forrest-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Steven Noels <stev...@outerthought.org>
Subject Re: Forresbot WANTED
Date Wed, 27 Nov 2002 21:10:37 GMT
Nicola Ken Barozzi wrote:

> But there is a shortcoming: some projects need to create artifacts prior 
> to doc generation: currently the Forrestbot cannot do it, while Gump can.
> Also there is an issue about delivering the site and the usage of 
> resources: using Forrest in Gump runs for Apache projects makes this 
> scenario all easier to set up.

Easier?

Do you mean Gump will do Javadoc generation? Or unit test runs? And 
which distribution methods does Gump currently offer, apart for what it 
needs for its own results (result web pages and nag emails), in 
comparison with Forrest?

Forrest is currently able to use cvs, local copy, scp and possibly other 
Ant tasks as delivery methods. So given the proper runtime environment, 
it is capable of doing most of what is needed for project site 
generation and publishing. Forrest's scope _IS_ site generation.

Now if we need Javadoc, we can add it to the Forrest generation process. 
But still, I think we might be mixing the concerns between a lot of 
different functions:

  * cross-building HEAD CVS trees
  * building nice websites from a collection of xml documents
  * aggregating & filtering Apache HTTPD usage stats
  * running unit tests
  * syntax-higlighting entire source trees
  * ... other future niftyness

Should all of this be done by one system? Nope.

You were the guy who came up with Vindico IIRC. So why should we now 
jump to Gump instead of designing a proper solution? Because Gump is 
there and Gump works? Yes - but it will only work as long as it is used 
for what it has been designed for.

Maybe we are trying to put too many functions in one entity. Maybe some 
superservice (cron) should exist which triggers Ant scripts which drive 
the functions listed above. Or shell scripts if Ant doesn't quite cut 
it. Maybe Forrest is just a service for such an ├╝berserver. But thinking 
Gump is that superserver, or might organically evolve into one, I'm not 
sure.

> Now, there always has been an overlap between Gump and Forrest.
> 
> Gump builds projects with the latest and greatest code, and can publish 
> artifacts of this. It's reasonable that Forrest can be used to create 
> some of these artifacts (documentation).

I violently disagree that documentation is an artefact of code. Javadoc 
is. Documentation has a separate concern, aim, goal, life cycle, etc... 
Why is there an HTTPD Documentation Project? Because it didn't quite fit 
into the normal repository / way of working, I guess... Why is it that 
the idea of having a separate CVS module for Cocoon docs has often been 
mentioned already? And now the Wiki kind-of fullfills that role somehow?

> On the other hand Forrestbot builds sites, but cannot include easily 
> artifacts from project compilation (some kind of catch22).
> 
> Here are some things to consider (all non-exclusive):
> 
> 1) Enhance Forrestbot by making a tag that enables it to run build 
> targets and scripts prior to doc generation. This will create the extra 
> stuff we need. In this way Forrestbot becomes a nightly build system, 
> and the forrest build becomes just one of the possible things that are 
> run on that code.

+1, for a selected number of tasks (Javadoc _might_ be an option)

As of currently, not many people took a thorough look into the existing 
Forrestbot, I guess. Let's try to build on what we have, instead of 
jumping into yet another direction.

> 2) Enhance Forrest by making it able to mount resources from urls, so 
> that systems like Gump can publish the artifacts in Forrest format 
> somewhere and have the forrestbot pick them up.

+0.5

> 3) Simply make projects that are Gumped use Forrest as a site generation 
> system and be able to publish the sites from the Gump run. These runs 
> differ from the Forrestbot ones because they are done without forcing 
> the latest code in compilation.

-1

> 4) Make Gump build its site with Forrest instead of Anakia; this will 
> keep Gump artifacts separate but still skinned by Forrest. The 
> Forrestbot remains the same and publishes only the bare site.

don't see the point in this

> Comments, suggestions, etc.

Sorry ;-)

</Steven>
-- 
Steven Noels                            http://outerthought.org/
Outerthought - Open Source, Java & XML Competence Support Center
Read my weblog at              http://radio.weblogs.com/0103539/
stevenn at outerthought.org                stevenn at apache.org


Mime
View raw message