[sc34wg3] More slides from Extreme (Subjects and Proxies)

[Robert Barta]

On Wed, Aug 04, 2004 at 10:41:50PM +0100, Ann Wrightson wrote:
> Model sketch for TMRM subjects and proxies as just presented at the Extreme
> Q&A session. Comments welcome! (Robert - wd be particularly good to hear how
> you think this plays with \tau)

Ann,

I cannot find anything which would collide with \tau.

The procedure, though, would be much more formalized:

 - there would be two maps, READ2 and SNOMED

 - each of them would have as "disclosure" a constraint set:

   READ2:
      every topic must be associated with a property 'read2'

      no two topic can have the same value for this property.

   SNOMED:
      ditto but with property 'snomed'

   Up to this point they can operate independently. Applications
   can use (either at boot time or on-the-fly) the constraints for
   identification purposes.

 - for mediation between the two taxonomies, there is a relation

       mediate \subset READ2_ids x SNOMED_ids 

 - from this relation we can derive a combined map

     READ2 + SNOMED

   by simply first throwing every assertion in a big pot. Then
   we take the constraints above and add:

   "No two topics in the combined map can have the property/value
    pairs read2/g33 and snomed/111111".

   We repeat this for every element in the set 'mediate'.

   This can be done by really creating this map, called
   'materializing' in the DB arena, or it can be done on the fly
   (virtualized).

Software-technically it simply happens by wrapping a virtualization
layer around the map READ2 + SNOMED. Any request for

   (READ2 + SNOMED) / snomed_ct:111111

would then be forwarded to the SNOMED map. Any request to

   (READ2 + SNOMED) / read2:g33

to READ2.

Again, lotta technical variations are possible, like inventing
a new, combined identifier from the set 'mediate'.

\rho