[Mageia-sysadm] maint db, rails and hosting

[Michael Scherer]

Le mercredi 05 janvier 2011 à 17:38 +0100, Romain d'Alverny a écrit :
> Hi there,

> What's your take on this?
>  - hosting Rails apps; how/where to host them (remember we have Gandi vhosts);

if rails support fastcgi, we can host it on alamut. But the obvious
problem is that we have perl ( catdap, mga mirror, bugzilla ) and ruby
and python. 

And please, try to use postgresql rather than mysql. We do have a
postgresql expert on board ( Olivier ) and postgresql is much more
integrated with puppet at the moment.

And for rails application, people have asked to consider redmine
( stormi, xrogaan on irc ), gitorious ( several people ) and teambox
( well, you, and shikamaru is packaging it ). So rails is likely to hard
to avoid.

>  - how we can deal with open contributions, specifying base
> requirements for the tools we expect to receive (being square, but
> flexible enough to accept more contributions); and what these
> requirements would be;

IMHO, you should stick to a minimal number of frameworks. Easier to say
that to do.

Having several framework lead to several template engines, thus making
web team life more complex ( ie django + tt + ruby one + php ) by
requiring duplication of templates code everywhere, even if this issues
can be minimized by using css and so on.

It lead to people who will debug to learn all languages ( I can code in
everything proposed,  but not everybody can ), causing problem. After
spending the nigh profiling catdap. 

More ever, taking framework without ensuring the team have the needed
competence is a risk. So a minimal number of people should know before
deploying a technology.

So a list requirement of frameworks would be :
- work with postgresql ( everything work with mysql, but we went on
postgresql when possible :
https://www.mageia.org/pipermail/mageia-sysadm/2010-October/000099.html )

- work with a template engine that is supported. TT is required for
sympa && bugzilla && catalyst application, django for transifex. So
having something similar in syntax would likely help ( but I guess the
closest we can have is django template rather than TT ).
  
- see how it can be handled translation wise. I know that people
prefer .po in general, and that's supported by tx and most tools. So
avoid custom format if possible.

- obviously must have a proper stable branch/security policy , etc, same
old criteria already laid out.


>  - how it's acceptable that people use infrastructure outside of
> mageia.org to bootstrap their contributions (like, still pointing that
> a master branch should be reasonably migrated to mageia.org infra, or
> other guideline);

It is acceptable, this is just slightly disheartening after working days
and night to setup something and see that people do not even ask.

But I think something outside should not be seen as official, as we
would otherwise face governance issues ( with practical impact ), and
that we should give incentive to people to use our infrastructure.
One of the goal is also to avoid the mess of having several user
databases, as it was identified as a problem in the past. 

Problem that could occurs are "the member of the team who manage the
project is hit by a bus and no one can request the main password", or
"someone decide to fork and close the project" ( ask to fusion inventory
people how they forked the launchpad project ). 

Or the various issue of "provider change his mind and decide to make
free user pay" ( like bitbucket when acquired by atlassian ). 



>  - how we can better advertise our infrastructure and services (how
> they work, how they can be used) *wink* and the sysadm team needs
> (more contributions/contributors)

A more complete wiki page should be a start. I suck at graphical design,
and more on a wiki http://mageia.org/wiki/doku.php?id=sysadmin ( i defy
people to find the link to svn without reading everything )

Setting up a websvn ( that I planned to do in my spare time ) would give
us more visibility too. 

Having a clear TODO list would help. 

I guess forcing sysadmin team to do reports, or blog about what was done
could help. Writing better instruction than the one I wrote too ( I also
do suck at documentation ).

Removing useless mail on the ml would help to reduce signal/noise ratio.
( that should be easy once I finish sympa and catdap ).
-- 
Michael Scherer