[topicmapmail] external resources reifying

[Jan Algermissen]

> Olivier Lamborelle wrote:
> 
> Jan,
> 
> Thanks again... I think our "divergences" are getting narrow... :-)
> 
> I got it about subject identity, it seems so obvious to me now that the address of the external resource should be the subject identity of the topic
> that reifies this resource, as you wrote it the first time:
> 
>      <topic id="a-document">
>          <subjectIdentity>
>          <resourceRef xlink:href="http://foo/doc1.xml" />
>          </subjectIdentity>
>      ....
> 
> So I actually agree with the whole description of this topic you gave, with all the metadata in it. This corresponds to my original intention
> (except I wrongly put the link as a first occurrence but now it's much better).
> 
> Now, I still have a problem with the description of a topic that would have an occurence refering to the external document reified above. Your last
> suggestion was:
> 
>      <topic id="valve16">
>        <occurrence>
>          <instanceOf>
>            <topicRef xlink:href="#functional-description">
>          </instanceOf>
>          <scope>
>            <topicRef xlink:href="#expert-technician" />
>          </scope>
>          <resourceRef xlink:href="http://foo/doc1.xml" />
>        </ocurrence>
>        <occurrence>
>          <instanceOf>
>            <topicRef xlink:href="#functional-description">
>          </instanceOf>
>          <scope>
>            <topicRef xlink:href="#spacecraft-captain" />
>          </scope>
>          <resourceRef xlink:href="http://foo/doc2.xml" />
>        </ocurrence>
>      </topic>
> 
> 
> Two things still botters me:
> 
> The first one is the most practical: I would like the occurence to refer to the topic defined above and not directly to the external document. This
> would be coherent with the fact that the external resource is now reified, so why not point to it.

Because that is not how the systax is intended to be used. An XTM processor will do the right
thing (merge the surrogates for the resource based on the URI) and you should simply use XTM
as it is intended to. After all, it is just one way to serialize the abstract graph that is
your topic map.

 This has the advantage to really concentrate all
> information about the resource in 1 topic: address and metadata.

I think your understanding is too document centric. XTM is just on possible serialization
syntax for topic maps and all that matters is that the right links are in place that
allow you to reconstruct the graph from the syntax. So using <topicRef> in <occurrence>
would (set aside that it is invalid anyway) not buy you anything, because what matters is
that the link is there and <resourceRef> sufficiently (and intentionally) does this.

> 
> However, as you recalled me, occurrence can not contain topicRef (except as scope or instanceOf)... That's a problem for me...  I guess putting a
> topic ID in a ressourceRef would not stick to the XTM "spirit".. but maybe that would work..
> 
>      <topic id="valve16">
>        <occurrence>
>          ...
>          <resourceRef xlink:href="#a-document" />
>        </ocurrence>
> 
> (Maybe the occurrence type could indicate that the resourceRef will point to a topic ?)


I really don't see why this matters at all for you, why not use the syntax as intended? What
are the problems you have?


Or do you write your own engine, afer all?


> 
> The second one concerns the discussion about the distinction between attributes that are really attached to the resource itself, and the information
> about the occurence, that describe the nature of the relationship. I was trying to imagine that the resource could constrain or contain the natures
> of the various relationships that can point to them. I imagined that if you see a piece of text or a picture, you can probably already determine the
> nature of the various relationships. 

Oops...how would that work?

A picture of a person will never be used in a  "location" type occurrence.  From this point of view, these kind
> of information could be seen as attributes of the resource itself. 

No, because the subject that the resource is about (the person in this case) has nothing
to do with the properties of the resource as such (author,last update,size,mime-type etc).


However, I think I was going to far and was a bit theoritical.. so I'll finally
> stick to the "usual" concepts and follow your "advice":

Ehm...yes, I think that would be easier ;-)

> 
> > No, metadata is a description of the information resource (attributes). And the attributes do not
> > change (they are properties of the resource regardless of context).
> >
> > The scope applies to the relationship between the valve and the associated document (the
> > occurrence is the relationship, not the information resource). 'functional-description'
> > describes the nature of the relationship NOT the information resource!
> >
> > Be very careful to distinguish between the information resource and the occurrence (which (again)
> > is a relatioship).
> >
> 
> Thanks again Jan... This is helpful in the process of building my first topic map (I guess you noticed... ;-)  )

Be patient, it takes some time and practice. Feel free to ask more.


Jan


> 
> Olivier.
> 
> --
> O. Lamborelle
> System & Operations Engineer
> 
> Space Applications Services NV
> Leuvensesteenweg 325
> B-1932 Zaventem
> BELGIUM
> 
> Tel: +32 2 721 15 81
> Fax: +32 2 721 54 44
> URL: http://www.spaceapplications.com

-- 
Jan Algermissen                           http://www.topicmapping.com
Consultant & Programmer	                  http://www.gooseworks.org