The face-off between Tim Berners-Lee and Peter Norvig (it is probably too pointed to call it that), came from Tim's AAAI keynote flogging the potential of the Semantic Web, and the importance of various enabling technologies in bringing that to pass. Mr. "When I invented the Web" is the right man for the job. His vision and practicality in bringing us this most wonderously cobbled together platform has succeeded, as wiser pundits than I have observed, in part because of its ability to fail gracefully. Document not found? Look elsewhere... the system doesn't fail, though. Website gone? Pity, but there are lots of others, and the system doesn't fail. The system... the Web... is simple and resilient, and is layered on another elegantly simple and resilient system -- the Domain Name Service.
But the Semantic Web... well, semantics aren't so simple. If the news reports about Tim's talk (and my inferences) are close, the points he made are roughly:
- Persistent Identifiers are critical
- Developers should use semantic languages and RDF
- Tagging will improve semantic efficacy
- The Web is the database, and RDF defines its syntax
Persistent Identifiers are indeed central to the problem. I may assume too much, but I think Tim's perspective is that URIs as they exist today -- essentially, URLs -- are enough for all purposes. On this, I would disagree.
From a technical point of view, he is, of course, correct. The operational characteristics of URLs, in conjunction with the DNS system, are sufficient to meet any identity requirement for online resources. It is in the realm of policy that problems emerge. Tim's assertion that important identifiers be persistent speaks to this, though not convincingly enough in my judgment.
The problem is that Identifiers are overloaded and multifaceted. They play different roles at different points in the lifecycle of the resources they identify. Changes are made to meet the exigencies of changing business models, and naming-theoretic arguments just don't hold up to those imperatives.
Norvig's observations that users often don't understand the significance of the technical decisions that they make (or don't make) bolsters this view.
<unsubstantiated supposition follows>
So, without branded identifiers -- identifiers whose form and syntax proclaim that they are managed according to publicly defined policies -- achieving robust identity networks will be more difficult.
image: Woodpecker, taken in May in the Nisqually National Wildlife Refuge, downstream from Mt. Rainier glaciers. Andy Powell and I walked the refuge on his visit here for the DCMI Usage Board meeting. I should note that Andy's admonition about identifiers... that departures from standard Web protocols in defining identifier systems will inevitably reduce their long term persistence, is always prominent in my thinking.