W3C home > Mailing lists > Public > public-rdf-dawg@w3.org > October to December 2004

Re: SPARQL Variable Binding Results XML Format draft - for review

From: Dave Beckett <dave.beckett@bristol.ac.uk>
Date: Wed, 15 Dec 2004 11:53:40 +0000
To: andy.seaborne@hp.com
Cc: public-rdf-dawg@w3.org
Message-ID: <20041215115340.19d7870b.dave.beckett@bristol.ac.uk>

On Mon, 13 Dec 2004 15:31:14 +0000, "Seaborne, Andy" <andy.seaborne@hp.com> wrote:

> Publishing this draft is acceptable to me.
> 
> 	Andy
> 
> Comments:
 
> 1/ Unbound variables should not be mentioned in <result> to avoid
> the issue with
> 
> """
> ISSUE: This is a problem, you cannot distinguish a bound variable
> value with an empty string literal, from a variable with no binding.
> """
> because we now have all the variables declared in the header.

OK, changed so that only bound variables in one solution appear
inside a <result>.  I'm unsure if this is better than
<foo unbound="true"/> since it changes the contents of the child
elements of <result> potentially and requires lookups by
variable name (xml element name) rather than by order in the
<result>

In particular, it works badly with XSLT1 at least in a simple fashion
unless I work out how to deal with multiple parts of a document in
XSLT1, which I can put in the example XSLT.

> 2/ """Note: The blank node label I may not be the same string used
> as the blank node label in the query graph."""
> 
> A serializer might not actually know the blank node label in the
> query graph so can't guarantee this feature.  It may just know
> whether two bNodes are the same or different.
> 
> I think this should say that the blank node label I is scoped to
> the result set   XML document and does not make any association
> with graph bNode label.

That's better wording.  Changed.

> 
> 3/ Observation: I thought you were proposing using rdf:datatype,
> rdf:nodeID.  I see "datatype" and "id".  That's fine - it drops the
> need for rdf: namespace.

Yes.  But I think 'id' might be confused with html:id or xml:id so
I'm changing this to bnodeid.


> 4/ """ISSUE: XML Style - dawg-results needs a better name"""
> 
> It's <dawg-result> in the example above this line, not "dawg-results"

I've changed to a root element <sparql>

> 5/ Could the example be extended to have the same bNode in one
> place as well as a different bNode in another place?  Currently,
> there is just two different bNodes which does not illustrate the
> use of the same id=""  

OK, I've extended the example to ask for a foaf:knows, so there is
they are each friends of each other.

> ... As HTML/XML IDs are different everywhere,
> the name "id" might be confusing.

Changed to bnodeid.  [Or should this be bnodeID like RDF/XML ?]

> I don't speak fluent RelexNG but the following seem odd:
> 
> A/ Use of #decimal
> 
> datatypes xsd =" http://www.w3.org/2001/XMLSchema#decimal"
> 
> Why #decimal? Is this (xsd) just for xsd:anyURI later?

Totally wrong URI, fixed.  I'm now using xsd: for giving XSD types for
  bnodeid='..' XML names (xsd:NMTOKEN)
  uri='...', datatype='...' URI references (xsd:anyURI)

> B/ nameAttr
> 
> var = element * {
>       xmllang?, nameAttr?, hrefAttr?, datatypeAttr?, idAttr?, literal
> }
> 
> nameAttr = attribute name { string }
> 
> I don't see where the attribute "name" fits in.  Is this remains
> from the <var> form?

That's right it's junk and been deleted.

Dave
Received on Wednesday, 15 December 2004 11:57:04 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 16:15:21 GMT