in reply to
XML for databases?!?! Is it just me or is the rest of the world nutz?
XML and databases includes at least 2 different aspects:
- XML as an exchange format for data, which might come from a DB or not. You just use it as a neutral, standard format, that takes care of annoying things like character encodings. CVS is great if the data source is "close" to the recipient: if they belong to the same organization, if they changes are synchronized, if the quality of the data can be checked and if the recipient can refuse to accept it if it is bad. Otherwise it is much easier to work with XML: the fields are tagged, which means that you can tell easily if one is missing, it is generally easier to deal with changes in the format (don't process extra tags, default non-existing ones) and to work with data coming from different sources (add a simple layer to normalize the data to a single DTD/Schema). You can do this with CVS but generally XML tools make it easier for the programer. Hence "XML for DB: use it as an exchange format". In buzzword-speak I guess this is described as: "XML is very well suited to loosely coupled applications".
- then there are some emerging XML Data Base systems (Berkeley DB XML, XML DB, see Ron Bourret's site for more products). These are used to store XML documents natively. The problem with XML documents is that their model is a tree, which does not fit nicely in relational tables: either you define a schema for a specific DTD, and it becomes difficult to change it, and you loose one of the great benefits of XML, or you design a generic schema, but then, unless you built a whole layer on top of the RDBMS, in fact developing a dedicated XML DB, performances are real bad, your queries need too many joints to be fast. XML DBs are either built on top of OODBMS (Object Store was used a lot for SGML DBs) or on top of RDBMS, but offer a tree model of the data. This can be very convenient for things like technical documentation or even product data (complex products are usually designed in a top-down way, which leads naturally to a tree model, relational databases are used because they are stable and work well, but a lot of CAD manufacturers are moving towards OO systems to store design data). Hence DBs for XML: "Store XML data in an XML DB".
Does this answers your question?