Kevin, (01)
I'm sorry that I made some harsh comments about your note. But I
was frustrated by hearing the kind of remark that has killed more
promising projects than any other: (02)
KDK> ... but I think that such a powerful framework is not really
> needed for this particular use case. (03)
That's the same kind of remark that Guha made about RDF, which is
the single biggest disaster that was inflicted on the Semantic Web. (04)
At the time that Guha and Tim Bray designed RDF, commercial web
sites, large and small, were built around relational databases.
If they had adopted something like JSON as their notation for
n-tuples, they could have had a seamless mapping of web data
to and from any RDB. (05)
Another opportunity they overlooked was the popularity of UML
diagrams by software developers. The notations for type
hierarchies and E-R diagrams, for example, could easily be
adapted to web data. In fact, they could be used as
readable adjuncts to any kind of Description Logic. (06)
Furthermore, there were many years of outstanding R & D work
on extensions to RDBs and OODBs presented at the yearly VLDB
conferences. Much of that work went far beyond the limitations
of commercial SQL systems. (07)
If the SemWebbers had integrated their web-based innovations
with the best R & D that had been done by the OMG, the VLDB,
and the AI community, they could have had an outstanding
foundation for the future. (08)
Instead, they took some DL technology, encrusted it with the
most horribly bloated and inefficient syntax the world had ever
seen, and made it incompatible with the mainstream of software
technology. Then they wondered why the Semantic Web "didn't
reach its full potential." (09)
Some further comments: (010)
JFS>> If you start with FOL, you have 120 years worth of very well
>> documented definitions... (011)
KDK> You at least need to reference a particular definition. (012)
I mentioned Mathematica. But another option is the mathematical
toolkit of the ISO standard for Z notation. See pp. 86 to 127
of the following book: (013)
http://spivey.oriel.ox.ac.uk/mike/zrm/zrm.pdf (014)
Z was designed by a bunch of mathematicians who went overboard
on using Greek letters and special symbols. But the ToolKit could
be translated to any CL dialect. (015)
KDK> ... in XPath (the source of the arithmetic definitions used
> in XBRL), you write x+y, x*y, and x/y. Since you're writing the
> same thing, why is your standard (CL) better than their standard
> (XPath)? (016)
XPath is a syntax-based tool for following paths in graphs, whose
semantics could be completely undefined. (017)
KDK> I can see why you (as an ontologist) would prefer a CL-based
> definition, and I can see why web developers would prefer the
> XPath-based definition. Why should your preference get to trump
> theirs? (018)
I have no quarrel with anybody using whatever implementation tools
they find useful. I worked at IBM for 30 years, and I have an
extremely high regard for practical solutions. But if you want
to share the data in those graphs with other kinds of tools, you
need some agreement on the common semantics. (019)
KDK> I also know LISP, and Perl, and Java, and XSLT, and all of
> them also have extensive libraries of all kinds of things already
> defined for you. And I absolutely wish along with you that they
> didn't all define the same things redundantly, and often subtly
> incompatibly. (020)
That's very good. With that kind of background, I don't understand
why you disagreed with what I said. (021)
KDK> Combinatory logic, by the way, is even "simpler" than FOL:
> the whole thing can be defined with only two operators, named
> "S" and "K". And you can define FOL in combinatory logic.
> So why don't we adopt that as the one, true uber-standard?
> What could be simpler? (022)
Combinatory logic is just a different syntax for FOL. It could be
defined as a dialect of Common Logic, if anyone wanted to do so. (023)
We've been trying to get people to focus on the semantics. That's
why the body of the CL standard is independent of any concrete
notation. The three concrete dialects are specified in normative
annexes, but the standard explicitly says that any notation that
conforms to the CL semantics can be called a CL dialect. (024)
JFS>> That's why there is no need to define new kinds of languages.
>> The hard work has been done, and you can focus on the content. (025)
KDK> No, what has been done is not "the hard work", it is the easy part! (026)
We have to be clear about what parts we're referring to. I think we
can agree that defining the content is hardest part. In designing CL,
we considered our work to be an engineering project, not a research
study. (But in defining the CL model theory, Pat Hayes and Chris Menzel
did make some interesting innovations that were needed to support some
of the rather free-wheeling features of RDF and OWL.) (027)
But as I said before, the definition of the abstract syntax and
semantics of Common Logic required only 12 pages. That is certainly
not a lot of hard work. But that kind of common foundation is
necessary if you need to share data among a wide range of different
systems with many different notations. (028)
I haven't analyzed the logic part of XBRL, but I would be very
surprised if it couldn't be specified as a dialect of Common Logic.
But that exercise is necessary to make sure, in your words, that
they don't "define the same things redundantly, and often subtly
incompatibly." (029)
John (030)
_________________________________________________________________
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
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (031)
|