- From: Sergey Melnik <melnik@db.stanford.edu>
- Date: Sun, 17 Jun 2001 15:49:30 -0700
- To: Graham Klyne <Graham.Klyne@Baltimore.com>
- CC: RDFCore WG <w3c-rdfcore-wg@w3.org>
Graham, as I already emphasized earlier, I think anonymous resources shoud be eliminated from the spec, at least from whatever is the RDF model. I guess we have to keep them in the serialization syntax for backward compatibility. In this case, see my suggestion below as just a possible way of making the spec more precise. Another approach is to leave the generation of URIs to the parsers, and deprecate the practice. Seems that we have to wait with this issue until the "core" model is in place. Sergey Graham Klyne wrote: > > Sergey, > > I vaguely remember this being raised, and must say I have deep reservations > about that approach. > > Right now I cannot say exactly why I think this is a problem, and I won't > fight the case here, other than to note that I might object later if/when I > can crystalize my concerns. > > (Part of my objection is that any mechanism that is based on merely > statistical uniqueness (i.e. digest-based) has no place in the core of the > RDF framework. In an appendix to RFC 2938, I have calculated some > probabilities of collision using MD5; the numbers of digest values in use > does not need to be astronomically large (c. 10^18 or so) before the > probability of collision becomes quite significant. Applications may > choose to take such statistical chances, but I don't think a fundamental > technology framework like RDF should do so.) > > #g > -- > > At 06:58 PM 6/15/01 -0700, Sergey Melnik wrote: > > > do you really mean this last claim? > > > I suspect you meant that all parsers should assign a predictable genid > > > given a common RDF/XML description mentioning a resource. 'all parsers > > > should assign the same genid to the same resource' would be magic, since > > > many times parsers won't have that information accessible. > > > >Dan, > > > >just to provide a pointer, a whole while ago I proposed an algorithm for > >doing exacty that "magic". A summary of the proposal can be found at > >http://nestroy.wi-inf.uni-essen.de/rdf/sum_rdf_api/ under "URI > >generation for anonymous resources" (by Peter Hannappel and Reinhold > >Klapsing). I believe, if needed, this algorithm can be tweaked to > >eliminated the problems wrt using XOR (as pointed out by Brian long > >ago). > > ------------------------------------------------------------ > Graham Klyne Baltimore Technologies > Strategic Research Content Security Group > <Graham.Klyne@Baltimore.com> <http://www.mimesweeper.com> > <http://www.baltimore.com> > ------------------------------------------------------------
Received on Sunday, 17 June 2001 18:23:24 UTC