal?
I kinda doubt it.  I think a longer release cycle will mean developers
will get to spend more time working on their packages and projects.  You
might know that nine months will be enough for you to get the build
system out even if no one else helps but what about the rest of the
promised infrastructure gals?  What we need is for more developers to
spend time on the projects related to infrastructure.

We need to prioritize and assign tasks.

First: What things have been promised or are otherwise blocker
infrastructure goals for FC5 that we must have adequate developers
working on?  What things should they start working on next so we can
meet our goals for FC6?  How many man hours will it take to complete
each of the stated infrastructure goals?  Are the goals something that
multiple developers can work on or only certain specific developers?
What things are always going to need some maintenance work so we'll have
to permanently allocate some developer time to it?
 
And then: What developers from within the Red Hat fence can be assigned
to work on this instead of the projects they are working on now?  What
is needed to empower developers outside the fenceline to step into the
roles necessary to make meaningful contributions to the infrastructure?
Seth has bandwidth, hardware, information on the things the buildsystem
must do to be acceptable to Red Hat, and a job that gives him time for a
lot of Fedora hacking.  What would be needed to (for instance) hack a
new front-end for bugzilla that allows tracking of new packages in
Extras from initial request, through review and fixing, into acceptance?

I don't think we'll have all the infrastructure done simply by giving
you nine months to work on it.  I do think we can make progress by
saying Seth is going to have this specific piece of the infrastructure
that we have promised for FC5 done in nine months.  While he's working
on that we can have the other tasks promised for FC5 completed if we
assign Developer A, B, and C to work on them.  All other goals are
allowed to slip into the FC6 (9 + 6 month time frame) although
developers A and C can work on them if they finish their projects
early. 

If people want to commit Fedora to 6 months for FC5, there will also
have to be a commitment of more developers to divide up the necessary
work to make that happen.

-Toshio
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : http://lists.fedoraproject.org/pipermail/devel/attachments/20050605/62751178/attachment-0002.bin