Re: Using XMLSchema-instance attributes in RDF/XML Syntax (fwd)

Hi Patrick,

I appreciate the reply, even if it isn't what I wanted to hear :-)  If I
understand correctly, you are saying that, for various reasons, it is not
appropriate to ever expect to be able to validate an RDF/XML instance using
XML Schema (except possibly through some non-standard means such as
transforming some hybrid RDF/XML into different forms depending on whether
RDF or XML Schema processing is desired).  This seems unfortunate since my
vision of the two is complimentary instead of competing.

I can appreciate the desire to unify RDF and XML Schema, but I would suggest
that at least for the short term that the unification can be very modest --
something akin to my original suggestion :-) or some variation such as
http://lists.w3.org/Archives/Public/www-rdf-interest/2002Aug/0124.html. 
Then at least it will be possible to create a valid RDF/XML that can be
validated with XML Schema.  If the grand unification of RDF and XML Schema
is closer than I think it is, that would be great and you can skip this
short term step, but I suspect it will be a while before I see an RDF
datatyping spec which unifies the two, much less widely supported,
non-experimental tools that support everything that XML Schema currently
supports, such as regular expressions patterns, unions of datatypes,
extension and restriction of datatypes based on facets, etc.  (I just now
scanned some of the OWL specs, and they look promising, but I am sticking to
my current convictions for the moment :-)

I have also interspersed some additional comments below.

Respectfully submitted,
 Tom Habing

Patrick Stickler wrote:
> 
> Thanks for the pointer Dan. I was about to reply
> directly to Thomas on this.
> 
> I can very much appreciate the utility that would be
> afforded XML Schema users in being able to express
> local datatyping using xsi:type, but there remain several
> questions about its use that have not been clarified:
> 
> 1. Must the datatype identified by xsi:type conform to
> the XML Schema specification? Some preliminary
> research I've done to determine this shows a strong
> perception that this is true. Yet RDF datatyping should
> work equally well for any arbitrary datatype which
> conforms to the minimal characteristics defined by
> RDF, including but not limited to  XML Schema datatypes.

My answer to this is, of course xsi:type should conform to the XML Schema
specification; that's the whole point.  Why can't RDF datatyping and XML
Schema datatyping coexist?

> 2. Will the XML Schema WG/community have a problem
> with RDF adopting xsi:type as a key term in its own
> vocabulary if that  means extending
> or interpreting its semantics to apply to datatypes not
> bound by the XML Schema specs. I suspect not.

Because I want to see RDF and XML Schema peacefully coexists, I disagree. 
Based on the above, the whole purpose of xsi:type is to indicate that some
part of a document should conform to some type defined in an XML Schema. 
Therefore, I believe this is the only way that xsi:type should ever be
used.  RDF should never use xsi:type (except to ignore it) unless it can
also be used in the way expected by XML Schema.

> 
> 3. Is introducing this co-dependency between the two
> standards absolutely necessary? One could easily
> construct a generic tool that incorporates an XML
> Schema validator and which validates typed literals
> in RDF/XML instances without parsing the RDF/XML
> into a graph.

This may just be semantics, but I wouldn't necessarily call what I suggested
creating a co-dependency between RDF and XML Schema.  It is more like
defining a dividing line between RDF and XML Schema -- simply, any attribute
in the XML Schema Instance (xsi) namespace is within the sole purview of XML
Schema, and RDF processors should keep their hands off.  Note, I am
referring only to the XML Schema _Instance_ namespace, not the XML Schema
namespace (xsd).  I like the idea of using things like <xsd:Integer
rdf:value='1234'/> and having that interpreted as valid RDF.

> 
> Given the other aspects of RDF validation, as well as
> those of upper layers such as OWL, I expect that in the
> long term, validation will be performed on the graph,
> not on the XML, and so any utility derived from adopting
> xsi:type would be limited and short lived.

Even so, why can't we have both, if desired, especially for the short term?

> 
> While I'm very sympathetic to easy validation of RDF/XML
> containing XML Schema datatyped literals, I don't think
> it is advisable to employ xsi:type, for the reasons outlined
> above.
> 
> Cheers,
> 
> Patrick
> 
> _____________Original message ____________
> Subject:        Using XMLSchema-instance attributes in RDF/XML Syntax (fwd)
> Sender: ext Dan Brickley <danbri@w3.org>
> Date:           Wed, 21 Aug 2002 09:52:49 +0300
> 
> of interest re datatyping options...
> 
> dan
> 
> ---------- Forwarded message ----------
> Date: Tue, 20 Aug 2002 16:20:31 -0500
> From: Thomas G. Habing <thabing@uiuc.edu>
> To: www-rdf-interest@w3.org
> Subject: Using XMLSchema-instance attributes in RDF/XML Syntax
> Resent-Date: Tue, 20 Aug 2002 17:20:35 -0400 (EDT)
> Resent-From: www-rdf-interest@w3.org
> 
> Hi all,
> 
> I posted the following to www-rdf-comments recently, but it didn't generate
> any comments or followup, so I am posting here to see what happens :-).
> Does what I am proposing make sense, is it too simplistic, or am I just
> missing something?
> 
> ---
> 
> I have been trying to figure out how I can use the various
> XMLSchema-instance attributes (especially xsi:type, but also xsi:nil,
> xsi:schemaLocation, etc.) in an RDF/XML document.  I want to create valid
> RDF/XML, but at the same time I want to be able to validate at least
> portions of the RDF/XML using XML Schema.  Some of my XML Schemas require
> the use of the xsi:type attribute in the instance documents in order to
> validate.  However, RDF insists on treating these xsi:attributes as RDF
> property attributes which causes the RDF to be invalid.
> 
> I can understand this in the original RDF M&S since it predates XML Schema
> by a year or so, but I am surprised to see no mention of this issue in the
> newest "RDF/XML Syntax Specification."
> 
> I have seen some of the discussions in the various lists of using xsi:type
> for data typing in RDF.  I don't claim to understand most of the issues
> associated with this, but I would like to humbly suggest that at the very
> least there should be some language in the "RDF/XML Syntax Specification" to
> the effect that attributes in the XMLSchema-instance namespace should be
> ignored by RDF parsers, similar to what is done with the xml* attributes.
> 
> Thanks,
>         Tom
> --
> Thomas Habing
> Research Programmer, Digital Library Projects
> University of Illinois at Urbana-Champaign
> 155 Grainger Engineering Library Information Center, MC-274
> thabing@uiuc.edu, (217) 244-4425
> http://dli.grainger.uiuc.edu

-- 
Thomas Habing
Research Programmer, Digital Library Projects
University of Illinois at Urbana-Champaign
155 Grainger Engineering Library Information Center, MC-274
thabing@uiuc.edu, (217) 244-4425
http://dli.grainger.uiuc.edu

Received on Wednesday, 21 August 2002 13:50:54 UTC