- From: Sergey Melnik <melnik@DB.Stanford.EDU>
- Date: Wed, 17 Nov 1999 19:19:29 -0800
- To: RDF Interest Group <www-rdf-interest@w3.org>
I've just discussed these two points with Stefan and would like to jot in down on the list: (1) as I pointed out in one of the previous emails, reified statements should have unique IDs. Thus, we need a standard Skolem function that returns an ID for a given subject, predicate and object. In my current implementations I'm using the following fct: reifiedID(subj, pred, obj) = crc64(subj) XOR crc64(pred) XOR crc64(obj) which delivers a 64-bit integer, where crc64() is based on some standard polynomial. However, we want to have a URI so we'd better generate a UUID like uuid:069d9fb6-943e-11cd-a35c-0000c08adf56 according to some similar (e.g. MD5 based, [1], ...) algorithm. (2) Another point are the so-called "noname" resources. It is convenient to have them as of the current serialization syntax, but the problem is that two different parsers will generate different URIs for the same noname nodes even if they fetch the same document from the same location. Generating these URIs should also be done algorithmically. However, we didn't think much about the algorithm. Any ideas? Both algorithms must be formally specified and should be an integral part of the new specs. Sergey [1] UUID: http://www.ics.uci.edu/~ejw/authoring/uuid-guid/draft-leach-uuids-guids-01.txt (has anyone got a link to newer specs?)
Received on Wednesday, 17 November 1999 22:14:49 UTC