At 14:31 03/10/2006, Peter Hunsberger wrote:
>On 10/2/06, peter murray-rust <pm286@c...> wrote:
>>At 04:56 03/10/2006, Dave Pawson wrote:
>
><snip/>
>
>>and then I want a linkbase that defines a bounded hyperdocument. I
>>pick up the index document and get everything else
>
>Once more, I'd strongly argue for the formal separation of
>relationship management and link rendering.  Here the semantics here
>seem clear enough; the very fact that you want strong typing indicates
>that it is the relationship between the two types that is of primary
>importance and not the link traversal.

Yes


>As such (for the more general question at hand), if you are compiling
>a list of linking use cases I think you first need to compile a list
>of relationship types and then enumerate those for which link
>presentation is applicable and what types of presentation makes sense
>for each relationship type.  Some example relationships types would
>include:, primary key/foreign key relationships (including id/idref?),
>object inheritance, hyper links and hierarchical inclusion (ie. XML
>parent/child element relationships).

Some of these may be relevant. CML is not based directly on a 
relational model because that is probably not the way chemists think. 
What we use is more like an object model. Inheritance is valuable but 
we do not want to use XML Schema which is far too complicated and 
certainly not how chemists think.



>> >I.e. is there a required/optional set of link properties that 
>> would assist in
>> >your particular usage (presentation I believe is your issue)?
>>
>>No - presentation can be done with normal stylesheets. Semantic
>>annotation is what I care about. What does it mean, not what does it
>>look like to a human.
>
>I assume you are in touch with the conceptual graph and related areas
>of  research?

I have talked over the years with Topic Map people if that comes 
under conceptual graphs. But I would like the relationships all to be 
expressed within the XML documents and use XML syntax (anything 
beyond that is too complicated and components outside the document 
can easily get lost). This is why XLink is appealing.

P.



Peter Murray-Rust
Unilever Centre for Molecular Sciences Informatics
University of Cambridge,
Lensfield Road,  Cambridge CB2 1EW, UK
+44-1223-763069