Author Sing Li has posted a tour of Geronimo's architecture on IBM's developerWorks site, including rationale and a fairly in-depth examination of the subprojects that are included as part of the open-source application server.

He discusses the difficulties in writing an application server and some of the difficulties such a project faces, as well as discussing specific Geronimo project technical goals that required writing a new open-source application server in the first place: scalability, manageability, and configuration management are highlighted. A sidebar points out that the Apache License is used throughout Geronimo as a primary value-add, similar to the reasoning behind Apache's Harmony project. Both Geronimo and Apache share Geir Magnussen, Jr., as a project lead.

He also discusses Geronimo-specific jargon to enable developers to understand what elements go into the Geronimo server.

With IBM now employing many of the core Geronimo developers through their acquisition of GlueCode, Geronimo's development and documentation processes are likely to advance quickly. Already, for example, there's an Eclipse plugin for Geronimo available from IBM.

This has the potential to tilt the open-source application server market away from JBoss slightly, because of IBM's marketing. However, Geronimo is not certified as a J2EE application server yet.

What do you think are the long-term effects of all of this? Some have said Apache is being used by IBM to somehow wrest control of Java away from Sun. Do you think this is true? If so, what does it mean for Java, or for Apache?

Resources:

• Jeremy Boynes on Geronimo, also from developerWorks

I just wanted to remind the news author that JOnAS exists (J2EE1.4 certified) and is included in redhat distribution !

For me, I feel that Geronimo, JBoss and JOnAS are growing fast !

After OS and Database, seems it's the middleware layer who is going to be "attacked" by open source. I think the market share of open source products in those layers will soon overcome closed source softwares.

Around me, I see more and more servers deployed with Linux and PostgreSQL (or firebird, or mysql). I think that, soon, more and more deployed application servers will be open source.

I think BEA should join Objectweb and contributes like redhat does :)

Stéphane TRAUMAT
scub.net
JOnAS Live author

may the source be with us!

Take a look at the release notes. They advertise the missing features. They are not afraid - they are not hiding anything. And it has a transaction log.

They have the proper focus from the outset.

Why Geronimo has got so much publicity? Because the credit of Apache (good), which is an independent software house. If IBM's acquisition of Gluecode changes that, Geronimo may lose support. IBM knows that. This is why IBM lets Eclipse goes as an independent oranization. There are many plug-ins of Eclipse for IBM' competiters.

If Geronimo keeps its independency. Any one: IBM, HP, SUN, you or me can use Geronimo as a base, putting nice user interface on it and sell it calling it MyJ2ee or YourJ2ee server.

Even you do not want to provide a product based on Geronimo, you still can sell service. Let's see the automabile market. Ford provides cars. Ford provides services to the general public though its dealers. But there are many many independent, smaller companies providing service for Ford cars (and other cars).

IBM will have stonger influence to J2EE with Gluecode. Who knows how IBM will play with that.

IBM also wants to use Geronimo. A few years ago, IBM's java IDE was VisualAge. It spent huge amount of money to advertized it. But one day, IBM said VisualAge was discontinued and recommended Eclipse. Will that happen again?

Wei Jiang
Perfecting J2EE!

A few years ago, IBM's java IDE was VisualAge. It spent huge amount of money to advertized it. But one day, IBM said VisualAge was discontinued and recommended Eclipse. Will that happen again?Wei JiangPerfecting J2EE!

Well, Eclipse actually is what was originally planned to be VisualAge for Java 4.0.

The different versions of IBM VAJava was written in smalltalk, and the new Java-based rewrite of VAJava was eventually open-sourced an released as Eclipse 1.0. IBM still kept a lot of Eclipse extensions proprietary, and bundles lots of these proprietary extensions together with the Eclipse base platform as a commercial product called "WebSphere Studio Application Developer".

Seems like they intend to do something similar here. Have an open-source basis platform they can use as an entry-level application server, and then lead enterprise customers onto their expensive enterprise-class applications servers. And just like in the Eclipse example, other commercial (and open source) actors will be free to build other products on top of the same base platform.

Sound like a good solution for all parts to me :-)

How many other companies has IBM bought only to have any innovation grind to a halt?

I don't know...how many? If innovation has ground to a halt, then Eclipse must be a product of my imagination.

How many other companies has IBM bought only to have any innovation grind to a halt?

I don't know...how many? If innovation has ground to a halt, then Eclipse must be a product of my imagination.

In all fairness, the answer doesn't have to include all companies IBM has ever bought, for the point to be relevant. If it's a significant number - five? ten? one? - then it's worth considering.

How many other companies has IBM bought only to have any innovation grind to a halt?

I don't know...how many? If innovation has ground to a halt, then Eclipse must be a product of my imagination.

In all fairness, the answer doesn't have to include all companies IBM has ever bought, for the point to be relevant. If it's a significant number - five? ten? one? - then it's worth considering.

To buy an existing company with a user base to prevent them from competing too strongly with you is a technique often used by most large companies.

But in this case the bought company isn't really a competitor to IBM, and it seems more like IBM bought them to speed up their development rather than slow them down. After all it is much better for IBM to be a defining actor on the development of such a new open source project than to let it happen in the 'competitor camp'. In addition IBM will probably want to capitalise on being able to build on these technologies themselves, as well as putting themselves in the best possible position for users that wants to upgrade to 'enterprise-class' appservers later on.

Observing last few posts regarding Geronimo, common theme is calling it as a "entry-level" AS.
Now what the heck this is supposed to mean?
Everyone knows what kind of shit IBM WebSphere AS and you call it as enterprise AS and call Geronimo as "entry-level".
I guess the quality of codebase and the coherence and modularity of architecture does not contribute to be an enterprise AS in the posters' view!

Everyone knows what kind of s*** IBM WebSphere AS and you call it as enterprise AS and call Geronimo as "entry-level".

Can you clarify this ??

WAS is terrible.
We have been using it in production for a couple of years now and it has been, and continues to be, very painful.

Installation is complicated.
Any other appserver, you can pretty much have a unzip installation - certainly for cluster slaves. There is nothing in the installation that cares what machine it is on. Not with Websphere. The best you can do is have a scripted deployment. And have fun there... We had a sequential army of IBM consultants come in to deploy Websphere - each with their own automated scripted installation.

Deployment Manager is well flaky.
Frequently doesnt deploy to all members of the cluster (error deploying on one node, but doesnt report it).
Leaves machines in inconsistant state.
"Resynch nodes" seems to be the voodoo cure - to be applied whenever anything doesnt seem to be working quite right.

Websphere pollutes the Application Classpath with old versions of common libraries. We had an app which wouldnt run on websphere because the app was using ORO - and websphere had some old ORO version hidden inside one of its own jars.... thanks IBM!

JVM locks up during GC.
Fair enough I have seen other JVMs crash and lock up - trouble is with WAS, you have a choice of 1 JVM...

Error reporting is diabolical.
It will silently fail to deploy an application - no error nothing. Or the best is when you get "Error deploying application"... and thats it! No more useful information! It obviously had more information when it generated that useless error message - but chose to keep it a secret.

The list goes on. We thought we could live with Websphere - but dear o dear what a PITA....

If Geronimo is the entry level, Websphere is the exit level...

-Nick