One of the really nice things about ApacheCon is that it makes it possible to finally put a face to a name.  On Day 2 I finally met Peter Royal, a fellow Apache MINA committer.  After meeting Peter, I was introduced to Brian McCallister.  I talked with Brian about my asynchronous database driver project, ADBCJ.  He told me that he always wanted an event-driven database driver.  I showed him what I was doing.  He suggested a different approach that was analogous to to SAX for processing query results.  I really liked the idea.  He showed me code he was working on for doing event-driven HTTP.  I started thinking about how I could restructure my code to support a more fundamentally event-driven approach.  I’ve come up with a solution that will better facilitate my current approach.

On Day 3 I attended Filip Hanik’s presentations on “Hacking Apache Tomcat” and “Zero Latency HTTP – The Comet Technique”.  Through both of his presentations, I couldn’t help but think that using MINA would be a great fit for a Tomcat Connector.  Towards the end of the day I approach Filip and talked to him about creating a MINA based Tomcat Connector.  He didn’t know of anyone working on such a project and indicated that because of the conservative nature of Tomcat, he didn’t think it was likely that a MINA based connector would be included in Tomcat.  This is understandable but it didn’t detract from desire to pursue such a project.

Filip and I also spoke a bit about Tribes and how it works.  Tribes is basically the Apache version of JGroups.  Tribes doesn’t try to do reliable multicast as JGroups does.  It uses multicast for peer discovery and then establishes TCP connections between hosts for communication.  I like this approach because it’s reliable and works for small clusters which is what most people actually use.

Peter Royal gave his presentation on MINA and I think it went pretty well.  It’s difficult to encapsulate the benefits of event-driven networking in general and specifically the benefits of MINA in a one hour presentation and I think Peter did a good job of this.

I attended a presentation on JCR and Jackrabbit.  There’s a lot to like about JCR and it’s something that I need to play with more.

I attended a presentation on Apache Harmony, the Apache implementation of Java SE.  Harmony seams to be coming along nicely.  I really like the modularity of Harmony.  Harmony is modular at both the JVM level and at the class library level.  This is one area where the Java standard could be improved.  The JRE is simply too big and has too much cruft that can’t be removed for reasons of backwards compatibility.  If Java were more modular and each module could specify the version of it’s dependencies, older modules, such as Corba, would only be installed if there was an application that actually needed it.  Additionally, Google is using the Harmony class library for Android which is very interesting.

Finally I attended a presentation on Google and open-source by Greg Stein.  Greg said that about 80% of all Google code could be open sourced.  If the code doesn’t drive Google’s bottom line then why not open-source it?  I think this is something I wish my employer understood better.