Jordi Gutiérrez Hermoso <jord...@octave.org>:
> 2012/4/16 Sylvain <b...@beuc.net>:
> > Next time,
>
> How about we fix Savane, this time, or whatever other time? I want a
> better Savannah. I sent the student your way so you would figure out
> another way to fix Savane. What do you suggest should be done? Where
> are these plans of yours for fixing Savannah even laid down? The
> webpages apparently have plans laid down that don't coincide with your
> vision, so what is your vision, and why isn't it written down in a
> prominent place?
>
> I want code. How do we get that code?
>
> - Jordi G. H.
There was a point at which I began to take a swing at fixing Savannah.
But I abandoned the idea. I don't think the architecture is salvageable -
all that PHP talking directly to the back-end database, no proper layering.
This lack has practical implications for supporting things like remote
scripting of forge actions via XMLRPC and mail robots. You can't do it
without the missing middle layer of coherent data structures. The set
of cron jobs used for doing updates that really ought to be triggered
directly by user actions is another symptom.
In my opinion the whole architectural lineage descended from
SourceForge has reached hard limits; Savane's problems and the immense
difficulty of fixing *anything* are a symptom of this, and not really
the maintainers' fault. And yes, I've looked at FusionForge - I think
it's a heroic but ultimately doomed effort to save a broken
architecture by wrapping more duct tape around the weak joints.
I think there's a crying need for a clean-sheet forge design based
around something like an ORM and designed from the ground up for
scriptability and lossless import/export of project metadata. I've
already done some of the hard parts with forgeplucker. Even so, in the
normal course of events we'd be looking at 4 or 5 man-years of
development work to even get to a beta. But...
There's a special-purpose ORM-like thing, a message queue manager
called "Roundup" (presently used to implement bugtrackers) which
already does about 85% of the needed things and has sound internal
architecture. As soon as I can get my decks clear - maybe around
mid-year - I plan to start writing a forge layer over Roundup.
I think a year's steady work on this would obsolesce all the existing
designs. No, I'll be plainer than that: it will nuke them from orbit.
--
<a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>
_______________________________________________
Savane-dev mailing list
Savane-dev@gna.org
https://mail.gna.org/listinfo/savane-dev
