Re: on "ORDER with IRIs" comment and user-selectable collation units

Dan Connolly wrote:
> Hoehrmann writes, 'section 10.1
> notes "IRIs are ordered by comparing the character strings making up
> each IRI" it's however not clear how character strings are compared'
>  --
> I would have thought it's obvious that we mean lexical order
> by unicode code-point. I thought there must be
> some anchor in [CHARMOD] for that, but while I was looking, 
> I found...

I belive the problem stems, at least in part, from the fact that rq23 does not 
define "<"  on xsd;strings.  It defines it on (numeric,numeric) and 
(dateTime,dateTime).  If rq23 extended "<" to include (xsd:string,xsd:string) 
then the text above could refer to that otherwise it will have to explicitly 
state that the character strings are compared by codepoint sequence.

This also applies for sorting by a variable that is always an xsd:string, so I'd 
like to see "<" defined by (xsd:string,xsd:string).

> C066  [S]  [I]  Software that allows users to sort or search text SHOULD
> allow the user to select alternative rules for collation units and
> ordering.
>  --

Do we allow explicit calls to fn:compare()?
which does allow various collations.

One comment - xpath-functions seems to allow the default collation to be set 
externally to the query.  For interoperability between client and server, can we 
state that the default default collation (that used if there is no other 
information) is by codepoint?


> There was some discussion in IRC of whether the SPARQL QL spec has
> user-selectable collation units, but I didn't understand the conculsion,
> if there was one.
> So, firstly, question of fact: where's the relevant text in rq23?
> Does it specify user-selectable collation units?
>  If so, let's have a test. Is there one? Has anybody passed it?
>  If not, does it say why not? [CHARMOD] C066 says it should,
>  so we owe and explanation as to why not.

Received on Monday, 8 August 2005 17:36:44 UTC