The Asian Institute of Technology recently hosted the 12th International Conference on Advanced Database Applications in Bangkok, Thailand, where the latest challenges in database design and applications were discussed.

Professor Vilas Wuwongse, vice president for external relations at AIT and the conference chair, said the hottest topic in databases today is the native XML (eXtensible Markup Language) database.

More and more applications are communicating with each other via XML as a medium, and most of these applications store their underlying data in relational databases. The next logical step is therefore to develop native XML databases that can communicate directly with each other over a network, cutting out the need to translate to and from another format.

The most obvious gain is in performance and scalability, but there are other benefits such as journaling and temporal (time-based) features.

Vilas said that even today's relational database was nothing more than a conceptual view of data. Underneath the tables and joins, the database still has to get down to the nitty-gritty and store the information in an efficient way, both in terms of storage and in terms of performance. Ultimately, data is still stored in a binary tree (b-tree) form.

Today's first generation XML databases are simply putting on another layer, translating XML to relational database SQL. However, many of the conference speakers were working on direct XML to native-format solutions. This is still a hot research topic and different methods are still being designed, tested and compared, each with different pros and cons.

XML databases also put a new meaning on the significance of temporal records. While a lot of discussion has been made into embedding temporal data into databases since the relational database days, the concept of time was always something to be added on, rather than being inherent in the data. This is because when a record changes, it changes entirely and the old record no longer exists.

However, because an XML database is essentially a document or story (as opposed to a line) that evolves over time, embedding temporal information about how it is changed is not just more practical, but ultimately more useful.

IBM, one of the conference sponsors, offers what it calls PureXML in version nine of its DB2 database.

The second major thread of the conference was how to deal with the information explosion that the world is undergoing. In his keynote, Professor Masaru Kitsuregawa from the University of Tokyo said that searching will soon be very difficult and that even the best search engines, such as Google and Yahoo, were already showing the limits of what is possible with today's technology.

The results of a search can be measured in two ways: recall and precision. Getting 100 items out of 1,000 on the web is a 10 percent recall rate. On the other hand, a 10 percent relevance rate, getting 10,000 hits of which only 1,000 are relevant, is not useful either.

Japan's Ministry of Economy, Trade and Industry, Ministry of Education and Ministry of Foreign Affairs have announced a major research drive into developing a better search engine that not only understands keywords and context, but entire sentences and relationships.

One of the other interesting topics was that of image and video search and developing ways to index and search through that data.

Vilas admitted to being a little disappointed that not many Thai researchers were interested in the conference. He said that the greatest number of papers this year was from China, at 34. Aside from his own team at AIT, he said that only Chulalongkorn University was doing some interesting research into time series data.