On Thu, 2012-03-15 at 09:06 -0500, John F. Sowa wrote:
> On 3/14/2012 8:25 PM, Paul Tyson wrote:
> > Nobody should believe URIs by themselves will solve all problems.
But
> > thinking with and about URIs is a good problem-solving methodology.
>
> Every professional field was defining terminologies with standardized
> names long before computers came along. There are large numbers of
> ISO standards for them. And there is a trivial way to make them URIs:
>
> 1. Put all the terms and their definitions in a file.
>
> 2. Make that file available on the WWW with an http address.
>
> 3. Put # in front of each id to index the file.
>
> 4. Presto Zingo! All the unique identifiers are now URIs.
>
> We have universal agreement that URIs are useful. End of debate.
> (01)
If that is your conception of useful URIs, then indeed it is the
end--but only because you have reached the limit of your understanding
of URIs, not because there is no more to be said. No business or field
needs just another data dictionary, no matter if it is dressed up in SW
garb. (02)
(By the way, I suggested just such a program to the OMG a few years
back: a standardized way to prefix ISO terminologies, so for instance,
everyone didn't invent their own URIs for 2- or 3-digit country codes.
Last I checked, there were not even official URIs for ISO standards
themselves, so you can't make reliable "rdfs:isDefinedBy" assertions for
ISO terms.) (03)
> > Or contrast that with a decision to accept the ontologies provided
by
> > your COTS enterprise software packages. If you are lucky enough, or
> > small enough, to be able to run your business on only one such
package,
> > and never have to export any data, then you are golden. Otherwise
you
> > must work out some integration strategy, and again, in general and
over
> > the long run, an integration strategy based on externalizing the
> > necessary elements as URIs will be cheaper and more effective than
any
> > other.
>
> What planet have you come from? (04)
I come from a small, sparsely-populated planet where...nah, let's not go
there. (05)
> Every professional field and subfield
> has unique identifiers that can be converted to URIs by the very quick
> and simple method above. (06)
And if that is all they plan to do with them they should not bother. (07)
>
> But the problem of relating traditional terms to technical terms is
much
> harder. In medicine, for example, you also have the problem that
> the traditional terms in different countries with different languages
> and methods of diagnostics and therapy evolved in different ways. When
> you try to develop international standards, you discover that many
> technical terms have no one-to-one mapping. (08)
Yes! It requires people to think, observe, analyze, experiment, test,
and rework. All activities that are well supported by SW technologies
and ready-to-hand mainstream IT tools. (09)
>
> Then when you get to so-called "modern" technologies like computers,
> you might think the problem would be simpler. But no! Just imagine
> an engineer going from IBM to HP to Apple to Microsoft. Each step
> is a huge culture shock with different practices and terminologies.
> Even the move from the sales division of any of those companies to
> the engineering, manufacturing, accounting, or legal departments
> of the *same* company is a culture shock. (010)
It's veering off the thread a bit, but in both these situations there
are good reasons *not* to have unified terminology, whether it is a high
tech company or not. Competing enterprises must be different from each
other--both to seek market advantage and probably also for sociological
reasons of team-building, establishing tribal boundaries and reinforcing
membership, etc. Inside the enterprise, salesmen have different world
views (and consequently different vocabularies) than engineers,
accountants, and machinists. A prime fallacy of enterprise software
projects is to bulldoze all these distinctions away (to the IT guy, it's
all just data!) and impose an artificial "single source of truth" on
everyone. And by the way, these are "best practices", so all your
competitors are doing exactly the same thing (and therefore no one can
be "better"). (011)
>
> For another culture shock, go from a sales office of one company
> to a sales office of the same company in a different country.
> When I was at IBM, I was talking with a colleague who was born
> in Switzerland and worked at the IBM Paris headquarters. He was
> completely fluent in French, German, and English. But if he visited
> an IBM customer in France, he made sure that he brought somebody
> from a French sales office to make the presentation.
>
> > Contrast that with a decision to model things of interest to the
> > enterprise using UML/SysML, or EXPRESS in the mode of ISO 10303 or
> > ISO 15296. Development cycles are long and costly, and uninformed
> > by immediate feedback.
>
> What?!?!? The people who designed the Semantic Web technologies were
> looking at problems that are orthogonal to those issues. In fact,
> many people in the W3C recognize that UML diagrams are a good SW
> notation that people can read and use on Day 1. (012)
I think I mostly agree with what you said in those two sentences. UML
shines for software modeling but as you noted the Semantic Web aims to
accommodate everything so once you have modeled the software aspects of
SW, UML won't carry you much farther. (013)
>
> Furthermore, the tools for processing UML are completely integrated
> with mainstream IT, (014)
Do you have some examples of tool integration? I must be near another
main stream because none of the big projects I have worked on in the
last decade have made a serious effort to use UML. (015)
> and they have a simple mapping to the SW tools. (016)
Mapping, yes; simple, I don't know. I'd like to believe the OMG Ontology
Definition Metamodel will let us all sing in ontological harmony one
day, but at this point ODM's a much tougher sell than just getting a few
URIs up and running around the enterprise. (017)
> If the W3C had adopted UML diagrams in 1998, the Semantic Web would
> have been integrated with mainstream IT a dozen years ago. (018)
My main stream must here again be different than yours. Mine adopted
HTTP/HTML/XML reluctantly, tardily, and poorly. I don't expect them to
do any better with the Semantic Web. (019)
Regards,
--Paul (020)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (021)
|