[tm-pubsubj-comment] RE: [topicmapmail] Re: [topicmaps-comment]Genetic PSIs [Re: Topi c Map domain, paradigmatic PSIs ...]

["Mason, James David (MXM) " <mxm@ornl.gov>]

At the risk of redundancy, I'm going to retransmit what I said earlier to a
more limited audience. 

I do not believe that it is our task in any of the standards
committees/discussion groups/consortia party to this discussion to preserve
knowledge: our task is to make a contribution to enabling others to preserve
knowledge. Collectively we are tool makers, not librarians (though some of
us have worked as librarians). We are building a framework for creating
catalogs, collections, and the like, not actually building the catalogs or
acquiring the collections. Anything we do should not be predicated on having
any particular resouces. 

All the schemes for massive replication of data, etc., sound well
intentioned, but as I suggest below, I can turn to the other computer
sitting next to the one I'm writing this message on and show (if you have a
clearance to see it) an example that breaks them all.

I take John Sowa's advice to heart. He likes (paraphrasing his words) broad,
shallow, very generic ontologies, as opposed to the SUMO and Cyc people who
are trying to build comprehensive, deep, detailed ontologies. 

I think we need a simple framework that requires very little work on our
part (like *no* replication of data by us). Then we can leave it to other
people (1) to build the larger framework of data preservation (notice I
don't say knowledge preservation) and (2) populate that framework with
knowledge built with our tools.

(Perhaps we need to preserve knowledge of our own deliberations, but that's
something that perhaps we have resources to do.)

Jim Mason



Here is what I said to the original audience:

I want to back this discussion up a bit, like to the very beginning. 

Before we ask *where* things need to be on the Net, we ought to ask
*whether* they have to be anywhere on the Net at all.

If these "paradigmatic PSIs" are indeed the definitions of components of the
Topic Map mechanism, like the  "role" role, and we put the in the standard,
is that not sufficient? Why do these things have to be *anywhere* other than
in the text of the standard?

If you tell me that the "paradigmatic PSIs" have to be somewhere reachable
by Net for TM systems to work, then I have serious problems with their very
existence. Take, for example, my own applications. Nobody else in the TM
community can see the networks I work on, and those networks can't see
anything that's on the Net that most of us generally use. If my TM
applications won't work without seeing something stuck somewhere out on the
Net, then I'm shot down. Forget about using TMs for anything.

Now, if, as Lars Marius suggests, he wants a PSI to explain Ontopia, that's
a different matter. He can put it somewhere on psi.ontopia.net, and that's
good enough. Whether I can reach it or not matters only if I want to see
Ontopia's self definition.

If you really want to put something on the Net, you're back to an issue
that's been around as long as I've been messing with standards, though that
began before most of you were probably aware of the Net, and we didn't build
it into any standards. You're talking about registered data and having a
Registration Authority to maintain it (or you're talking self-registration,
as in "psi.ontopia.net"). SC34 and its predecessors have been down this path
several times before. Think of (1) ISO 9070 and (2) ISO/IEC 10036. 

The former is for "SGML Public Text", and Charles made it a self-registarion
standard. If you proclaim something to be "SGML Public Text", it *is*
registered as public text; nothing more needs to be done. Whether that thing
in a <!DOCTYPE is reachable on a network or not doesn't affect the validity
of the DOCTYPE statement or the instance document that contains it (though
it may make a mess of subseuquent validation if you can't reach it). 

The latter is for assorted font and glyph stuff. It calls for a RA. When
AFII, the RA, went out of business, SC34/WG2 had to jump through all sorts
of hoops to get the RA moved to a university in Japan. Not a good thing.
(GCA was also a RA for "SGML Public Text": but now GCA is IDEAlliance. What
does that do for anything that was registered?)

If you're wanting an enduring body to host a Web site, you're asking a lot.
I can pretty much guarantee that neither ISO nor ANSI nor JTC1 will do this
for you. I don't think LOC would do it either. I can put anything you ask
for up on the Y-12 (not ORNL) site (so long as it's not illegal data), but I
won't make any guarantees about its staying there for ten years. I wouldn't
bet on OASIS, either. Likewise GCA/IDEAlliance. Likewise the W3C. OCLC has
recognized this sort of problem, and that's why they established the PURL
mechanism. But I gather that's a front for something else, and the something
else takes maintenance. Who's going to do the maintaining? I certainly
wouldn't put something at topicmaps.net or isotopicmaps.org. Who's
guaranteeing them?

All this brings me back to the initial question: if we standardize
something, why does it have to be anywhere other than in that document? If
we define something to be the "role" role, that should be the end of it. 

I've argued with Charles about a lot of things over the years, including
this subject. I hate to admit it, but I think he was right in 9070. You
shouldn't commit future users to needing some resource that needs
maintenance. It's asking too much. Self registration is the way to go unless
you're actually standardizing the essence of the very lowest level of
something (like assigning code points in UNICODE -- and look at the task
they have on their hands for stability, not to mention volume). 

If we want to say what a PSI should look like and how it should be
documented, that's fine. We can make the standard an example of itself. If
we want to issue further guidelines for those who come after us, that's fine
too. That is, I believe, what Bernard's committee is about. But let's not
make the standard depend on anything more than its own words, wherever they
may be (like on paper in a drawer).

Jim Mason

James David Mason, Ph.D.

Y-12 National Security Complex
Bldg. 9113, M.S. 8208
P.O. Box 2009
Oak Ridge, Tennessee 37831-8208

+1 865 574 6973

Chairman, ISO/IEC JTC1/SC34
http://www.y12.doe.gov/sgml/sc34/ornlsc34oldhome.htm