A conference like no other

Apache probably is the strongest brand in the open source space, but the conference itself focusses strongly on content. Here you will see no sponsored talks by commercial vendors, no sales people trying to sell you anything, it’s all about the code, the community and collaborating with each other. In that sense it’s quite different from most other conferences and if you like meeting and discussing fellow developers, this is a great place to visit. Many events facilitate discussion, and power and internet connectivity are available everywhere.

What open source is all about

Brian Behlendorf summarized the three main cultural elements of Apache quite well:

• write good code and debate it to the bone
• be humble
• collaborate

In essence, Apache is a meritocracy, of which only individuals can become a member. It’s sometimes also described as a do-ocracy as projects are driven by contributions: if you want something done, just do it. Another important aspect is that everything that is done on the Apache projects is discussed and archived on the mailing list. All discussions, code diffs and decisions must be recorded there.

Presenting Apache ACE

Tuesday evenings “birds of a feather” session featured a discussion about Apache ACE, where questions mostly centered around the use cases for ACE and possible integrations with other OSGi components. One of the conclusions is that there are probably three different phases of deployment:

1. Using Apache Felix File Installer, which allows you to drop components in a local folder to have them installed.
2. Using Apache Felix Karaf’s provisioning components, which allow you to define features which basically group components and allow you to define dependencies on other features.
3. Using Apache ACE, which allows you to group components and automatically deploy them to many remote systems.

Friday’s OSGi track started with an introduction to OSGi and moved into more advanced topics during the day. The Apache ACE talk was received well, with several people expressing an interest in wanting to use it and contribute to it.

Final thoughts

Summarizing the week, Floris and I had a great time talking to many interesting people and learning about various projects. ApacheCon is a great conference, and I’m already looking forward to the next one.

Just as I expected, general attendance at JavaOne was dramatically low. Where one had to check in on sunday or monday on previous years to avoid long lines on the morning of the keynote, I could walk up to a counter directly, sign up, collect my goodies elsewhere and be in the keynote room only a minute or so later. It’s a funny feeling when compared to other years. It all went so quick I almost missed a big sign of Hans that was put up outside the Moscone center. Appearantly, he symbolizes the Java community!

With Oracle still in the process of buying Sun, I couldn’t help but feel that there was less news to announce than normal. The opening keynote certainly was pretty boring, partner after partner announcing how great their relation with Java and Sun is. At the end, Larry Ellison did appear on stage, together with Scott McNeally and Jonathan Schwartz. Not sure if it was bad rehearsal or a total lack of chemistry, but it came across a bit awkward. The other keynotes were no better, the exception being the final one by James Gosling, who did manage to show quite a few cool projects.

Appstores were definitely a big theme during the conference. If you did not have one yourself yet, you were definitely not cool. I signed up for the Sun one, but I don’t think it will become a big success. There are not that many killer desktop apps around in Java, especially not ones I would want to pay for. But then again, I never got the whole ringtones market either, and that’s huge.

A couple of highlights of the rest of JavaOne:

• Rich Hickey, author of Clojure, stating that mutable stateful objects are the new spaghetti code, dismissing Java and object oriented languages in general in a future where concurrency and parallelism are more and more present.
• At zembly.com they have tools to search for web service APIs and test them. If plugged into an IDE you can even drag and drop them into a source file and get a snippet generated automatically.
  – After a talk about Filthy Rich Clients at earlier conferences, Guy and Haase did a talk about applying the 12 Disney animation principles to user interfaces. Interesting concept, but they should have spent more time coming up with good examples. Also, two or three bad jokes per presentation are enough guys!
• Don Brown showed the new Atlassian plugin framework, and gave a good overview of some of the pitfalls of using OSGi. The plugin framework can be dropped in arbitrary applications, and he demonstrated that live on stage.
• Glassfish V3 was demo’d at one of the keynotes, running on Apache Felix, and launching in less than 2 seconds, showing almost instant develop/compile/run cycles by updating bundles automatically from within an IDE.
• Runescape, a massively online multiplayer game, designed to run on stone-age hardware. What impressed me most was the fact they created their whole 3D authoring environment themselves, instead of relying on industry standards like Lightwave or 3D Studio, claiming to work a lot quicker that way.

Want to know more? Come and visit us in Arnhem for the JavaOne Reloaded event. For more information, see a previous blog entry.

Today, the long anticipated Bug arrived in the mail. I met Ken almost a year ago now, at EclipseCon, where he did a very nice demo of a new device. Call it LEGO for adults. The box contains hotpluggable hardware modules, such as motion sensors, a touch screen, a GPS locator, a generic I/O board for custom extensions which makes the hardware as dynamic as the software inside. That software, of course, is OSGi based, so there is full modularity there too. Add a well designed SDK for developing and deploying bundles for the system and you have a system that is very appealing to developers and can be used for all kinds of technology demos. Today’s blog ends with a nice picture of the opened box, showing some of the goodies inside. Expect more in depth articles in the near future!

Jon Bostrom did a talk on the state of OSGi on the mobile phone as used by Sprint in their Titan platform, for which an SDK will be released in the upcoming weeks. He envisioned a more open phone where people can actually add and update middleware services. A local webserver can even be used to share stuff with others directly. All of this builds on OSGi, the security model, and things like deployment packages. His main argument for all this openness is that a phone should be able to adapt to all kinds of new technologies as they become available, and that means making everything optional and updatable. That means the platform will be open to for example different UI technologies, like eRCP/eSWT, Flash, JavaFX, JavaME and basically anything else that will come out.

A big part before and after lunch was taken up with discussions with all kinds of people in the hallways. The sessions at Øredev also allow plenty of time to discuss things between sessions, which is a very smart move by the organizers. Later in the afternoon, BJ Hargrave presented a talk about OSGi now and in the future. Most of us know the now, so I’ll focus on the future. Release 4.2 is being finalized for a late Q2 2009 release. This is a release which focusses on OSGi in the enterprise. The enterprise expert group formed two years ago now and the new things in this release will be: a distributed service registry, including service registry hooks, a spec called “blueprint” which basically standardizes some of the Spring DM dependency injection stuff, a bundle tracker, a standard mechanism for launching multiple frameworks in a single JVM (which felix already supports) and a unified launcher.

The features for release 5.0 are a bit more vague. Nothing has really finalized, but there are a couple of interesting things ahead that basically evolve around a framework 2.0 version, that breaks backward compatibility and adds modern features like annotations, collections, etc. Still, this new framework will probably expose older versions of the framework packages too so it won’t totally break backward compatibility after all, something that’s only possible in an OSGi framework!

The closing keynote by a guy from Oracle about identity almost bored me to sleep, but the beer kept me awake for the dinner and subsequent Boom Chicago stand-up comedy, which gave us a lot of laughs in a very interactive program.

The final day started with an opening keynote with great food for thought from Robert C. Martin, who talked about the craftsmanship called software development. He had a couple of important points to make, that all evolved around being agile and doing test driven development. He believed that iterations should be short. One or two weeks. He also believed strongly in writing tests first, writing no more than necessary for the test to fail and then implement no more than necessary for the test to pass again. Doing that means that you always have code that at least “worked a couple of minutes ago”. As simple as that might sound, it is a very valuable concept. There were a lot of other topics in his talk, such as taking personal responsibility, always doing the best job you can, writing great code and always making sure the code you check in is at least a bit better than the code you checked out. He also stressed the need for constant improvements on the codebase, because you never get it right the first time, or the second, or the third. There is no such thing as perfect code. Finally, he commented on what he called “the grand rewrite”, completely redoing an existing project again from scratch, and told us it is a very sure, slow and painful way to complete failure.

After the keynote I attended sessions about customizing Silverlight 2 controls, developing an iPhone application from scratch in Objective C, and a second session by Jon Bostrom on OSGi. This time he got more technical and it was nice to see a general interest from the people in the room when he mentioned things like deployment admin and its role in provisioning software. I attended a second talk by Terence Barr, about LWUIT, but Terence gave the impression he was tired and showed mostly demos he’d shown in an earlier session already.

A session by Mike Jennings about Google Android, even though it was in the last time slot, was absolutely packed, which shows how much interest there is in this open platform. Mike explained how to do development for Android, showcasing both command line tools and Eclipse integration and talking us through an application that uses the tilt sensors and touch screen. He stressed that Google has no interest in controlling what applications you put on your phone, a strategy that is clearly pitched against Apple, who have a very tight and sometimes extremely silly form of control (or censorship) over what goes on the iPhone.

Summarizing Øredev 2008, it was a great conference. The level of the speakers is remarkably high and everything around it is organized well. It was a true joy to be there, to share knowledge and I sure hope to be back next year.

The opening keynote was by Ted Neward who talked about The Next Big Thing(tm), which he thinks is a renaissance of languages. The main reason for that is one that we have acknowledged within luminis, and that is that current languages (Java, C#, …) fail miserably when it comes to handling concurrency and really exploiting distribution of tasks amongst all available cores and nodes. Functional languages like Erlang, which we studied in a research project early this year, cope a lot better with that. With all these languages, the gap between researchers, who create these languages, and practitioners, who just want to use them to get real work done, also becomes more obvious. This gap is mainly caused by different interests of both groups and their inability to effectively give feedback to each other. An example he used was the time it took for object oriented languages to become popular: about 25 years, from the initial Smalltalk language to C++, which first used the concept of objects in a commercial setting. Ted argued that objects are not all that great, and that we probably need other ideas to go along with them to improve upon the way we work.

During the first day, I attended quite a few sessions on different subjects. A session on the Java ME SDK showed the integration of all ME components (CLDC, CDC) in an environment with better emulation and on device debugging support. The presentation itself had some basic demos, and one showing a SunSpot which showed off the sensors API. In overtime they mentioned the BluRay project, which is also supported, and the fact that they will add support for JavaFX. Later that same day, Josh Marinacci actually did a good presentation on JavaFX script, showing only three slides and developing sample applications on the spot. He showed some of the nice features of the new language and also demoed the Photoshop CS3 integration. A third presentation by Sun, done by Terence Barr, contained some demos that exposed the new features of the Java Mobile Services Architecture. A Marble Madness demo, showing off hardware accelerated Java3D and the motion sensor API was nice. He also showed how a single JavaFX application could be deployed on the desktop, as an applet and on a phone, but I’m still wondering who would like to use a UI designed for a phone on a 30″ cinema display. Then again, that does mean you can show about 200 applications side by side.

Rickard Oberg did a very good session on Qi4j, a domain-driven development language that builds on standard Java. Basically it tries to make the gap between code and the domain model as small as possible. He explained how an application, originally created using Spring, would have looked had it been done using Qi4j. It definitely sounds like Qi4j is ready to be used in a research experiment, possibly by also rebuilding some existing piece of code. I discussed this idea with Rickard, who was very enthousiastic about the idea and would love to hear the feedback that came from that.

Ken Gilmer from BugLabs showed “the bug” again. For those new to the bug, it’s basically Lego Mindstorms on steroids, really cool stuff for geeks! Having seen his talk already at EclipseCon 2008 earlier this year, some of it was familiar. The new thing was the fact that he showed they completely published the specs for the “bug” components so anybody is free to create their own hardware modules. After the talk I briefly talked to Ken. They’re still having some issues to solve before they can ship the thing abroad. An interesting thing to note is that he was interested in our device manager donation to Apache Felix, so perhaps we can get some feedback from them on how that runs on Concierge.

In the evening, a session about the principles of open space, coupled with an experiment with that, were conducted, but after dinner I spent most of the time in a 2 hour session with Alef Arendsen who demo’ed Spring DM and did a nice job of explaining how they use OSGi. After a long day, at around 10 in the evening, I stepped out of the conference building for a refreshing walk back to the hotel. To be continued!

Near the end of the day, we showed a demo of the L-iQ Provisioning Server. The interest was pretty overwhelming and we got to show it in action many times, doing demos for three hours straight. Our main demo showed Apache Felix and a simple paint program which visualizes the dynamics nicely.

We ended the evening with a couple of beers on the roof of our hotel with some people from ICW who were staying at the same hotel.

Day 2, still sunny but cooler, started with a keynote by Peter Kriens who took a long trip down memory lane, going back to the days at Ericsson where he first started on what later became OSGi. He ended with a small peek into the future and ended by stating the mission of the OSGi alliance now, which is to become THE component framework.

Later in the day we had presentations on several topics. There were two parallel tracks so I could not even attend everything. A talk on JOnAS 5 gave an overview on the architecture of the first open source, OSGi based application server and how they can dynamically load and unload services. They already use iPOJO, OBR and have a componentized EJB 3 implementation called EasyBeans. Michael Keith from Oracle explained how JPA was adapted to work on OSGi and wondered why there was not yet a standardized persistence solution for OSGi. Jan and Markus gave an interesting talk on the Eclipse Communication Framework and R-OSGi, explaining how they can do all types of distributed services. The Enterprise Expert Group will also address this soon, and we briefly talked to Eric Newcomer about that. Richard did a nice overview of iPOJO, going through over 100 slides in 45 minutes while still making a lot of sense. iPOJO is definitely the dependency and component management solution for the future, as soon as it gets an API so we can use it instead of our dependency manager.

From the various sessions we visited, there were some that were particularly interesting and worth mentioning here.

For everybody who has ever built protocols, Apache Mina is a great framework. It abstracts away most of the complication of the NIO API and contains many building blocks for creating scalable servers and clients that use a custom protocol.

If you need to do enterprise integration, and you’re familiar with the Enterprise Integration Patterns book, then Apache Camel will help you write these integrations. Camel itself is packaged as a set of Spring components that can be integrated into any application easily. Configurations are either done using XML or with a fluent Java API. The latter looks really great and allows you to express complicated rules in a very easy way. Another interesting thing you can do with Camel is implement Business Activity Monitors (BAM). These are rules that trigger exceptions if some condition is not met. For example, you can specify that whenever a message A comes in, you would expect a response message B within 3 seconds. You can furthermore say what should be done if that is not the case, for example post an alert message C.

One of the Apache projects that has really adopted OSGi and Apache Felix is the ServiceMix project. They managed to implement JBI on top of OSGi and are working with the spec leads to make JBI fit more naturally with OSGi (which currently is not really the case). This is really one of the projects that will help move OSGi into the enterprise.

The Apache Sling project had an interesting BOF. They are close to doing their first release and are trying to grow their community. One of the nice aspects of Sling is that it packages Apache JackRabbit in OSGi bundles. Of course this should really happen within JackRabbit but it’s a great initiative. Sling itself is a web application development platform that uses REST principles to build content repositories.

Apart from the sessions, we also had several talks with people about Apache Felix. During the BOF about Apache Sling we discovered they had a very nice web based management console for OSGi. Since this was not in any way Sling specific, we agreed to donate it to Felix, as it would be a nice addition to our text and Swing based shells. You can follow that process on either the Sling or the Felix mailing list.

Furthermore, we discussed the possibility of a simple, pluggable installer that combines efforts from Peter Kriens, ServiceMix and Sling to create an easy to use, convenient installer that will speed up development.

A third interesting topic was the issue of testing OSGi bundles on multiple frameworks and rebranding. Even though we at Felix ensure we only use standard OSGi features that are part of the spec, some people think that bundles that are part of the Felix project only work on Felix. That’s of course not true, but we don’t explicitly test and guarantee the working of these bundles on other frameworks either. So, like with many hardware components that often get rebranded, we discussed the value of doing this for OSGi bundles. That way, companies like JayWay and luminis could provide branded bundles to their customers and make sure they comply with their quality and testing standards.

As a general trend, a lot of projects are now moving towards Java 5 as their minimum version, mostly stating generics and the concurrency API’s as their prime reasons. This poses an interesting question for OSGi development, since not all embedded JVM’s are Java 5 compliant yet. Should some kind of new Foundation profile be defined that includes the most important features of Java 5?

Finally, we showed some people our software component provisioning server and got positive reactions and some good feedback. In general we had a great time and learned some new things. Next year, we will try to organize more events around OSGi and Apache Felix in a further attempt to grow our community.

In case you did I’d like to urge you to consider our talk about OSGi security.

The program looks very interesting and OSGi is well represented throughout the conference.

Marcel Offermans and Karl Pauls will give a tutorial on the first day about Building Secure OSGi Applications.

This Tutorial focuses on embedding various OSGi framework implementations namely, Eclipse Equinox and Apache Felix, into applications as a means of plugin mechanism while taking advantage of the often overlooked benefits of this solution: security. The OSGi specifications have an extensive and very powerful security model that eases the difficult task of building secure, dynamically extensible, applications so don’t miss out on this opportunity.

People in and around the Netherlands, however, have another possibility to get to see this as we will give a boiled down version of the tutorial at the NLJUG jspring 2008 as well.

Thats all for today and we look forward to seeing you in Santa Clara!

regards,