Re: RDF 1.1 Semantics: LV and typo

Thanks for the quick response! Comments in line.

On 24/09/2014 19:58, Patel-Schneider, Peter wrote:
> Hi:
>
> This is an unofficial reply.
>
> LV is mostly around in the 1.1 semantics to provide a link back to the LV in the older semantics.  It could be removed without any trouble.
>
> However, LV is not defined circularly.  LV is defined in terms of ICEXT, I, and rdfs:Literal.
>
> ICEXT is defined just above, and that definition uses IEXT, I and rdf:type.
>
> IEXT is part of every interpretation.
>
> I is defined in the semantic conditions for ground graphs.  For IRIs, I is defined in terms of IS, which is part of every interpretation.
>
> So, no circularity.

To clarify, the unknown quantity on the right hand side for me is 
rdfs:Literal. The class extension of rdfs:Literal has not been 
previously defined / mentioned up to that point. Hence it seems to me 
that LV (which is not defined up to that point) and the extension of 
rdfs:Literal (which is not defined up to that point) are both unknown 
quantities. (My expectation was that rdfs:Literal would be defined in 
the semantic conditions in a similar way to, say, rdf:Property.)

> Defining LV as you suggest would have some interesting consequences.   You would be able to reason that a literal value (an arbitrary element of LV) is one of the literals that you can enter.  I don't think that you can conclude anything from this in RDF, but in OWL you can.

Understood. I see now that in D you could have a datatype without 
concrete lexical strings for some values (e.g., owl:real?) and with this 
definition, those values would not be a rdfs:Literal, thus causing 
problems saying that, e.g., every datatype is a subclass of literal. 
(The key realisation for me here is that L2V is not necessarily 
"surjective".)

In this case, I guess like the notion of rdfs:Class, the notion of 
rdfs:Literal is "new" in a sense and cannot be (easily) "derived".

> You should think of IC and IV simply as shorthands.  They can be replaced everywhere in the current semantics by their definitions, ICEXT(I(rdfs:Class)) and ICEXT(I(rdfs:Literal), respectively.

Understood.

>
> peter
>
>
> PS:  If you want the tyop to be noticed officially you should probably send in a separate message to that effect.

Will do.

Thanks,
Aidan

> On Sep 24, 2014, at 3:20 PM, Aidan Hogan <aidhog@gmail.com>
>   wrote:
>
>> Hi folks,
>>
>> Regarding the RDF 1.1 Semantics document.
>>
>>
>> What is the purpose of LV? (I know there's a note in 9.1, but it's in a separate non-normative section.)
>>
>> The definition:
>>
>> 	LV is defined to be ICEXT(I(rdfs:Literal))
>>
>> ... seems to me to be circular -- in that it defines one previously unknown quantity in terms of another previously unknown quantity -- and perhaps it's unnecessary?
>>
>> Would LV be equivalent to something like:
>>
>> 	{ v : there exists l s.t. IL(l) = v }
>>
>> ... in which case, could ICEXT(I(rdfs:Literal)) not be equated directly with the above set as a true semantic condition, dropping LV altogether?
>>
>> 	ICEXT(I(rdfs:Literal)) = { v : there exists l s.t. IL(l) = v }
>>
>> (If this equality doesn't hold, I'd be interested to know why. If it does hold, it would seem a better option than introducing LV.)
>>
>>
>> I have a similar issue with the definition of IC:
>>
>> 	IC is defined to be ICEXT(I(rdfs:Class))
>>
>> ... it does not appear to be a semantic condition but rather an extension of the notion of an interpretation with the abstract idea of a class. Not sure what I would suggest here other than to formally extend the notion of an interpretation outside of the semantic conditions? Probably this could have been built-in earlier since, e.g., RDF also has some implicit notion of classes (the class rdf:Property).
>>
>> (Apologies if this has been discussed before.)
>>
>>
>> Minor Typo: Alex Polleres -> Axel Polleres
>>
>> Best,
>> Aidan
>>
>

Received on Wednesday, 24 September 2014 23:49:08 UTC