Re: [Vala] mod_vala (was Removal of Mono)

["Michael B. Trausch" <mbt zest trausch us>]

On Fri, 3 Jul 2009, Jeremy wrote:

> Michael B. Trausch wrote:
> > On Tue, 30 Jun 2009, Jeremy wrote:
> > > > Maybe mod_mono (for apache) is an example of the sort of dependency I
> > > > was thinking of:
> > > > http://ibbie.xanga.com/686874553/item/
> > >
> > >   That would be my blog post. I do still think a "mod_vala" would be a
> > > *great* idea. Unfortunately, I feel it best to suggest that focusing on
> > > improving the documentation might be a better goal. Both flavors of
> > > syntax are great; but anyone coming in, either knowing another language
> > > or just new to programming, will be discouraged when they can't find the
> > > answers. After all, it's really difficult to RTFM when there's no
> > > (finished) FM.  I mean, imagine if you did "man sbrk",  and at "RETURN
> > > VALUE" it just said "To be completed..." (;
> >
> > Hrm.  A "mod_vala" might be okay, but it would be necessarily non-portable. 
> > Why not have a FastCGI library for GObject instead, and then Vala (or C) 
> > can be used to create applications that use the FastCGI protocol run run 
> > behind a Web server?
> >
> > The FastCGI library could be a bare, low-level thing that just implemented 
> > the protocol, and have a layer that sat on top of it to provide more 
> > reasonable interface to the proess and make it easy to work with.  That 
> > would accomplish something along the lines of a mod_vala, but it would not 
> > only work with Apache, but a fair lot of other Web servers, as well.
> >
> >     --- Mike
>
>   You know, I was actually thinking about that after I'd sent the e-mail. 
> Either FastCGI or SCGI should work, and wouldn't be locked into a specific 
> server. There'd be no issue of constantly re-running a binary (ala CGI) since 
> it'd essentially hang out and wait for the server to call it - essentially, 
> the application becomes a sort of data source. At least, that's how I 
> understood it worked, please correct me if I'm wrong.

I am actually working on a FastCGI library for Vala.  I had written one for 
C#, and was using it until a decision was made to migrate the project to Vala 
for technical reasons (native-code software kicks the pants off of managed 
software when it comes to memory consumption, and Vala makes it easy to 
maintain).  I do still very much like Mono.

Currently the FastCGI code base is private, and it depends on two bug-fixes: 
one in GLib, and one in Vala that depends on the one in GLib.  It is going to 
be published under the LGPL after a (rough, initial) port is complete, and 
you or anyone else who desires to use it as a base is more than welcome to do 
so.

After it is ported, I will be making a branch that will work with current Vala 
and GLib versions.  It will require some extra code to work around the bugs 
that there are bugfixes available for in the GNOME Bugzilla.  I am trying to 
advocate for the GLib/GObject fix to be applied, to make the bugfix in Vala 
easier, but as GLib/GObject doesn't seem to have an active maintainer ATM, 
that might not be possible.  I am considering a different bugfix to Vala that 
I may submit that will work around the issue in GLib, but it would be a nasty 
hack and I am not sure that I want to even propose it.

Anyway, working around the issue in the FastCGI library is possible with some 
extra code (and slightly increased memory overhead) and when the port is 
finished, I'll be publishing it as a full project on Launchpad and maintaining 
it actively.  I can publish it earlier if there is interest in helping it 
along, after I have the base spec implementation done---I do plan for it to be 
more than _just_ a FastCGI library, I want to provide some level of HTTP 
request abstraction that is just slightly higher than the FastCGI protocol 
itself.

        --- Mike