Ivan's Blog

I read an article by Rob McCool entited “Rethinking the Semantic Web” in the latest (January/February 2006) issue of the IEEE Internet Computing magazine. One central aspect of the paper is that it takes it as self-evident that “RDF is complicated”, ergo the “the Semantic Web is way too complex”, ergo something new has to be found and used instead

I would really like to see this misconception disappear. We should make people realize that, in fact, RDF is simple. After all, all there is to RDF are triples. A triple (s,p,o), as it is usually denoted, is nothing else than a named link between two “things” (“web resources”) or possibly binding a resource to a simple a. In some ways, not much more complicated than hyperlinks, except that there is a label added to the link. Nobody considers the concepts of hyperlinks as complicated any more… Or, to take another example, all techies have worked with property/value pairs for years, let that be in Java libraries, XWindows preference settings; well, an RDF triple is not much more than that except that it uses URI-s to name things. RDF is just a set of those triples or, if you are of a visual type, a directed and labelled graph representing those. That is it! The rest is syntactic sugar. Whether the RDF triples are encoded in RDF/XML, turtle, microformats, you-name-it, is a matter of convenience, of good tools, and of personal taste. Why trying to find a controversy (e.g., heralding microformats as something fundamentally different) when, in fact, there isn’t? Don’t take me wrong: microformats are really good stuff, that is not the point. But let us just realize that they are all part of the same landscape…

Of course, I am no fool. I know that the RDF/XML format is, well, not the prettiest syntax around even if one is used to XML. And that the RDF community did not make a very good job at the dawn of the Semantic Web to clearly separate the (simple) model from the (complex) syntax in that particular serialization. But this should be put behind us. I also know that reading the formal semantics of RDF is certainly not easy, and that is probably a British understatement. But… how many SQL users have read the formal semantics of SQL? How many users of a particular programming language have read the formal semantics of that language? Those documents are for the few implementors, not for the lambda users…

And, of course, there is the issue of using ontologies, RDFS, OWL, all those complicated things… Yes, those can become complex. But that is always the case with everything we use. A fundamentally simple and very popular programming language like Python has very complex constructs, but not everybody will use those. Similarly: nobody is forced to use RDFS or OWL; you can be a perfectly respectable Semantic Web citizen if you use the basic RDF only which, I believe, is simple and powerful.

The Semantic Web, in my view, has great potentials. Let us not spoil it with such misconceptions… and let us get the message out on this!

Category: /WorkRelated/SemanticWeb; Posted at: 12:43 UTC; Permalink

I scanned through the proceedings of a Eurographics/ACM Siggraph Workshop on “Computational Aesthetics 2005”. Lots of intereting stuffs, although a large portion of the papers were, actually, on non-photorealistic rendering, but that does not make them less interesting.

A paper that caught my eyes was T. Isenberg et al “Breaking the Pixel Barrier”. The authors argue (rightfully so…) that vector graphics is a good thing, that we should not be blindfolded by pixel-based graphics, that vector graphics is better for zooming, no need for re-sampling, etc. And I suddently feld awfully old. The authors seem to have forgotten that computer graphics started with vector graphics, and that pixel based graphics came only later! I started computer graphics with line drawing displays in the early 80’s, pixel based graphics came much later (and it actually took some time for the CAD community to accept them…). Lots of good work done in vector graphics was pushed aside by pixel-based graphics, photorealistic rendering… Anyway. It is good we remember these things and that the non-photorealistic rendering community may revive vector graphics.

Oh yes, and… the hope for the authors is that Windows Vista™ will have a vector-based rendering engine. Which, by itself, is true and has its advantages, but never heard of SVG, and its usage in Firefox 1.5 or Opera 8? This reminded me that it would be nice to take some of the NPR tools and see what it would require to add an SVG output…

(Unfortunately, though the workshop proceedings are online, only EG members can get to the full papers.)