forrest-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jeff Turner <>
Subject Re: [RT] CLI and Forrestbot (was Re: New Forrestbot in the works (Re: forrestbot configuration))
Date Thu, 02 Oct 2003 09:08:41 GMT
On Wed, Oct 01, 2003 at 11:29:29AM +0100, Upayavira wrote:
> Jeff Turner wrote:
> >On Tue, Sep 30, 2003 at 02:52:30PM -0400, Dave Brondsema wrote:
> > 
> >
> >>I'm trying to set configuration files for forrestbot to generate several
> >>sites.  However, it seems that the <prepare> section of my conf.xml file
> >>has to be set with the same values as I already have in
> >>  Why can't forrestbot just use
> >>
> >>Moreover, how to I specify forrest.maxmemory using the
> >>forrestbot?  I have it set in, but I get an out of
> >>memory error when I run forrestbot.
> >>   
> >>
> >
> >The current Forrestbot implementation (XSLT generating Ant) is inflexible
> >and convoluted.  Over the last few days I've been rewriting the
> >Forrestbot using Ant 1.6's <import> facility.  There's 'get-src', 'site',
> >'deploy' and 'notify' targets in a generic script, and each project then
> >has a script which overrides key properties.  The targets themselves can
> >also be overridden.
> >
> >It's worked wonderfully well.  *Much* cleaner and more flexible.  Ant's
> ><import> is a killer feature.
> >
> >As for your specific question, the generic script <import>'s
> >, so it does use
> >
> >I'll commit my stuff tonight.
> > 
> >
> Jeff,
> Is there a reason why you use Ant for this sort of thing, as opposed to 
> a more pure Cocoon CLI/bean approach? (other than because the CLI/bean 
> isn't good enough yet).
> To my mind, this is the sort of thing that I see the CLI/bean doing 
> eventually. It can already send pages directly to a modifiable source, 
> so with an SCPSource you could remove the final 'copy' stage. There is 
> already a CVSSource which can be used to read out of a repository 
> (although that would require a change to your sitemap, which is not 
> preferable). The cli.xconf can now handle multiple sites (using <uris> 
> elements). Then, once the CLI/bean can do all this, it can either be run 
> from the command line via cron, or potentially be called from within a 
> web interface (run as an Avalon component via the flow).
> Can you see the CLI/bean taking over this role at some point in the 
> future, or do you think there is some intrinsic role for Ant to play here?

Well where does the CLI's job begin and end?  Its contract seems pretty
clear to me; it wants a directory of content and cocoon.xconf file, and
spits out a bunch of content.

But there are jobs to do before and after that.


 - Collate or generate content to be included in the site.  For Cocoon
   this means copying lib/jars.xml to src/documentation/xdocs/.  It might
   mean invoking Javadoc, or invoking Maven targets to generate all those
   pointless stats people love.  All far too general for the CLI to do.

 - More mundanely, obtain a project's static XML from somewhere (CVS,
   scp, ftp).  You're right, this could be done, slowly (opening a ssh
   connection for each file I imagine), with a CVSSource.


 - Notify someone of the render status (via email, <echo>, sms etc).

 - Publish the results (local copy, scp, ftp).

Most of that seems more suited to Ant than the Cocoon CLI.


> Just musing.

View raw message