- From: patrick hayes <phayes@ai.uwf.edu>
- Date: Wed, 12 Jun 2002 16:55:18 -0500
- To: "Jonathan Borden" <jonathan@openhealth.org>, Jim Hendler <hendler@cs.umd.edu>
- Cc: www-rdf-logic@w3.org
>Peter Crowther wrote: > > >>... I believe humans can, with sufficient >> effort, make *some* stuff work well enough to trust without having the >> formal semantics. In particular, the following aspects make it easy: >> >> 1) Bilateral communications rather than peer-to-peer, allowing effective >> communication between producer and consumer of specification; >> >> 2) Well-understood problem domains, such as finance, giving a higer base >of >> common understanding to start with; >> >> 3) Restricted problem domains, such as a credit card application, giving a >> limited scope for any such communication; >> >> 4) Past experience of similar problems, giving a history of known >solutions; >> >> 5) Shared language between producer and consumer of specification; > >Certainly this is important. But what about "shared language" on the >semantic web ... how does one really define what the semantics of a given >URI is, in a formal sense. Easy. A URI is an individual name. In any interpretation, it refers to an entity in the universe of the interpretation. There may be more to say, of course, but this much enables us to at least get started. >It seems that unless this is done in an >unambiguous fashion, any formal infrastructure built on top is sort of like >rearranging, err straightening, deck chairs ... Good analogy. At least the inference machines can be built so that they keep the deck-chairs straight. Not much, but a start. Better than each deck chair having its own compass. > >> >> 6) Limited scope of implementation, for example a single banking system >> communicating with a central card issuer system; >> >> 7) Limited variation of environment, for example a credit card system that >> deploys particular card swipe hardware and software. >> >> All of these simplifying factors were present in your example. None of >> these simplifying factors are present on the semantic web. I consider the >> comparison between the two cases to be specious for that reason. >> > >I wonder if the semantic web can meaningfully work without some of these >assumptions. An actual example of a working semantic web application that >doesn't make some of these would be helpful in convincing me otherwise. I think everyone agrees that we will need to rely on such assumptions for the forseeable future. I don't quite understand why people seem to think that to insist on the SW content exchange languages having a clear semantics is somehow taking a stance against the existence of social conventions. (This weird idea seems to arise from some equally weird flaw in Jim Hendler's education.) The point is, rather, how to have some confidence that these social conventions are preserved when the people aren't in the room, when innumerable pieces of software are exchanging information with each other on a large variety of topics, drawing conclusions (using techniques, and other information, unknown to the people who wrote the original content) and then passing those conclusions on to other pieces of software, all this happening so rapidly and on such a scale that it is completely impossible for human beings to check it all. (And of COURSE the code was *written* by people. Sigh.) That is the problem that requires us, as designers of the protocols that these pieces of software will use, to make every effort to ensure that the meanings that are finally produced, or the actions that are finally taken, do in fact still conform to the social conventions used by the people who originally composed the content and who write the code which is manipulating it; and that this is true no matter how long the chain of inferences, or how varied the processes and processors that produced them, as long as they all conform to the specs we write. This isn't easy to do in any large network of systems. The SW, if it ever exists, will be a system on a scale hitherto unimaginable. I'm not asking for cast-iron guarantees that it will all work properly (still less for machine-checkable proofs that it will, which is why Jim's reference to Millo et al is completely beside the point here) and of course Im not arguing that things like 404 errors will never happen, or that the world is not a scruffy, untidy, place. I know it is; that is the point. And for the record, I never, ever, use the phrase 'pure logic'. Pat -- --------------------------------------------------------------------- IHMC (850)434 8903 home 40 South Alcaniz St. (850)202 4416 office Pensacola, FL 32501 (850)202 4440 fax phayes@ai.uwf.edu http://www.coginst.uwf.edu/~phayes
Received on Wednesday, 12 June 2002 17:55:21 UTC