Thanks for that.

My concerns still rest with the capabilities and expectations of the
client:

Can a client deal with an arbitrary vocabulary "on the fly"? e.g.. in
your example, would the programmer of the client have to be aware of the
fruit vocabulary in order to deal with the data coming back? If the
latter the client would need some way of indicating to the ZNG server
what vocabularies it knows how to deal with.

In your example the vocabulary is simple enough that there is an obvious
(and unique) RDF encoding of the data. For more complex vocabularies
there may be syntactically different RDF expressions of the same
information - would a client be able to deal with these variations or if
not how would it specify which variation it could deal with?

In general ZNG is aimed at lowering the complexity/barriers. A client
that can adapt to handle arbitrary vocabularies and variant RDF
expressions within such a vocabulary strikes me as being a very complex
programming task (in AI techniques etc.) as opposed to a client which
indicates its programmed limits to the server (who in turn indicates its
programmed limits to the client and the session aborts if there is no
overlap) - which is the current case with Z39.50 and current drafts of
ZNG.

Hence my feeling that any record schema should be uniquely "named" and
that the client should request this schema and the server either return
in that schema or fail the request. The schema "named" could be a
vocabulary and RDF structure for the vocabulary, but may not be.

In the WSDL schema for RDF, I have left the recordSchema in the request
optional but that is on the assumption that if omitted a particular
schema (which may be a particular vocabulary and RDF structure) if
understood.

Matthew



> -----Original Message-----
> From: Poul Henrik Jørgensen [mailto:[log in to unmask]]
> Sent: 26 September 2001 15:30
> To: [log in to unmask]
> Subject: Re: ZNG, RDF Syntax and RDF Schema
> 
> Hi Matthew.
> 
> You questions below induced me to think the matter through once more.
> Below is an attempt to illustrate my proposal with examples.
> Your comments are very welcome - and useful.
> 
> 1. Metadata Vocabulary
> ======================
> Assume, that farmer Joe has declared the following vocabulary
(Namespace)
> -
> by means of RDF Schema or otherwise:
> 
> Object Class: CID=Fruit. Superclass=Entity
> Object Class: CID=Apple. Superclass=Fruit
> Object Class: CID=Pear, Superclass=Fruit
> 
> Property: PID=Taste. Domain=Fruit. Range=("Sour", "Sweet")
> Property: PID=Colour. Domain=Fruit. Range=("Red", "Green")
> 
> 2. Metadata Instance
> ====================
> Being a forward-looking businessman, the Joe has established a ZNG
Server
> with information about his products.
> In reply to a query about his products, he informs that apples are Red
and
> Sweet, while Pears are Green and Sour.
> Joe has not bothered to declare a special format for his information.
So
> he
> just defaults to the RDF Serialization Syntax:
> 
> <?xml version="1.0" encoding="UTF-8"?>
> <rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
> xmlns:rdfs="http://www.w3.org/2000/01/rdf-schema#"
> xmlns:gs="http://FarmerJoe/Meta/GreenStuff/">
>         <rdf:Description rdf:about="http://FarmerJoe/Products/Apple/">
>                 <gs:Taste>Sweet</gs:Taste>
>                 <gs:Colour>Red</gs:Colour>
>         </rdf:Description>
>         <rdf:Description rdf:about="http://FarmerJoe/Products/Pear/">
>                 <gs:Taste>Sour</gs:Taste>
>                 <gs:Colour>Green</gs:Colour>
>         </rdf:Description>
> </rdf:RDF>
> 
> 3. Something Else
> =================
> Do you think, that this would be practical, i.e. to just reference the
RDF
> Serialization Syntax and the Vocabulary Namespace?
> 
> - Or do you have a better solution to the problem of how to format
> metadata
> without a predefined format?
> 
> Best regards,
> Poul Henrik
> 
> 
> -----Original Message-----
> From: Matthew Dovey [mailto:[log in to unmask]]
> Sent: 26. september 2001 13:08
> To: [log in to unmask]
> Subject: Re: ZNG, RDF Syntax and RDF Schema
> 
> 
> Poul,
> 
> So by "default" format are you suggesting that
> 
> a) ZNG recommends that anyone with a vocabulary (such as Ray at
present)
> should use RDF to specify an XML instantiation of that vocabulary.
> Providing that the structure of this XML instantiation can be given a
> unique identifier (a URL), we can use this as the RecordSchema
> designator in ZNG
> 
> or
> 
> b) the RecordSchema in the ZNG Request or Response could just
reference
> a vocabulary - in which case the XML instantiation is derived from the
> vocabulary in RDF. This assumes that given a vocabulary it is possible
> to derive a single or canonical RDF structure (i.e. how do we cope
with
> two servers having very different RDF expressions of the same
> vocabulary?)
> 
> or
> 
> c) something else?
> 
> Matthew