[Mageia-webteam] Initial hosting requirements for maintainers db

Michael Scherer misc at zarb.org
Wed Jan 12 13:29:53 CET 2011

Le mercredi 12 janvier 2011 à 12:29 +0100, Romain d'Alverny a écrit :
> On Wed, Jan 12, 2011 at 11:52, Michael Scherer <misc at zarb.org> wrote:
> > Non integrated because that use a totally different system ( ie gem/rvm
> > vs rpm/urpmi ) with totally different versions ( ie, defined by coders
> > instead of the one agreed when we decided to use the distribution ),
> > totally different update mechanism, and with different requirements.
> Perfectly aware of that. Or we could as well host it somewhere else,
> just to host the application on a matching stack. I'm not saying it's
> the best solution, but it may be a temporary option to keep open.

Then, as decided in the past for the case of the forum hosting done by
MLO, that would mean no access to ldap if the sysadmin team do not
manage the server. 

However, having a separate account database could be done indeed.
But I think we will quickly agree that it is not a good idea.

> > Another option that I consider would be to use another stack. Nanar
> > already told me that he would be able to do it quite fast in catalyst,
> > and I consider myself being able to do it in django without much problem
> > if I dedicate enough time ( I am quite rusty but for a simple CRUD
> > application, it would be quite ok ).
> Obviously we still have to learn to let it go at some times, delegate,
> welcome newcomers and new ideas, new stacks, if we want ever to grow
> the whole thing. It has a cost (flexibility and work), sure, but it
> has a payback too.

I am just mentioning it for completeness, telling another alternative
that should be also be kept open.

I am not against rails hosting. I would have told exactly the same to
someone asking to use python-virtualenv and django, or similar
technology on any other language.

I just ask that using it do not totally disrupt our hosting and
procedures. And I would not ask if I was not sure that it could be

The current rails stack in Mandriva do not seems to be unusable to the
point to compile our own, as gitorious worked with it among others. And
while I understand that it can be a pain to use older versions of the
stack for a coders, it does seems a so strong requirement. That's still
ruby, that's still rails. 

> It's not by sticking to a single stack mostly only handled by
> sysadmins (Catalyst in this case) that we will achieve that delegation
> and growth. I'm not saying it's crap, I'm saying it does not fit the
> pool of volunteers we have for contributions in web development.

I didn't ask to stick to one stack so sorry if I gave that impression. 

And given the current state of the server, we are already supporting 2
stack ( django is the other one for transifex ), and also applications
without any kind of web stack ( like bugzilla and sympa ), and planning
to support a 3rd one ( symphony for stormi web application ). 

I would even be ok if people used custom fastcgi from my point of view
( but you prefer to have people use a popular web stack, I perfectly
understand that you set some requirement on that part and concur in that
sense ).

I am not even against supporting a 4th stack ( namely rails ), despites
the fact that is is more than most commercial shop, since we engineered
the server for that ( ie fast-cgi ). We anticipated the fact that we
would have to cope with everybody using his preferred stack, be it here
or upstream ( ie transifex vs epoll vs gitorious vs custom php
application ).

So my request is "make and test the application to work with :
- fastcgi, should not be a problem, rails support it
- postgresql ( again, i do not see any complex request that would cause
trouble )
- our version of rails and ruby ( ie 1.8 or 1.9.1 ), and rails 2.3.

I do not have the impression to ask for the moon, I do not ask for a rpm
package to be done, nor to have a rigorous code review regarding
security, or specific development method, requirement on the
documentation, memory usage, time of response, etc.

There is nothing exotic to my eye :
- standard version of ruby 
- a well know database, respecting SQL 2003 standard 
- a specific version of rails, which is unfortunately not the latest

Nothing that can't be installed ( server just run Mandriva 2010.1 so a
vm or rvm could perfectly mimic the setup ruby wise ), nothing that
would requires esoteric knowledge.

Michael Scherer

More information about the Mageia-webteam mailing list