[topicmapmail] PSIs - alternatives

[Simon Grant]

This strikes me as very clear and persuasive reasoning.

At 20:50 2006-06-21, Steve Pepper wrote:
>The purpose of the subject indicators (or resource descriptors) in the
>Published Subjects (or PRI) paradigm is to provide *humans* with just enough
>information to decide whether or not the PSI/PRI suits their needs, and they
>should therefore contain the minimum number of assertions necessary to
>achieve that purpose. The more assertions, the greater the likelihood that
>people will avoid using the PSI because they disagree with the opinions that
>the minter of the PSI holds about the subject in question. This defeats the
>whole purpose of PSIs, which is to establish agreement concerning the
>*identity* of subjects, not agreement concerning *opinions* about those
>subjects.

Indeed, one of the key precursors to finding agreement is to be sure 
to pick something people are able to agree on - the identity of 
subjects - and to avoid things people are likely to disagree on.

I think what would be useful to look at is the information necessary 
within the PRI/PRD system as a whole. There is
1. as Steve above (providing humans with information)
2. whatever else may be necessary towards a self-sustaining 
infrastructure and motivation to use it.

It seems to be Steve's position that all of 2. can be provided 
outside the PRD, with the PRD used exclusively for 1. This is 
plausible to the extent that proposals for providing the rest of 2. 
are plausible. What I read here is that one class of approaches, 
involving central repositories or directories, is not thought to be 
plausible by many.

Steve's requirements, set out in section 2 of "The Case for Published 
Subjects" at
http://www.ontopia.net/topicmaps/materials/The_Case_for_Published_Subjects.pdf
also seem very clear and persuasive. Section 2.4, "Easy for humans" has
1) given a subject, it should be easy to find an identifier (if one 
already exists);
2) given a subject, it should be easy to mint an identifier (if 
nothing suitable already exists);
3) given an identifier, it should be easy to find out what subject it 
identifies; and
4) given an identifier, it should be easy to insert it into a 
document, index, database or other source of information.

and section 2.5, "Easy for computers" has
"Any processing that requires more than simple string comparison 
(whether it be string normalization, network retrieval of a resource, 
or some form of computation), will invariably lead to more 
heavyweight and less reliable applications."

A key consideration in discussion here is, what if (as seems likely) 
people persist in minting "their own" identifiers for subjects where 
there are already identifiers? There are many reasons why this is 
likely to happen, including ignorance, laziness, and failure of 
discovery mechanisms, but also dissatisfaction with the accuracy or 
reliability of another authority. And when it has happened, as it 
inevitably will, following Steve's "easy for computers" requirement 
it will be essential for everyone to converge on the same one. Will 
this happen, and how long will it take? What will happen in the meantime?

My suggestion addresses both the situations where people won't do the 
merging, and the situation where alternatives exist prior to merging. 
Having a list of accepted equivalent PRIs included means that the 
machine comparison is still relatively easy - it would involve 
fetching the two PRDs and then string comparison of two lists against 
each other. That's all. Maybe Steve and others think that even that 
is too much to ask. If so, I have no answer. But is it really too 
much to ask, when it would produce what looks like a workable 
short-to-long term solution without the need for anything centralised?

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