On Thu, Feb 11, 2010 at 12:15:30PM +1000, Chris Curran wrote:
Having some mail client issues. Resending.
On 02/11/2010 11:17 AM, Susan Lauber wrote:
The time rapidly approaches when we need to start updating the guides
for a new version. In this example - start work on F13 version of the
guides.
Basically - I wanted to know where to start contributing changes
intended for the F13 version on guides. And where to point new
contributors as well.
In the meeting tonight, a discussion started about branching
philosophies.
#action laubersm Bring branching topic to docs list to decide how we
want to do this going forward, and record it for posterity.
I have looked at a couple of guides in git. They seem to have a master,
F11, F10, etc, and various F??-tx branches.
I think we generally agree that the F??-tx branches are for the
translation teams.
The rest seems to indicate a need for a F12 branch and then master
becomes F13.
But when should the F12 branch appear so that work on F13 can begin?
I did not look at all the guides - there may be other schemes in use
already.
So I see 2 major parts to the question:
1. Is this the branching scheme that works best for Fedora Docs group
and guides?
What are the alternatives?
pros/cons? etc
The alternative is not branching. Just to build and forget. This isn't the
kindest approach and goes against some of the theory of open source.
I would say branch at release. We do not have the resources to maintain
old forks of books but we should keep the source for posterity.
I think this is the most rational choice too, and mimics what we do
elsewhere in Fedora. Granted, the No Frozen Rawhide process we're
trying for F13/F14 changes that a bit, but the reasons why that's
changed aren't really relevant for docs if history is a guide.
2. WHEN do we branch?
I would argue for at the date of translation string freeze. Only critical
updates (like an error which causes users to delete their root partition)
found in translation should get updated after that point.
Most of the world over, a branch happens at the time you need to
indicate "This bunch of code/docs/stuff is not going to change
drastically from here on out, and will be treated as stable."
Translation seems as good a way to anchor that as any, and better than
some others. (Release day isn't a horrible choice, but it's kind of
arbitrary when you think about this in terms of "when is this stuff
considered stabilized.")
Should the F12 branch been made at release of F12 so that master moves
to the next version right away?
It should be made at translation. I would argue for having master to just
keep going. This allows for continual development. I know how much some
people *hate* RAD, agile or SCRUM approaches but unless I get three major
contributors that is the only feasible approach to take. Fedora
continually changes during a release. It is not the same Fedora by the end
of the 6 months when the next one starts coming out and the documentation
is usually out of date for most interfaces of developed projects.
I would argue here for a master/lastest branch and then legacy branches.
Sounds right to me. The nice thing about git projects is how cheap
the branches are. You can collaborate with people on a
"new-section-on-empathy" branch that is ugly and broken at various
times, and when it's all done, merge it cleanly back to "master" in
the middle of the cycle.
If so, how do we handle mid release updates to a guide?
We should have a master or latest branch and some method for pushing more
frequent updates. The present method, with the CVS/php/voodoo monstrosity
is a failure. It takes days to push a book and translations. This is not a
conductive environment for updating documentation. I know someone will
respond with the "Zikula is coming" or some such but this is really
holding us back. Is there anything stopping Infrastructure just giving us
limited access to puppet so we can push updates in a more timely and
effective fashion?
Until something changes in Infrastructure do not expect *any* mid release
updates, ever. It is too much work and takes a fanaticism and commitment
which I certainly lack. Masochism is not everyone's idea of a good time.
If I was part of the Docs team leadership and I really wanted mid
release updates, I'd expect to make more of an effort to make specific
assignments for those updates to new people who are showing up asking
to help. (It'd be good to have more active mentorship of those folks
on the list, period, but that's a side topic.) For instance, I might
assign someone to look at a specific guide, and confirm it is accurate
for the existing release -- pointing that person at the project's
Fedora Hosted wiki page with instructions on how to get the material,
build, and read/test it.
I'm not saying I think mid-release updates are intrinsically valuable,
just making a suggestion for people to consider iff. mid-release
updates are important to them.
Should master stay the current version until the next release alpha?
Which is about now.
Or even longer?
Master should just go with the latest and greatest. I think this is one of
Fedora's strengths which the Documentation Project should latch onto and
hold for dear life. We can channel the energy which most people, including
myself, love about Fedora and open source; the life which drives continual
change. People want to document new features and technologies far more
than they want to do many other things. We need to channel this energy
into our docs. We need people working on documenting the latest and
greatest bits of code. Master should be as fast as the writers want to
ride the lightning, if that is Rawhide, so be it. I am presently intending
to go this direction with the Virtualization Guide.
+1. I think this is the sanest approach with a six-month release
cycle for a number of reasons -- not just for reasons of people-time
when it comes to maintenance, but also when it comes to coordination
and tracking effort.
