Software AG has established partnerships to have Tamino XML Server
working with the main J2EE Application Servers : IBM Webspere, HP Bluestone,
BEA Weblogic and Sun iPlanet.
In this article I put some considerations
about those appservers complexity.
I once got the question from a friend:

"Why is it so complex ? Is it needed ?"
Is it the price to pay
for robustness and scalability of applications ?

Battling Complexity

This week I have been testing Tamino
with one of these products (Websphere).
You can not imagine how I found it
hard to have some Tamino EJB up and running.
I'm not an EJB specialist but
I have been reading about EJB infrastructures for a long time. Also I have been
busy with Java for more than four years since I left the University. I worked
on JavaBeans, OrbixWeb, Visigenic, JBuilder, Visual Café, Silverstream,
TogetherJ, Bolero,... I never had such a feeling of complexity.
You need to
go back to your "Java Enterprise in a Nutshell" to understand your first "Hello
World" Deployment Descriptor.

It eventually worked very well but I
wonder how people that don't know XML, Java, IIOP and the rest will ever cope
with those infrastructures... I can tell you I meet a lot of people that can't
write a XSL stylesheet and that are very good in IT.

I have met
colleagues at Software AG that don't like OTMs as Corba or EJB. They prefer to
perform stateless Synchonous calls through a series of Message brokers that
integrate complex applications. Why would we duplicate the complexity of
applications into statefull objects ?

Yesterday, in a presentation in
La Hulpe, Arthur Andersen explained how they implemented a WebSphere
infrastructure with Delta Loyds. At the end of the presentation they outlined
that concern : the productivity of the EJB Developments !! They hope that
better development environments will support the infrastructure they adopted.


Scalability ?
On a an other side, I have also been chatting
with some J2EE application server freacks. They claim that appservers are very
scalable, you can implement clusters, components are reusable,... They don't
see the problem in implementing EJBs, "that's easy, after a good training"
(pre-requesites: HTTP, Java, CORBA, RMI, JNDI, JMS, JTA, JSPs,...).
The
message is the following : You will have your business runnning on the internet
24/24 and 7/7, so you will have a lot of connections, a lot more of data. You
need fail-over and load-balancing.
So, you need an Application Server.

Complexity Vs Scalability, Security, High
availability

Again, that is the question about the J2EE appservers
story:
Are we obliged to use complex Distributed Component Models to
have reliable software ?

I would not be surprised if someone
would propose an open source infrastructure that would allow to code regular
Perl, PHP or Java code on top of any database and then deploy the applications
on some Scalable, Reliable, Sercure, Highly Available, Managable Linux Cluster.


But will it be really simpler than J2EE ?

Flag for follow up
!