Re: what functions should we include in SPARQL 1.1?

On 10 Nov 2010, at 17:05, Andy Seaborne wrote:

> 
> 
> On 09/11/10 23:24, Axel Polleres wrote:
> >
> > On 10 Nov 2010, at 04:02, Andy Seaborne wrote:
> >
> >>
> >>
> >> On 09/11/10 17:42, Axel Polleres wrote:
> >>>>>     which we may want to reuse, or should we uniformly refer to the
> >>>>>
> >>>>>        http://www.w3.org/2009/sparql-functions#
> >>>>>
> >>>>>     for all functions (also the xs: ones?
> >>>>
> >>>> Reuse where possible.
> >>>
> >>> Remembering that we had this discussion partly before, one argument against reuse and for uniformly using a single
> >>> namespace was that this single namespace would indicate exactly the interchangeable functions within the SPARQL1.1 spec
> >>> and that these functions then would be usable without a namespace prefix as simple keywords of the language.
> >>
> >> Must have missed that - I don't recall a proposal to make the names
> >> keywords in the language for all mandatory functions from F&O - is the
> >> proposal to add new keywords for all required functions, then map to
> >> IRIs (sparqlfn:) and owl:sameAs to fn:.
> >
> > that'd work for me.
> 
> It was a question - is that the proposal you are making?

Yes. 

1) owl:sameAs is one possiblity. 
We could describe such things in the sparqlfn: namespace document, right?

2) Alternatively/additionally we could also redirect that URI to the fn: one.
To be discussed.

3) this one is more speculative... could we also do something similar to RDFa's namespace profiles?
   (this is more a question than a proposal so far, because I don't know/understand that mechanism well enough yet)



>  It's not
> support - when there is a detailed proposal on the table I'll decide.
> 
> I don't recall it coming up before and I don't understand the argument
> for adding more keywords to the language (it's a grammar change albeit
> with no issues arising). An open set of keywords based on prefixless
> usage is a problem for the grammar, e.g. function called "select".

It's simply weird for users IMO, if they have to use prefixes for some 
of the default supported functions, but not for others.

best,
Axel


> Using the same URIs if it means the same thing makes more sense to me
> but the net effect of your proposal is that I-as-implementer will have
> to add IRIs both ways round.
> 
> Since sparqlfn: will not be needed (if there are SPARQL-unique
> functions, they have keywords so you don't need IRIs anyway), the
> argument for one namespace does not work for me.  If an implementation
> adds an extra fn: not in the required set, then doe sit also appear in
> sparqlfn: - presumably not but then the app writer sees fn: but not
> sparqlfn:.
> 
> It will be a bit more diffcult to work with service descriptions if
> inference is needed.  Leigh Dodds' survey uses use of fn: and of
> per-system naming.  The expecations of all fn: have not arisen in my
> experience in practice - documentation helps:
> http://openjena.org/ARQ/library-function.html
> 
>         Andy
> 
> >
> > Axel
> >
> >>
> >>          Andy
> >>
> >> PS
> >> http://spreadsheets.google.com/pub?key=0AkNZYESXv3IndGwyRkRXZ2hES0RjM0c3MHhLa05vTmc&gid=0
> >>
> >
> >
>

Received on Wednesday, 10 November 2010 17:14:45 UTC