- From: Graham Klyne <GK@NineByNine.org>
- Date: Mon, 25 Nov 2002 11:52:56 +0000
- To: pat hayes <phayes@ai.uwf.edu>
- Cc: "Jeremy Carroll" <jjc@hplb.hpl.hp.com>, w3c-rdfcore-wg@w3.org
At 02:18 PM 11/21/02 -0600, pat hayes wrote: >>Also when we went down this canonicalization route we were very aware that >>there are perfectly good RDF implementations that cannot tell when two XML >>literals are identical. > >I don't think that position is now tenable. I feel should hesitate to speak here, but I shall anyway... The more I discuss RDF and its associated concepts, the more it seems to me that this idea of identity is slippery and subtle, and often treated glibly, en passant. I think it's not hard to agree on identity when dealing with a concrete representation (at the level of a sequence of octets), but one when starts to look for deeper forms of equivalence life always seems to get complicated. When do two different octet sequences indicate the same name? When do two different names indicate the same value? Etc. I think the well-ordered world of formal semantics imposes a similar well-ordering on the concept of identity, but doesn't always explain how that ordering is to be achieved. In the case of XML literals, what form of equivalence do we mean? We have several choices: - equivalent octet sequences - equivalent Unicode character sequences - equivalent infosets (whatever that means) - equivalent RDF graphs (OK, not an option for XML literals, but it is another notion of equivalence in XML data) Shallow notions of equivalence are easy to determine, and are sufficient to infer deeper notions of equivalence. Deeper notions of equivalence can be arbitrarily difficult. So what's my point? What is the position that is not tenable? I think we can fairly safely assume that RDF implementations can determine character-level equivalence between XML literals. But it will be problematic for some applications (think: DPH) if the ability to determine deeper levels of equivalence is required. I'm not sure why we *need* any deeper level than character equivalence. I think this falls into the same category as being able to detect entailments: some true facts may be overlooked, but we don't end up concluding falsehoods. #g ------------------- Graham Klyne <GK@NineByNine.org>
Received on Monday, 25 November 2002 09:36:36 UTC