Gilles Scokart wrote:
-----Original Message-----
From: Steve Loughran [mailto:[EMAIL PROTECTED]
Sent: Wednesday, November 08, 2006 12:37 PM
To: [email protected]
Subject: Re: future Ivy development
I'd probably put priority on getting the first edition under
the apache
banner out the way; build process, testing, documentation and
the like.
A goal of a 2.0 release would be to get more developers than
just Xavier.
-steve
I think you get a good point there. The question is : what is the best
aproach to achieve that?
From my point of view, I already had the oportunity to look in the sources,
and I tried to made some changes to it. But I only managed to post 1 or 2
patch because I was not sure to break any flow of execution. The problem is
that there is so many way to use ivy. Much more that what I'm aware of. (on
the other hand, it is also for this reason that I like ivy).
For that we need test data; not just use cases but real ivyconf.xml and
ivy.xml files used in various projects.
any OSS project that uses ivy should be under the gump, which can be
used for regression testing nightly builds. We can set up test projects
there too. Now gump is wierd as it sets up its own classpath, and hides
problems (missing artifacts) that you'd otherwise see. So the test
projects need to be able to make assertions about things on the
dependency graph that are independent from ant's special
classpath-under-gump logic.
I like the idea of easyproglife to start by listing some use cases. I think
this can be a good starting point to clarify the different possible usage.
+1
Rewrite (or refactor) a 2.0 version would also make it easier for new
developpers to join.
Prefer refactor/evolution over rewrite. The costs of a rewrite are
pretty steep; look at Axis2 is a case in point.
