W3C home > Mailing lists > Public > public-esw-thes@w3.org > November 2004

Re: SKOS dodges the identity crisis?

From: Graham Klyne <GK@ninebynine.org>
Date: Fri, 19 Nov 2004 09:21:50 +0000
Message-Id: <5.1.0.14.2.20041119090232.032c4fc0@127.0.0.1>
To: "Miles, AJ (Alistair)" <A.J.Miles@rl.ac.uk>, "'www-rdf-interest@w3.org'" <www-rdf-interest@w3.org>, "'public-esw-thes@w3.org'" <public-esw-thes@w3.org>

At 18:46 18/11/04 +0000, Miles, AJ (Alistair) wrote:

>Hi all,
>
>I did some more thinking about SKOS, and wrote up an idea at:
>
>http://esw.w3.org/topic/SkosDev/DodgeIdentity
>
>I would very much like to know if you think this looks sound, workable,
>reasonable, viable, or not ... all thoughts welcomed :)
>
>We really have to get this sorted.

FWIW, I don't see anything broken in your approach.  I think a number of 
approaches are possible, and whichever one you choose I think what's most 
important is to use it consistently.  So if you choose that your URIs 
indicate concept descriptions, and build your vocabularies and assertions 
consistently on that basis, I think you'll be OK.  I also don't think that 
any putative advances under discussion are likely to cause future breakage.

Wandering off into more philosophical aspects of your comments:
[[
However, this would have some undesirable effects. For a start it means all 
of the labels will get mixed up, and we won't be able to tell which labels 
came from which thesaurus. And if there were definitions or scope-notes, 
these would get mixed up too.

For very practical reasons, it makes good sense to keep URIA and URIB as 
distinct nodes within an RDF graph.

 From a philosophical point of view, some person wrote the description of 
URIA, and a different person wrote the description of URIB. If URIA denotes 
a concept in someone's head, and URIB denotes a concept in a different 
person's head, well how can we ever know they were thinking exactly the 
same thing?
]]

In general, I'd say that you cannot know they were thinking the same thing.

If the two "resources" are to be labelled with different strings, I would 
say that makes them ipso facto different resources - they must have 
differing interpretations in any model.  They may be substantially similar, 
but they remain different.

I think the AI folks have been wrestling with this for a while:  when are 
things to be considered equivalent and when are they different?  When is a 
"small" difference significant?  (e.g. when are 22/7 and 3.14159 considered 
the same and when are they different -- as numbers they're certainly 
different, but as approximations for pi in school level maths they might be 
regarded as equivalent.)  I think this is all a manifestation of the "frame 
problem".

So, what can we say:
(a) two occurrences of the same URI denote the same thing
(b) different URIs denote different things unless there's a clear assertion 
or inference that they are the same thing
(c) if there is any observable difference in expected/required behaviour, 
they definitely denote different things

Beyond this, I think one must generally assume that different URIs denote 
different things.

...

In thinking about this, I was wondering if we are to allow "level breaking" 
properties in RDF that allow examination of the URI used to label a 
resource.  I could imagine something like this as a special property in a 
CWM rule:

    { ?a foo:uriMatches "(some-URI-regexp)" }
=>
    { ... }

Does this mean that two different URIs necessarily denote different things, 
because the URIs themselves lead to observable differences?  I'm rather 
uncomfortable with this, but I'm not sure if there's a principled way to 
assert that this is not a valid property.

#g


------------
Graham Klyne
For email:
http://www.ninebynine.org/#Contact
Received on Friday, 19 November 2004 11:20:01 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:38:52 GMT