Re: 2001-09-07#5 Literals

Tim Berners-Lee:
>>
The idea of any "SHOULD" in the definition of equality for literals for
any
language give me the creeps.
>>

Which is to say 'should not' is inviting non-interop. Spot on. I reduced
this constraint to 'should not' for a number of reasons, notably
backward compatibility with text in the M&S, which fudges how lang tags
affect comparison, and of which 'should not' is a reasonable
interpretation, and the fact that a lang tag is optional. But really it
all boils down to this sentence in the M&S:

[[[
the language of a literal is considered by RDF to be a part of the
literal. 
]]]

The problem with that sentence is the word 'part'. Without falling down
a mereological rathole, using 'part' there without some kind of
explanation of what that means isn't helpful. Especially when the lang
tag is mandated to add no information to the consequent graph, and, the
processor is expected to indicate how it deals with lang tags. None of
this helps interoperability.

If we want to harden this to must we have to admit that a literal cannot
exist without a lang tag and that a literal+xml is some kind of natural
type, or at least hold a notion that each 'part' of a literal affects
decisions regarding its equality. 

I like to fight the good fight against that: a lang tag is clearly not
necessary for all literals. A lang tag is not part of a literal anymore
than an RDF statement about Tim Berners Lee is part of him. It's an
interpretation of a literal that it's in English or French;
interpretation implies semantics and that implies it belongs in the
graph. That's to say, lang tags imply semantic equivalences over
literals and not just the syntactic ones made by most string ops as you
rightly point out. The layering is all out of kilter.

The issue is symptomatic of a general problem, Each xml attribute used
to qualify a literal is a hack to hang a property of a literal. 


Tim Berners-Lee:
>>
I am nervous about the effects of langauge tags, I must say.
>>

I'm more nervous about the way xml:lang has crept up from a syntactic
artefact of XML-RDF, to be a candidate for an abstract description of a
literal. And it makes abstract notions of RDF dependent on at least one
other spec, that isn't in my mind central to web architecture. 


Tim Berners-Lee:
>>
Presumably the alternative choices would have been to deprecate the
language tags on RDF literals, in favour of either using XML
with parsetype literal, <d:name rdf:parseType="literal"><span
xml:lang="fr">chat</></>
or RDF itself
>>

Or just allow statements about literals. 


regards,
Bill

..
InterX 
bdehora at interx.com 
+44(0)20-8817-4039 
www.interx.com 

 

Received on Friday, 19 October 2001 06:34:36 UTC