[Mageia-webteam] Webteam peers, bootstrapping

[Romain d'Alverny]

On Thu, Jan 6, 2011 at 18:23, Maât <maat-ml at vilarem.net> wrote:
>>> for example : forums we'll be deploying forums versions with git
>> Ok, but that does not necessarily mean access to the server (this can
>> be triggered indirectly).
> Yup but this means coding yet an other specific wrapper ?

Hmm yes. That's maybe too soon. But that would allow to integrate with
a test suite on the test server (+ reporting + blocking production
push in case of tests failures).

>>>>  * server deployment:
>>>>    - staging from a branch available to all peers
>> That is, ability to sync trunk (or whatever branch) to staging server.
>> That's an option that allows everyone to see up-to-date working apps
>> (and to quickly acknolewge broken ones). Or we can have a cron from
>> VCS to staging.
>
> This imply limiting ourselves to one trunk and limiting our capacity to thread features developement nope ?

Somehow, yes. :-/

> I'd rather see a model with many trunks/tags/branches available for (one for each active dev) and ~user dirs allowing people to switch their personnal instance to the trunk/tag/branch they want to play with...
>
> for a given project we would have a SVN repos organization like that :
> [...]

Well, that makes a lot of branches to create for each project and for
each new peer in a team.

That's where I like git and even more tools such as gitorious/github
(everyone gets his own branches on his own public space and can push
their changes back). It automates some tedious parts.

> Then for apache testing organization we could have :
> http://thesdeverver.mageia.org/manager => where $nickname can log on and click to switch it's own instance to what he wants (application,repository,tag or branch or trunk)
>
> http://the-dev-server.mageia.org/testing/~$nickname/ => where he tests the application he just switched to
>
> what do you think ?

Hmmm why not. Will need some work for the testing deployment
(especially for db-based apps).

As for the workflow, if it's not too inconsistent/repetitive (?):
 * a dev (team peer or not) works on her platform, tests there; keeps
a branch on mageia.org;
 * pushes changes for code review and merge into trunk (by any
available peer) from her branch;
 * tests/staging are done with the trunk - if it breaks, it depends:
fix it easily/quickly or reverse the merge;
 * the trunk is merged/branched in a production branch before being
moved to production server;
 * production fixes are merged back into the trunk;
 * and so on.

This in the perspective of encouraging small features increments to be
pushed to production as soon as possible (instead of waiting of having
all pieces fall together). That's not a general rule of course but...
a preference.


Romain