Xavier Hanin wrote:
On 11/8/06, Brett Porter <[EMAIL PROTECTED]> wrote:
Is there any scope to revisit collaborating on the metadata?
I think that moving to apache is a great opportunity to collaborate on the
metadata, and if we choose to go with a 2.0 version, it will allow to give
more flexibility to our point of view :-) So that's one more pro for a
2.0version :-)
I know
you ruled it out back in 2005,
I regret you understood it like that, I should had been more cautious when
speaking about that. The problem we identified at that time was that our
users liked the flexibility of our module configurations and configuration
mapping, and your point of view about that kind of flexibility was
different. That's why I thought we both concluded we were not ready to
collaborate on metadata. Anyway now I think both maven 2 and ivy are more
mature, thus it should be easier to identify use cases that must be
addressed and how we can address them.
but you've identified that getting
contributed metadata in has been a problem for Ivy, something Maven
is well familiar with and are (slowly) addressing and making good
progress (especially with tooling in the area now). Seems like a good
opportunity.
Yes, getting contributed metadata is difficult, and Maven has a lot more
experience on the subject than Ivy community has. So I'd love to benefit
from your experience, and so does the Ivy community I guess :-)
Maven is getting into it's next big development cycle so additions/
changes to the POM and revisions to the way the dependencies are
resolved are both up for grabs and high on my list of things to do.
This is great if we are both at a step where discussions can really lead to
change in the tools and a more homogeneous way to deal with that difficult
problem of managing dependencies.
Is it worth discussing?
I have a work colleague who is converting all the maven2 metadata into
RDF, including class information, with which we can mine some
interesting things.
My own goal is to use the data to audit the repo, look for anomalies
(loops, duplicate classes in different project's jars, missing links).
It makes a very good place to see if some of the semantic web vision
-individually published information- really can be merged together to
produce a unified whole, or whether consistency fails down as things scale.
Stefano M is also interested in some RDF-based Gump successor, which is
itself dependency driven.
While I personally dont think that much of RDF, I can see value in
having some metadata info that is entirely tool-neutral, which RDF is,
and in the N3 notation, almost human readable once you've had enough
training in it. Plus swiprolog has good rdf integration, so I can work
with a proper logic language while processing it.
It may make sense to have an RDF form as an interchange representation
of data, between m2, ivy and gump, and any other tools.
Where to discuss this? Here? [EMAIL PROTECTED] the maven mailing
list is too busy for a big discussion.
-Steve
