I can hardly believe that it is already nine years ago that Ericsson Research hired me to do a project called Perspective to write a 3D user interface (VRML) to personal information store. At that time I was living on the Swedish west coast and flew once a week to Stockholm. After a few weeks the manager Anders Danne asked me to join the department for a planning meeting in the beautiful Baltic skärgarden (archipelago) and give a presentation. Surprised I asked about what subject, “anything”, he answered. “Great”, I replied, “Do I get time to prepare it?”. “No problem, take whatever time you need”, he answered, right the answer I was hoping for. So I spent a week thinking very deeply and came up with a project to create a personal assistant that read your mails, saw where you browsed, knew where you used your phone, was aware of the programs you looked at, and tracked your location. The assistant did not only track all this information, it tried to infer relations, and provide timely advice to its patron. The key driver for me was that in the long run search is just not enough; we need our assistant to tell us things at appropriate times, not just to respond to our queries.
Though I was a trifle disappointed with the lackluster response to my presentation to, I was pleasantly surprised when Anders asked me to run a follow-up project based on the ideas in the presentation. So in January ’98 I sat down to try to code “Marvin”, the name of the alter-ego that should provide all these goodies. (Yes, you are right, that is where the name came from).
Why do I tell you this nostalgic story? This weekend I started going through a pile of CACM magazines and one of them focused on Personal Information Managers. More than eight years later I now see in print many of the ideas that I had, but with so few solutions yet. In my original presentation I sketched a scenario in 2003, assuming I was pessimistic. It is kind of discouraging to see that in even in 2006 most of these concepts are seen as pure research and futuristic.
The biggest disappointment was, however, to see that the ambitions of the projects were pretty low. The all tried to find concrete solutions using XML, folders, databases, RDF, and other stuff. I always intended to use Artificial Intelligence techniques for my project because I think that we need solutions where the computer has some understanding of our daily problems, that it has at least some common sense. Without this, as users, we will spent just too much time categorizing and cataloguing our heaps of information. None of the projects in the CACM even mentioned Artificial Intelligence. This holy grail of computing seems to have completely disappeared from the researchers radar screens.
So why did I leave this exciting and interesting field of research? The reason was that I based my working ideas on small units of Java code that categorized and catalogued information in a specially developed database, called an InfoBase. Due to the diversity of the real world, I had made this a pluggable system so that it was easy to add new plugins written in Java.
Based on this work, I was asked one day to present my ideas at Ericsson in Linkoping, a little village in the middle of nowhere, eh, Sweden. When I arrived after driving 4 hours through the snow, it turned out there was a confusion; they expected me to port the Java Embedded Server to their e-box in one day, and oh yes, they also needed a cool demo. The purpose of the demo was to show to IBM, Oracle, Nortel, Deutsche Telekom and many other companies the concept of a residential gateway. I made the grave mistake of succeeding that grueling day (notice that even Java did not run on the e-box yet). Thereafter I was sucked in the vortex of what became in the end the OSGi Alliance.
Eight years later I can look back at some very interesting years developing the OSGi technology, and I feel I worked on the future. Despite this satisfaction, there is a tinge of regret I did not pursue this incredibly interesting area of personal information management. Well maybe one day a big research company sees that OSGi technology is incredibly useful in this application area.
Peter Kriens
OSGi Evangelist
No comments:
Post a Comment