Re: my vote on the varieties of datatyped tagged literals

On Thu, 2011-09-08 at 17:40 +0200, Richard Cyganiak wrote:
> On 8 Sep 2011, at 16:42, Sandro Hawke wrote:
> >> On 7 Sep 2011, at 16:40, Sandro Hawke wrote:
> >>> Option 2 might be worse than Option 1; to put it simply, it seems to be
> >>> making tagged literals be datatyped literals by making up a new,
> >>> different, *non-XML-standard* sort of datatyped literal.
> >> 
> >> This is not true for 2c.
> >> 
> >> It is compatible with the XSD notion of datatypes. That was the whole point.
> > 
> > I'm giving you my skeptical look.  |-)  A datatype with an empty lexical
> > space?   I see some discussion in XSD 1.1 about "ineffable" values,
> > points in value spaces which have no lexical representation, but it
> > seems like quite a stretch to define an XML Schema Datatype with a full,
> > useful value space and an empty lexical space.
> 
> You claimed that it is non-standard. In what way is it non-standard?

Is "lexical form" and "lexical representation" the same thing?  Only the
latter is define in the spec; the seem to use the terms interchangeably.
I'll assume so.

Well, I'm confused by the same issue Pat asked you about, where there
can be lexical forms but an empty lexical space.   I might just not
understand what's going on, but it seems to me that datatypes are
supposed to be fairly straightforward mapping and this proposal does
something very odd with them.   I trust it fits with some reading of the
letter of the  law, but I'd be very surprised if the people in the XML
Schema Working Group would consider this okay.   They might or might not
notice or care.

So, that's what I mean by non-standard.  I mean, this doesn't look like
an XML datatype that people familiar with XML datatypes would consider a
proper XML datatype.

> > In terms of code: with this design, we need a different API for language
> > tagged strings than for other data values, right?   We can't use .lexrep
> > to get the lexical representation, since there is none.  
> 
> I don't know what you're talking about. I'll quote the design again:
> 
> [[
> The abstract syntax has a lexical form and language tag (like in RDF 2004). The value is assigned directly (like in RDF 2004), bypassing the datatype. The datatype has an empty lexical space and empty L2V mapping.
> ]]
> 
> Why does this require a different API for language-tagged strings?

I just can't get my head around the idea that a value like <"chat",
"fr"> would have a lexical representation ("chat") but that it wouldn't
be in the lexical space.

Also, it seems like a problem that the pair of lexical representation
and datatype does not convey all the information.

I'm totally willing to accept that there might be some elegant solution
in here; I just can't see it.

     -- Sandro

> > For my example
> > code snippets, your option 2c (which Ivan has labeled 2d) still looks
> > just like option 1, I think.
> 
> Yes – it doesn't require any string munging, which is an advantage.
> 
> Best,
> Richard
> 
> 
> 
> > 
> >   - Sandro
> > 
> >> Best,
> >> Richard
> >> 
> > 
> > 
> > 
> 
>

Received on Thursday, 8 September 2011 20:46:38 UTC