W3C home > Mailing lists > Public > public-rdf-wg@w3.org > February 2013

Re: agenda 27 Feb telecon

From: Peter F. Patel-Schneider <pfpschneider@gmail.com>
Date: Wed, 27 Feb 2013 06:42:30 -0800
Message-ID: <512E1B56.10601@gmail.com>
To: Pat Hayes <phayes@ihmc.us>
CC: Guus Schreiber <guus.schreiber@vu.nl>, RDF WG <public-rdf-wg@w3.org>, Richard Cyganiak <richard@cyganiak.de>

On 02/26/2013 10:53 PM, Pat Hayes wrote:
> On Feb 26, 2013, at 1:30 PM, Peter Patel-Schneider wrote:
>
>> Things to think about wrt the semantics:

[...]


>> 2/ LV is the union of all recognized datatype value spaces.
>>     No visible changes in RDF/S/D, I think, but has visibility in OWL, e.g., an
>>     instance of rdfs:Literal would have to belong to one of the recognized
>>     datatypes.  This would be non-monotonic with respect to adding more
>>     datatypes.
> Agree, and I will change this. As LV plays no role in RDF at all, I do not want it mentioned in the basic interpretation structure (KISS), but suggest introducing it in RDFS and defining it simply as the class extension of rdfs:Literal:
>
> LV=ICEXT(I(rdfs:Literal))
>
> RIght now, this is stated as a condition on RDFS interpretations, but the same equation can be used to define LV, which would not be mentioned earlier in the semantics. That should have no effect on OWL right?
I think that this will work.

[...]

>> Proposed fixes:
>>
>> 1/ Move built-in literals out of simple semantic conditions for ground graphs.  Does anyone use simple semantics?
> I agree, and this change is on my agenda. Simple semantics supports the interpolation lemma which is handy when defining the rule sets.
>
>> 2/ "Fix" datatypes, allowing other datatypes that have language tags (but no RDF syntax would permit them).  This makes rdf:langString a regular datatype. This would allow, for example, internationalized numerics, i.e., "123,1"@fr^^xsd:decimal.  (Well maybe not using xsd:decimal, but you get the idea.)
> I like this, and argued for it, but got strong pushback from Richard. I don't think it is worth re-opening that debate.  (I see no reason why we should not allow literals of arbitrary complexity, perhaps with more than two strings. For example, someone might want to define a datatype describing time intervals which has strings indicating the two end times and the timezone, three strings in all. The semantic model can take this into its stride by allowing L2V to map from any finite sequence of strings to a value. No semantic problem. On the other hand, I think those who write and deploy parsers might balk at that level of freedom in literal syntax. For an argument against forcing people to use a single string, see rdf:PlainLiteral. )
>
>>    Is this too radical for this late in the game?
> Unless Richard has had a change of heart, yes it is.
>
>>   However, who would notice if the change is only made in Semantics, without a corresponding change in Concepts?
> True. But it does seem slightly clunky to introduce this big generalization simply in order to define the identity map. Whatever. (BTW, this was one reason for building this into the simple interpretation case, so we could get the semantics written down without debating whether or not this counts as an L2V mapping.)

The reason to do this would be to make rdf:langString be a non-exceptional 
datatype.  Absent this, I would argue for rdf:langString to be made a datatype 
by fiat, as an exception to the usual datatypes.

>
>> 3/ Make RDF interpretations be {rdf:XMLLiteral, rdf:HTML, rdf:langString, xsd:string}-interpretations (as they are now in Semantics, but I do seem to remember something about rdf:XMLLiteral not being part of RDF any more).
> Whatever the WG decides are datatypes required for RDF, we should put them there. Semantics is a faithful servant of Concepts at this point.
>
> Pat
>
>

peter
Received on Wednesday, 27 February 2013 14:43:00 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 16:25:54 GMT