Hrabri, Andreas and me published a paper at CCGrid 07, which takes place in Rio de Janeiro this year. On yesterday evening, I started a 15 hours trip to Brasilia, and arrived this Sunday morning. The flight was ok, despite the fact that an amazing amount of little kids in the airplane where not able to sleep. Before this trip, most of my colleagues confronted my with news articles about drug dealer wars and shootings in Brasilia. Even the german ministry of foreign affairs suggests to always take enough money with you, in order to be robbed without getting harmed.

I had a nice talk with some brazilian business guy during the flight. He complained that he heard the same question about safety again and again in Europe, but agreed there is some real truth in it. As expected, it depends on the part of the city where you’re going around. The conference hotel is at the safe Barra beach, which has an impressive collection of luxury hotel buildings and fenced apartment villages. It is immediately clear that this is a tourist region, but people are very kind and friendly to everybody.

I got the advice not to talk to little boys asking for something, as well as to leave important belongings in the hotel. The only relevant problem is to get safe from the airport to the hotel. I had the luck to find a shuttle, which saved me from negotiations with some obscure taxi drivers.

Finally, I had a two hours walk on the beach, without any major problem. It is amazing to see how relaxed the people go around here – old people in Germany usually don’t wear a small bikini at the beach, and still look good. The weather is amazing, as well as the nature around here. The ‘cold’ weather conditions – 20-25 degrees – are comfortable for a european guy, but you quickly burn yourself in the sun.

During our Middleware and Distributed Systems lecture, one of the students raised an interesting question regarding message-oriented middleware. I presented CORBA Notification Service and JMS as examples for messaging technology, and explained for both the idea of a predefined message structure for central filtering capabilities.

Our student asked why these filtering capabilities demand a unified message structure on the wire. What about reflection ?

First, it is easy to see that reflection on incoming data demands a structural understanding of the wire representation. This could be either given to the broker in advance (e.g. by providing CORBA IDL to the message broker, or through a central type repository), or could rely on identifiable wire data structures. The first idea does not work in reality, since central type repositories never played in CORBA or other technologies. The second idea needs a cross-product wire format with some TLV encoding, which is not fulfilled by IIOP and the most proprietary JMS-enabled message broker protocols.

From this argumentation, we concluded that even though reflection might be possible, it is really hard to implement with todays systems. Therefore it is easier to combine filtering language and predefined data structure, so that implementors and users can express filters based on well-known message types.