Article:		Using Topic Maps to Extend Relational Databases
Subject:
Date:		2003-03-09 16:09:46
From:		kent fitch
Great article, but I think this point can be debated: "This approach is so powerful one might wonder whether all data could be stored as topics, associations, and occurrences. Well, it could, but it shouldn't. A Topic Map approach will never yield the same performance as data in tailor-made relational tables." A tradeoff between "power/flexability" on one hand and performance on the other has to be evaluated in terms of requirements (just what performance is required), available hardware, maintainability and simplicity. For a recent project (www.austlit.edu.au), the project team decided that an approach similar to the one described in this article was suitable, and indeed it has turned out to be in practice (for a full description see "Taking RDF and Topic Maps seriously - what happens when you drink the Kool Aid" (http://ausweb.scu.edu.au/aw02/papers/refereed/fitch2/paper.html)). Also relevant are is the Parsons/Wand paper "Emancipating Instances from the Tyranny of Classes in Information Modeling" (http://citeseer.nj.nec.com/parsons00emancipating.html) Kent Fitch Project Computing

Previous Message

Next Message

• 
  2003-04-01 04:16:11 Marc de Graauw [Reply]
  
  Kent, thanks for providing the link to your article. Very interesting.

  
  By the way, you are quite right with your performance argument. Given a suitable problem domain, as in the article you mention, a complete implementation in topics etc. can do the job efficiently.

  
  Cheers, Marc