On Fri, 2003-03-14 at 18:17, Dominique Devienne wrote:
> Could you please provide a link where to start looking to try to understand
> how you are doing it? Like maybe a ViewCVS link? Thanks, --DD

Right now we use three approaches we have a WerkzResolver for
short-lived graphs which we use for the CLI and a resolver based on the
graph package in the commons-sandbox which will be used for long-lived
graphs:

http://cvs.apache.org/viewcvs.cgi/jakarta-turbine-maven/src/java/org/apache/maven/jelly/tags/maven/

What's in the Plexus plugin is just brute-force walking through POMs. I
will eventually incorporate the use of the commons-graph package.


> -----Original Message-----
> From: Jason van Zyl [mailto:[EMAIL PROTECTED] 
> Sent: Friday, March 14, 2003 5:08 PM
> To: Jakarta General List
> Subject: RE: so many jars
> 
> On Fri, 2003-03-14 at 18:07, Dominique Devienne wrote:
> > Why can't the POM be stored in the artifact itself? 
> 
> That certainly can be done and has been suggested, but the POM can be
> uesd for so many other things. It's certainly one of the routes we could
> go.
> 
> > Assuming most Maven
> > projects generate JAR artifacts, couldn't you store the POM in the JAR,
> like
> > in META-INF/maven/project.xml, and then use that info at runtime? Seems
> like
> > all you would need is some bootstrap class (a modified forehead?) that
> knows
> > where to find JAR artifacts (the runtime equivalent to the current
> > maven.repository); When requested to start a main class from a given JAR,
> it
> > would read the POM from the JAR, and start looking up the necessary
> runtime
> > JARs dependencies (declared in the POM) in the repository(ies). If these
> JAR
> > dependencies in turn also contain their own POM, you can fully resolve the
> > dependencies (transitive closure).
> > 
> > Or am I completely in left field? --DD
> 
> Nope, that's what I'm doing to create avalon runtimes.
> 
> > -----Original Message-----
> > From: Jason van Zyl [mailto:[EMAIL PROTECTED] 
> > Sent: Friday, March 14, 2003 4:53 PM
> > To: Jakarta General List
> > Subject: RE: so many jars
> > 
> > On Fri, 2003-03-14 at 15:24, J Aaron Farr wrote:
> > > On Fri, 2003-03-14 at 11:14, Henri Yandell wrote:
> > > > at compile-time maybe :)
> > > > 
> > > > But you still end up with jar duplication, you just get to avoid
> having
> > to
> > > > think about it too much when compiling.
> > > > 
> > > 
> > > I agree that maven has an excellent jar dependency solution for
> > > compile-time, but I don't think there exists any (standard) solution for
> > > run-time dependencies in java.  
> > 
> > This is technically not a hard problem. For maven it's more an
> > admistrative problem. Eventually only the compile time needs will be
> > required in the Maven POM but in order to find the runtime dependencies
> > you need all the POMs of the compile time dependencies of the target
> > project. And there exist most of the POMs to do this it's just creating
> > a mechanism where the POMs are stored safely, sync'd properly and a
> > mechanism for retrieving them.
> > 
> > There are glimpses of this already in some of the maven plugins where a
> > single POM is used as the target and then all dependencies are traced
> > and gathered. I'm doing this to produce avalon runtimes with great
> > success and the method will eventually make it's way into Maven proper
> > after more field testing. 
> > 
> > > Using a system level $CLASSPATH variable
> > > becomes painful (as was mentioned).  The general solution seems to be a
> > > hierarchal set of /lib directories.  
> > > 
> > > To have a true run-time jar dependency solution you would need a
> > > standard installation and launch mechanism.  I suppose something based
> > > on JNLP (WebStart) plus some sort of "ports" or "emerge" system could do
> > > this, but I don't think it exists yet.  Interesting idea though. 
-- 
jvz.

Jason van Zyl
[EMAIL PROTECTED]
http://tambora.zenplex.org

In short, man creates for himself a new religion of a rational
and technical order to justify his work and to be justified in it.
  
  -- Jacques Ellul, The Technological Society


---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to