I've moved the jetty7 integration from my sandbox into trunk.  It's always built but not used by default.

To use jetty7 rather than jetty6 run maven with -Djetty=jetty7 or change the commenting in the root pom to


The Jaspic implementation seems to be working pretty well with the tck.

At this point I'd like to branch 2.2 off and integration the classloading stuff I did in my framework sandbox.  I don't really anticipate any more large-scale changes to 2.2, just fixes for various issues such as the mdb problems.

Alternatively I could create a branch of all of geronimo to play more with classloading.  However I'd rather this stuff was in the bright light of trunk development.

I'd also like to switch to using jetty7 by default.


david jencks

On Apr 21, 2009, at 11:50 PM, David Jencks wrote:

On Apr 15, 2009, at 10:42 PM, David Jencks wrote:

On Apr 15, 2009, at 10:33 PM, Jack Cai wrote:

I agree that a 2.2 release would be nice to do to push out things already in trunk, before our users wait for too long. :-)

I'm reviewing the list of planned features [1] and current status [2] of 2.2. The latter [2] is more up-to-date. It would be good to make clear the areas that need some more work, so that people like me can jump in and help. Currently the major development items I see -

1. TCK, need a committer to do the job
2. MDB problems mentioned above
3. JMS portlets update mentioned above
4. Farm/cluster management (do we still want this in 2.2?)
What's the problem with (4)?

I've been assuming that the classloader work Gianny and I have been working on in my sandbox would get into 2.2.  At the moment I think I have the classloader framework more or less working and I'm going through the plugins working on setting up the required jar dependencies.  Only some of them can be derived from maven dependencies.  This is turning out to be a somewhat slow process.

I finally got the server to run with the one-classloader-per-jar setup.  After struggling with this for a couple of weeks and seeing the difficultly of correctly configuring classloaders I don't  think we should put this into 2.2.  For one thing classloading seems to be pretty slow: it takes about 55 seconds to start the jetty-jee5 server.

At the moment I think a reasonable strategy would be to:

1. branch 2.2 off of trunk now
2. merge in the classloader work from my sandbox framework and local copy
3. upgrade trunk version to 3.0-SNAPSHOT
4. work on using osgi classloading instead of our homegrown solution.

For 2.2 it would be nice to get jaspi officially OK and in.  We finally got the tck from sun.  I haven't looked at it yet to try to figure out how hard it will be to adapt to our tck setup or to run.  If we can get it in we can probably also get the jetty 7 integration in.  Doing this before (1) might be a good idea.

david jencks

david jencks

And of course there are also testing and doc work.

Please complement and elaborate if necessary.

[1] http://cwiki.apache.org/GMOxPMGT/geronimo-22-release-roadmap.html
[2] http://cwiki.apache.org/GMOxPMGT/geronimo-22-release-status.html

- Jack

2009/4/16 Kevan Miller <kevan.miller@gmail.com>

On Apr 15, 2009, at 11:29 AM, David Jencks wrote:

On Apr 15, 2009, at 8:23 AM, Donald Woods wrote:

Should we try reverting trunk (2.2) to use the same levels of OpenEJB and Axis as in the recent 2.1.4 release, to see how close we would be to a release that passes the TCK?  That way, ActiveMQ 5.3-SNAPSHOT would be the major difference left to resolve for a 2.2 release....

I think it would be more worthwhile to look into what is going wrong with the mdbs.  David Blevins doesn't think any mdb-related openejb code changed and ActiveMQ broke at least one other thing since the last time mdbs worked well.

I agree. FYI, I tried to get TCK fired up, but am having some issues. David, have your run tck recently? Let's discuss on tck mailing list...

What's the status of JMS resources and the Admin Console? Seem to recall some missing function...