Re: WebID-ISSUE-61 (xsd): xsd datatypes [ontologies]

On 20 Nov 2011, at 12:49, Dominik Tomaszuk wrote:

> On 20.11.2011 10:54, Henry Story wrote:
>> 
>> In your example below you are putting syntactic restrictions on how the
>> datatype should be written out. But that does not give us the semantic
>> mapping we need. So unless the only way to write out the hex strings is
>> with 2 characters followed by a '-' with no initial and final white
>> space, and no way of entering new-lines, then you won't have a canonical
>> representation. If you don't have a canonical representation then you
>> have to specify a map to a canonical one. cert:hex does just that: it
>> says remove any non hex characters. That is what gives it the
>> flexibility of being useable in the wild with people cutting and pasting
>> information from their browser. It is also what allows us to make easy
>> to explain videos of how the key in the browser relates to key in the page.
>> 
>> If you don't think that is needed then one might as well go with well
>> known datatypes. xsd:hexBinary is already defined and it has the
>> advantage of even having another xsd:base64Binary way of writing things
>> out. xsd:integer has the advantage of being mathematically the simplest.
>> 
>> It's just a pitty they did not make more flexible hex datatypes when it
>> would have been so easy to do that.
> OK, but my point of view is machines like. OK you can write:
> rsa:modulus "SSBsaWtlIFdlYklELg=="^^base64Binary
> But is it valid modulus? No, it isn't.
> You can write:
> rsa:modulus "12:34..."^^xsd:string
> or mayby rsa:modulus "12-34..."^^xsd:string
> How to parse it? It is not possible to consider all cases.
> Some people vote for xsd:string (without restrictions). It has some benefits but also allows to write:
> rsa:modulus "I like WebID"^^xsd:string
> Moreover why xsd:string, and why not RDF literal, or PlainLiteral [1]?
> Semantic view is important but don't forget about abstract syntax.

Well that's why it's probably be better to go with xsd:hexBinary, because those problems have already been considered there, and it's the kind of thing people expect. Anyway as you see there is no ideal solution.

My guess is that xsd:hexBinary will probably kill this discussion, make SPARQL queries a simple ASK query, and so make the implementations a lot more efficient for very large providers, and a lot simpler for startup implementors.

Henry


> 
> [1] http://www.w3.org/TR/rdf-plain-literal/
> 
> Domel

Social Web Architect
http://bblfish.net/

Received on Sunday, 20 November 2011 12:25:34 UTC