Re: SEMIS Bulletin
Fred, Danny, Pete, Nagarjuna, Philippe, Edward, et al.,
I'd like to start with the response by Pete Kirkham,
since his remarks are closer to what is needed for
semantics than anything that is currently implemented
or proposed for the semantic web:
PK> So, to compare to ISO 10303, we've got external document
> reference IDs, EXPRESS and some metamodelling toolkits,
> all we need to do is agree the models by which we can share
> our semantic information. That's where STEP was 15 years ago
> when I was an undergrad. Agreeing the models is 99% of the
> task in metamodel based information sharing, and has taken
> well over a decade so far for engineering product data.
> That's in a domain which is a well understood and generally
> 'quite mathematical', and the only way that information gets
> usefully exchanged, even between two systems that are
> validated to conform to the standard model, is to have an
> engineer in the loop to catch exceptional cases.
For anybody who isn't familiar with STEP, EXPRESS, and related
technology, I strongly recommend them as examples of very
successful standards, tools, and methodologies that support
deployed applications that are far more "semantic" than
anything that is being proposed for the semantic web.
Following is a paragraph from one of their FAQ pages:
STEP provides a standard way to accurately and efficiently
exchange complete electronic product information. STEP allows
companies to effectively exchange information with their
worldwide partners, customers and suppliers, as well as
internally. In order to remain competitive in the global
marketplace, companies will have to ensure this exchange is
consistent, accurate and timely. By removing the barriers
that prevent maximum flexibility in design, manufacture, and
support, STEP will enable manufacturers to achieve new, higher
levels of quality and productivity while reducing costs and
time-to-market.
Source: http://pdesinc.aticorp.org/whatsnew/FAQs.html
This sounds remarkably similar to the hype that surrounds the
semantic web. But instead of hype, it's a description of what
people have been doing for the past 15 years. And they have
been doing it with tools, technology, and expressive power that
form a superset of what the semantic webbers are only promising.
In his conclusion, Pete says "hopefully less work re-done
(a common base for EXPRESS, UML and OWL would mean that much
duplicated effort could be avoided)". I would add SQL to that
list of work that should be integrated. My major complaint is
that instead of integrating and building on SQL, UML, EXPRESS
and other technologies, the semantic webbers have ignored the
rest of the world while redoing a tiny piece of it in a way
that's too little, too late, and too incompatible.
I'll start with Fred Kintanar's last sentence:
FK> By the way, I bought your Knowledge Representation book
> recently and it's great.
Thanks for the compliment. But that's not really fair since
I can't dump on people who say nice things.
But as I said about Pete's note, my major complaint about
the sematnic web is that instead of integrating and building
on successful and widely used earlier work, it completely
ignores everything that is currently in use. Examples:
FK> A big one is global scalability, where the key element
> is URI's (and their already deployed global acceptance).
Everybody agrees that you need unique global identifiers.
But they're nothing new: every telephone in the world has
a unique identifier, every postal address, every code for
electronic funds transfer, every book with its ISBN number,
every UPC bar code, Dun & Bradstreet (DUNS) numbers for
businesses, etc., etc., etc. But as usual, instead of
integrating and relating the existing identifiers, the
W3C throws in yet another independent scheme.
FK> I think the Web context of machine-to-machine interaction
> (including semantic web services) provides a source of new
> insights.
Of course. But things like that have been done with Arpanet
and other networks for over 30 years. TCP/IP, FTP, SMTP,
and many other protocols were in widespread use long before
the WWW. Except for the prettier graphics and simpler links
of the WWW, I was doing the same kinds of things I do today
on IBM's VNET in the 1970s and 1980s.
FK> Surely though the greatest value of the Semantic Web
> efforts is enlarging the Knowledge Representation community,
> and attracting a lot of new talent.
I agree that advertising is important, but what bothers me
is the similarity to the boom-and-bust cycles that have
plagued the computer field since the 1960s:
1. Some good technology is invented.
2. People overhype it by promising more than they can deliver.
3. Disillusionment sets in when the promises are not kept.
4. People run to the next new fad (which may be based on the
same technology, but with different buzz words).
FK> LISP has had its chance, and despite the seeding of Scheme
> and LISP in many of the world's best computer science students
> it is nowhere near XML's global deployment.
I cited LISP as an example of a far simpler and more powerful
technology that the W3C ignored. As I said, XML is great for
annotating documents, but incredibly ugly, inefficient, clumsy,
and inflexible as a foundation for knowledge representation.
See Philippe's note for further comments, which I endorse.
FK> Representation systems, and not just XML but other data
> and metadata formats as well, are architecturally crucial,
> especially when combined with URI's and interaction...
I agree. But see the next comment.
FK> If DL were dumped for CG's would it make any difference?
No. As I have said many times, DLs are a subset of logic
that could be and have been used with CGs. But the notation
I recommend for humans is controlled English (or other NLs)
supplemented with diagrams, such as UML or others, and
with Common Logic for computers (with translations to and
from the DL subset, the EXPRESS subset, the SQL subset,
the UML subsets, etc.).
Danny Ayers said,
DA> When I suggested it was 80% done I had something like
> the following definition in mind:
TBL> "The Semantic Web is an extension of the current web
> in which information is given well-defined meaning, better
> enabling computers and people to work in cooperation."
> ... -- Tim Berners-Lee
DA> The identifiers are in place, the languages have been
> developed and tools are springing up all over the place.
> This extension is well on it's way.
Tim's phrase "well-defined meaning" is so vague that it could
be interpreted in many different ways. I agree with Pete
that EXPRESS comes much closer to the goals of "well-defined
meaning" than RDF and OWL. Long before HTML came along,
manufacturers were already using EXPRESS to communicate
meaning across the Internet.
In summary, my recommendation is to replace the term "semantic web"
with "semantic system" and to integrate all related technologies
-- including SQL, UML, EXPRESS, OWL, RDF, and others -- on a common
logic-based foundation. The human interface should be controlled
NLs supplemented with graphics. XML could be used as an exchange
format for anybody who wants to use it, but more flexible formats
should also be permitted (for anybody who wants to use them).
John