- From: Graham Klyne <gk@ninebynine.org>
- Date: Mon, 16 Jun 2003 18:21:01 +0100
- To: <danny666@virgilio.it>, "RDF interest group" <www-rdf-interest@w3.org>
I think we're in considerable agreement -- I maybe wasn't clear enough. My suggestion was that just that the arithmetic *vocabulary* as implemented by CWM be written up so that it's easier for a community of use to build up. Any problems would then get identified and could be folded into a subsequent standardization process. I suggested the CWM vocabulary because it exists, and see no point in inventing another one. If there's another candidate, then by all means consider that. Or if there's a real problem with CWM's vocabulary, let's know about it. I think a fair number of people will copy CWM's vocabulary anyway, as has happened with predicates like log:implies and friends. (In general, I wouldn't especially advocate standardizing vocabularies, but arithmetic seems to be sufficiently fundamental... it's difficult to do anything seriously without getting some level of arithmetic involvement.) As for the service aspect, I think that smacks too much of computation, where RDF is about expressing truths. Ultimately, as far as RDF is concerned, property/ies indicates relationship/s between values. Whether those relationships are defined/presented in terms of the result of a well-known service (multiplication), or in terms of a mathematical description of the properties of such a service (e.g. 0*x=0, 1*x=x, x*y=y*x, succ(x)*y=x*y+y, etc) doesn't matter for that purpose, as long as the semantic interpretation applied yields the expected truths and non-truths. When these relationships are used in an inference process, then you may need some "service" to actually perform evaluations. The property URIs simply identify the relations; in performing inference, it's the inference engine's responsibility to apply processes (services) that are consistent with those relations. #g -- At 18:18 16/06/03 +0200, Danny Ayers wrote: > > This suggests to me that we don't really need anything new that isn't > > already being done. There's a lot of activity devoted to the development > > of rules [2] [3]. It seems that there already exists a design for RDF > > properties dealing with relations based on arithmetic operations [1], > > though I could see a case for making that work a little more > > "forceful" -- > > e.g. by publication as a W3C Note, with a view to taking it to the > > recommendation track in future RDF working groups. > >I think cwm is wonderful, but I'm not sure how far it should be considered a >role model. Being able to plugin a procedure at the drop of the hat is >sweet, but unless that procedure is made explicit then everything could get >very shaky. Being able to call on a reproducible algorithm/process/service >to carry out a conversion is one thing, having it be a black box is another. >Personally I'd favour drawing a line in the sand around simple comparisons >(==, > etc) applied to typed literals which could form a little package of >quasi-builtins for use with rules. Where any f beyond this is needed then I >think this would be best viewed as a service, even if it is carried out >locally as Chaals suggested. ------------------- Graham Klyne <GK@NineByNine.org> PGP: 0FAA 69FF C083 000B A2E9 A131 01B9 1C7A DBCA CB5E
Received on Tuesday, 17 June 2003 04:48:26 UTC