OSGi and the Java Embedded Server
Many people (ok, maybe the older ones) know that Java Embedded Server was the pre-cursor of the OSGi Service Platform. Few people know what the exact relation is.
Java Embedded Server was a development of SUN. It was the brainchild of Anselm Baird-Smith. I first met Anselm in 1998 in Raleigh, North Carolina. for a meeting to discuss Java for the residential gateway. He was the SUN representative and I was the Ericsson technical representative. I was dragged in the Ericsson e-box project due to my Java experience.
We had this meeting because the business people had given the technical people the assignment to take the Java Embedded Server product and make it ready to be used in a residential gateway. While we were going to be busy with this design work, the business people were supposed to develop the residential gateway market.
Quickly, a core group emerged that was doing all the technical work. Anselm, Ben Reed from IBM, and me for Ericsson. Work was progressing slowly. This was Anselm?s brainchild and obviously he was not going to let the others run over him. Several weeks passed with long dreary phone conferences. At a certain moment we decided to make a shortlist of three items a person so we at least had a manageable list.
During this period, there were troubles on the business side: SUN threatened to leave the OSGi. One day I got a call from the Rolf Johansson, the Ericsson e-box marketing person. He explained the problems and asked me if it was possible to make an alternative to Java Embedded Server together with IBM. Obviously: Yes!
The next few weeks Ben Reed, BJ Hargrave (IBM) and me worked almost 24 hours a day to create the alternative. Fortunately, we all worked in different time zones so we did not completely exhaust ourselves. When the deadline was nearing, Tommy Bohlin of Gatespace also joined our efforts. When we were closing our work, I got a call from the Ericsson guy, telling me that the alternative was no longer needed. Ouch. Do not get me wrong, I was thrilled to hear that SUN did not leave the group. However, what we had found out during our intense work was that IBM, Ericsson and Gatespace were having very similar ideas that did not match the current JES architecture. In the few weeks we had made enormous progress and we all distinctly disliked the idea of going back to the original discussions to make small changes to JES.
What to do? Naturally, we had extensive discussions over the phone. One thing that came out of these discussions was a list of more than 25 changes to the JES architecture that we thought were absolutely required. However, we did not have any backing from the business people to make these drastic changes and we were convinced that Anselm was not going to play dead and let us run over him.
The next meeting was in Cupertino at the SUN campus. Ben, BJ, Tommy and I decided to meet before the meeting in the Peppermill, a breakfast place at the other side of the road of the SUN building. At first we felt gloomy, but the breakfast must have strengthened us: we decided to go for all the changes though we were expecting a big fight.
With a full stomach we entered the building and were welcomed by Li Gong, a senior manager of SUN. This was his first meeting because it turned out that the whole JES group, including Anselm, had disappeared to a new startup. We looked at each other, and at that moment we realized that this was a golden opportunity. Clearly, after this the meeting went extremely smoothly because our opponents were brand new to the problem domain. The rest is history.
Despite the many changes we made during that meeting, and the at least a magnitude larger number of changes we made over the last few years, we owe the core architecture to Anselm.
posted by Peter @ Sunday, February 26, 2006