Comments on: Global identifier schemes don’t scale II http://phildawes.net/blog/2006/03/09/global-identifier-schemes-dont-scale-ii/ Mostly programming with a few bits of other stuff Thu, 20 Nov 2008 23:02:27 +0000 http://wordpress.org/?v=2.1.1 By: Semergence » Phil Dawes’ Stuff » Blog Archive » Global identifier schemes don’t scale II http://phildawes.net/blog/2006/03/09/global-identifier-schemes-dont-scale-ii/#comment-6328 Semergence » Phil Dawes’ Stuff » Blog Archive » Global identifier schemes don’t scale II Thu, 09 Mar 2006 08:37:03 +0000 http://phildawes.net/blog/2006/03/09/global-identifier-schemes-dont-scale-ii/#comment-6328 [...] Phil continues his thoughts with the assertion that [...] […] Phil continues his thoughts with the assertion that […]

]]> By: Stefan Plantikow http://phildawes.net/blog/2006/03/09/global-identifier-schemes-dont-scale-ii/#comment-6334 Stefan Plantikow Thu, 09 Mar 2006 12:20:06 +0000 http://phildawes.net/blog/2006/03/09/global-identifier-schemes-dont-scale-ii/#comment-6334 I can see your point and think you're right about the problems of unique identifiers, as identity is a matter of meaning, and meaning a matter of context (category theory, anyone?), for a global semantic web this raises the question how to describe context. Using unique ids to describe the shared context obviously leads to a vicious circle here, so some other means are necessary. I can see your point and think you’re right about the problems of unique identifiers, as identity is a matter of meaning, and meaning a matter of context (category theory, anyone?), for a global semantic web this raises the question how to describe context. Using unique ids to describe the shared context obviously leads to a vicious circle here, so some other means are necessary.

]]> By: Danny http://phildawes.net/blog/2006/03/09/global-identifier-schemes-dont-scale-ii/#comment-6336 Danny Thu, 09 Mar 2006 14:37:17 +0000 http://phildawes.net/blog/2006/03/09/global-identifier-schemes-dont-scale-ii/#comment-6336 I think you're absolutely right to consider URIs as fair game for re-evaluation, though I personally think they're the best thing since sliced bread. I don't know to what extent using quads globally might have helped, but I've a feeling it would be closer to shunted the problem along, rather than solving it. The fact that URIs have scaled on the current web suggests that there is something right here. But when it comes to the SemWeb, you can *almost* already make the kind of no-URI descriptions you describe: id::PhilDawes (weight 10st, date 24/12/2005) id::PhilDawes (weight 10st 3lbs, date 26/12/2005) consider id as a bnode, with associated properties maybe localId "PhilDawes", weight "10st", date "24/12/2005" etc. Danbri's <a href="http://rdfweb.org/mt/foaflog/archives/2003/07/10/12.05.33/" rel="nofollow">Identifying Things in FOAF</a> post puts the case well for the indirected id idea. However, there is quite a big *almost*. How do you identify the relationships between things? The value of "weight" property above is expressed as a string containing the units - that isn't likely to be a very common use of the term "weight". Although disambiguation of the same multi-reference form as the thing-identification could be done, if I remember correctly this led to one of the issues with the early Semantic Networks. You can build lovely big interconnected graphs, but then if you try to do anything with them (query, inference, transformation) you've got a problem. The use of global IDs for the relationships helps make these problems tractable. I think you’re absolutely right to consider URIs as fair game for re-evaluation, though I personally think they’re the best thing since sliced bread. I don’t know to what extent using quads globally might have helped, but I’ve a feeling it would be closer to shunted the problem along, rather than solving it. The fact that URIs have scaled on the current web suggests that there is something right here.

But when it comes to the SemWeb, you can *almost* already make the kind of no-URI descriptions you describe:

id::PhilDawes (weight 10st, date 24/12/2005)
id::PhilDawes (weight 10st 3lbs, date 26/12/2005)

consider id as a bnode, with associated properties maybe localId “PhilDawes”, weight “10st”, date “24/12/2005″ etc.

Danbri’s Identifying Things in FOAF post puts the case well for the indirected id idea.

However, there is quite a big *almost*. How do you identify the relationships between things? The value of “weight” property above is expressed as a string containing the units - that isn’t likely to be a very common use of the term “weight”. Although disambiguation of the same multi-reference form as the thing-identification could be done, if I remember correctly this led to one of the issues with the early Semantic Networks. You can build lovely big interconnected graphs, but then if you try to do anything with them (query, inference, transformation) you’ve got a problem. The use of global IDs for the relationships helps make these problems tractable.

]]>