[topicmapmail] PSIs - alternatives

[Simon Grant]

Useful to see Steve's presentation. I wonder why 
the name "Canute" (Knut) comes to mind? :-)

At 14:16 2006-06-20, Steve Pepper wrote:
>It is inevitable – because of the bottom-up 
>nature of the PSI minting process – that 
>situations will arise in which multiple PSIs 
>have been created for the same subject. Although 
>this can be handled quite easily using mappings, 
>as you suggest, this does require some work, so 
>the ideal solution – and the one I believe we 
>should strive towards - is one in which de facto 
>standard sets of PSIs emerge over time. 
>Publishers (“sponsors”) of PSIs who believe that 
>their PSIs are equivalent to others should be encouraged to consolidate.

The ideal, I would agree. The practical solution? 
I think there are indicators that this may not be 
so. At least people here don't all agree that it is a good practical solution.

(Simon Grant)
>A "democratic" alternative would be to have PSIs resolving to a file
>that named other, equivalent, PSIs, along with a description of the
>subject as seen by that authority.

(Steve Pepper)
>I think we need to be very careful about what 
>kind of machine-readable content we suggest that 
>people put in their subject indicators (or 
>Public Resource Descriptors, to use the 
>terminology of [1]). Having said that, some kind 
>of referral mechanism, by which publishers can 
>deprecate their own (outdated) PSIs in favour of 
>others, would probably be useful. [...] In 
>general I think we should encourage people to 
>put as few assertions as possible in their PRDs 
>(subject indicators): the purpose of a PSI/PRI 
>is simply to help us know when we are talking 
>about the same thing, not whether we hold the 
>same views about that thing; including 
>assertions (even “innocent” assertions such as 
>near-equivalence) are therefore not only 
>irrelevant, they are directly counter-productive.

Careful, I would agree. What I am suggesting is 
exactly a careful, and reasonably minimal, 
approach to this. I'd be happy without 
near-equivalence, just retaining equivalence, 
where practical. And where the subject is clearly 
discrete, such as a person, this should be 
straightforward. Disallowing near-equivalence 
would probably mean, in practice, that PSIs for 
abstract subjects would be more fragmented. 
People do seem to agree that a piece of text 
indicating what the intended subject is would be useful.

>On the other hand, that kind of assertion could 
>certainly be justified as part of a service intended to aid PSI discovery.

So I'm thinking of a pretty simple information model for a PSI / PRI.
- The usual Dublin Core-like stuff (title, description, author, date, ...)
and EITHER
- a list of equivalent PSI / PRIs
OR
- a single replacement pointing from a superseded 
one to the new one (in which case the DC stuff would not be needed)

Lutz Maicher's ideas are interesting, too. I 
wouldn't suggest that the putting together of 
PSIs is automatic - that would invite automatic 
abuse - e.g. I claim that the PSI of a politician 
I love to hate is the same as a despised animal. 
But a tool to help people manage PSI equivalences 
in their topic maps would be handy.

Simon

--
Simon Grant  http://www.simongrant.org/home.html
Please continue to use my established e-mail address
a (just by itself) (at) simongrant.org