Sorry to intrude but this is not right. This email is CC'ed to jboss-dev in
case someone there wonders about the relationship between the two projects.
> While some folks may view us as a competing open source project, we do
not.
> OpenEJB is designed to be an EJB 2.0 container system only, not a J2EE
platform.
> Proprietary and open source application servers use OpenEJB to obtain
robust and
> performant EJB 2.0 functionality. We don't see our project as competing
against
> jBoss, we see our project as a possible plug-in for jBoss or any other
platform.
> In this way OpenEJB is somewhat analogous to Apache's Tomcat: In the same
way that
> both jBoss and the Enhydra platforms use Apache's Tomcat for servlets,
they both
> can use OpenEJB to get plug-and-play EJB 2.0 functionality. We are no
more a
> competitor with jBoss then Intel is a competitor with Dell Computer.
This is not true. jBoss is an EJB container that happens to have a J2EE
structure around it to easily integrate with other J2EE modules. I.e., while
we do have a horizontal server kernel based on JMX, our focus is most
definitely on the vertical EJB container functionality. Because of this
OpenEJB and jBoss are very much competing projects, although jBoss also goes
the extra mile to provide 1) a usable server around it 2) integration with
servlet engines.
Your comparison with Enhydra and Tomcat is somewhat weird. In the Enhydra
case they don't have servlet functionality. In our case we already have an
excellent EJB container so we have no need for OpenEJB.
> In the end we would like to see OpenEJB used by both Enydra and jBoss as
their
> preferred EJB container system.
I can't speak for Enhydra, but this will not happen with jBoss, that I can
assure you. Why would it? I don't see any good reasons for this to happen
myself.
> To make this a reality we can not fold our
> project or resources into either jBoss or Enhydra -- that would compromise
our
> position as platform agnostic. We have already seen OpenEJB plugged into
OpenORB
> with little or no difficulty and believe that our final release will
quickly be
> adopted by organizations looking for fastest container system available.
Note that jBoss is *also* a pluggable EJB container system. In fact, if you
strip away the extra horizontal server functionality that we provide (JMX
bootstrap, logging, configuration, etc.) jBoss and OpenEJB are very
"similar", i.e. we have the same goal there. The jBoss EJB container
functionality is a JMX plugin to our JMX bootstrap, but can easily be put
into any JMX compliant server (hey, even WebLogic for that matter).
And, although your project seems to constantly state that "it is only an RI"
we instead make sure that what is in jBoss is excellent: we performance test
and tune our code all the time, and also have an extensive test suite to
make sure it work as intended.
How you can claim to have the fastest container system available without
having a real implementaion is a mystery to me. I mean, just what is that
statement based on? The possibility of someone perhaps making fast plugins
for OpenEJB? Feel free to expand on this.
> In addition to "plugability", performance is critical to the success of
OpenEJB.
> We have nearly eliminated synchronization of threads so that the container
is very
> scalable. With improvements planed to further eliminate synchronization
and to
> streamline operations within the container, we believe OpenEJB will be the
fastest
> container system available.
We have made quite a lot of performance tuning of the jBoss EJB container
and it is much more involved than simply removing thread contention
(although that was an important part of it). The jBoss container is now
ridiculously fast and stable. Since we also have in-VM integration of two
popular servlet engines (Jetty and Tomcat) we can get HelloWorld invocation
speeds from servlets at around 0.3 ms per call.
Feel free to show me OpenEJB benchmarks that are better than that, or even
close to it. I will happily see those results, since that just gives us even
more reason to push even further on our performance.
> We would love for jBoss and Enhydra J2EE platforms to use OpenEJB. We
openly
> promote this and are more than happy to work with those teams towards such
an
> effort. It makes sense for them to leverage the work we have done and
benefit from
> our container system. We are ready, willing and waiting to support this
> integration, but we need the support and voices of users like you from
those lists
> to make it happen.
So, to sum up the jBoss project has absolutely no intent on integrating with
the OpenEJB container system. Why? Well, why would we? AFAICT there are
really no reasons for us to use OpenEJB considering that the EJB container
architecture we already have is well-designed, stable, and performance
tuned.
Of course, if you want to integrate with jBoss (i.e. use our horizontal
tools such as JMX logging and configuration) then we certainly won't stop
you. Go right ahead.
I hope this clears up the confusion.
regards,
Rickard
