Archive of UserLand's first discussion group, started October 5, 1998.

Re: What is RDF?

Author:Paul Snively
Posted:8/21/2000; 8:23:49 AM
Topic:What is RDF?
Msg #:19873 (In response to 19859)
Prev/Next:19872 / 19874

Dave Winer: I hope to start a discussion that instead of rambling all over the map into someday panaceas that depend on a lot of magic, focuses on what RDF can do for us today, and what the costs assosicated with that are, so intelligent busy people can make a decision.

My own feeling is that people hear the catchy phrase "Semantic Web" and think, "yeah, that's a good idea, a web of meaning, not just presentation." Yup, sounds great! Where can I get some?

What they may not realize is that they've just landed squarely in the middle of the problem of distributed knowledge representation, an arena that has occupied AI researchers for decades—very nearly as long as there have been computers and networks.

It's no accident that Ghua came from the Cyc knowledge representation project before working on MCF/ProjectX at Apple and thence RDF at Netscape. RDF makes an explicit nod to "frame systems," a well-known knowledge-representation family of architectures in the symbolic AI community.

I see two major problems here:

  1. Knowledge representation is a hard problem; distributed heterogeneous knowledge representation bordering on impossible given the current state of the art. Part of the point of Cyc is to provide something approaching a lingua franca for knowledge representation through its Upper-Layer Ontology. The Open Knowledge Base Connectivity (OKBC) attempts, as its name implies, to be to knowledge bases what ODBC is to SQL-based databases.
  2. Knowledge representation without inferencing is all but useless, and RDF is, of course, silent on the issue of inferencing. To put it another way, Ghua almost certainly represents the entirety of epinions' ratings, accounting, and web of trust with RDF, but he also had to write a lot of software tying it all together.

My RDF vote, then, is for "a nice idea that will probably never go anywhere." Far from being too complex, RDF doesn't go far enough in supporting the construction of a real Semantic Web by specifying a (set of) inference procedure(s) and providing reference implementations at least in Perl and Java.

More to the point, though, I believe that once people find out how complex constructing a useful "Semantic Web" actually is, the idea will fall out of favor and we'll continue to face the limited sort of functionality on the web we already have.

There are responses to this message:

This page was archived on 6/13/2001; 4:56:08 PM.

© Copyright 1998-2001 UserLand Software, Inc.