Hi John, (01)
On Sat, 2012-02-25 at 22:57 -0500, John F. Sowa wrote:
>
> I have been working, writing, teaching, and preaching about semantic
> technologies for over 40 years. I'm not going to abandon them just
> because one system or another is flawed. (02)
And because of that, I assume your comments in this thread are in the
mode of a professor heckling his seminarians to defend their theses, and
are not charged with any personal animosity towards those who favor this
or that technology. (03)
For over 25 years I have helped with the computerization of business
processes in large commercial enterprises, the aim of which I have
always taken to be the conveyance of meaning from one mind to another
for the purpose of enabling, enhancing, and coordinating the activities
of many individual people to pursue common enterprise goals. My
operating formula is simple: I adopt any technology, methodology,
principle, or practice that helps this aim; I avoid what doesn't; I
assail what degrades it; and I assay what shows promise. (04)
>
> PT
> > It is a good lesson in what not to do with RDF, and affirms
> > what I have always thought about "big bang" transitions in
> > general. I completely agree with your and DS's assessments.
>
> I mentioned that example because the mistakes were very clear.
> But it illustrates a common problem: A bad decision was made,
> a lot of money was invested, reviewers were hired to bless the
> decision, and DS tells them something they didn't want to hear.
>
> The usual outcome is that management shoots the messenger, goes
> ahead with the plan, and crashes into a train wreck.
>
> > It is not, however, a good argument against my proposition that
> > enterprises who invest in URIs will get more utility in the long run
> > than those who continue to put all their intellectual property eggs
> > only into vendor-controlled baskets.
>
> Nobody argued against URIs. As Matthew said, databases have used
> unique identifiers in similar ways since the 1970s. Logicians have
> known the importance of identifying and resolving ambiguous terms
> since Aristotle.
> (05)
I confess I am not aware of any similar uses for URIs and RDB record
identifiers (unless you use URIs as primary keys). One URI use case is
to make a system of resolvable URIs for all the part numbers of interest
to a manufacturing enterprise. This would allow RESTful access to
information about any part. How would RDB or application-specific ids
support that in cross-application, human-friendly fashion? (06)
> > I have used RDF to improve or enable processes that would have been
> > difficult or impossible to do with some other technology.
>
> Those problems were solved with LISP back in 1959. The chief designer
> of RDF was R. V. Guha, who said that he wanted to use LISP notation,
> but the W3C forced him to use XML. (07)
Well I didn't mention the particular applications I had built, but
regarding solutions with LISP, I agree with you--as I agree with Paul
Graham and others who claim the world would be better with more LISP,
and with the recycled adage, "those who do not know LISP are condemned
to re-invent it". I regret that my career has not afforded me greater
opportunity to learn and use LISP. But that would certainly have taken
me farther away from mainstream IT than I find myself now--which is why
I'm surprised you introduce the topic in a discussion that started
around mainstream IT compatibility. (08)
But since you started, let me go on. In the late 1990s I learned
DSSSL--surely the most elegant information processing specification ever
to come out of ISO. It was based on Scheme, so I learned quite a bit of
Scheme along the way, and a smattering of LISP. When DSSSL's author,
James Clark, announced "DSSSL is dead" a few years later I refused to
accept it. I clung to the hope that somehow people will--nay, must--see
the beauty and utility of DSSSL. I hoped far longer than a reasonable
man should. But when my lamentations had ceased I took a look at Clark's
next creation, XSLT, and saw that he had put most of what was good, and
even some of what was beautiful, from DSSSL into XSLT. (09)
So W3C's grudge against LISPish notation was not only aimed at RDF; they
chose to re-implement in XML notation, at some loss of functionality, a
perfectly good Scheme-like transformation specification. But after all,
XML is just uglified s-expressions. And LISP's power comes not from
parenthesized lists alone, but from a very large language specification
that says how to interpret those lists. So maybe we're not in such a bad
state after all. (010)
Your assertion that RDF is flawed because it was not done LISPishly
puzzles me. RDF is an abstract syntax that constrains your utterances to
triples, forces you to use URIs as the first two components, and
suggests some weak semantics that include interpreting the components of
a triple as grammatical subject, predicate, and object. You can write
your triples any way you care to--the W3C provided a starter set of
concrete syntaxes, and is adding more. How would LISP have rectified the
specification? (011)
>
> PT
> > I have not reached or foreseen the limits of [RDF's] utility
> > in my current position, but my hand has been stayed in this
> > direction and now I can only watch as the conventional non-RDF
> > approaches run into trouble.
>
> Now you are talking like the zealot who convinced the IT manager
> in the horrible example to switch from Oracle to RDF. (012)
I don't know whether to take that as compliment or criticism, though I
think you intend the latter. My dictionary says zeal means "Enthusiastic
and diligent devotion in pursuit of a cause, ideal, or goal..." As I
explained in the preface to this note, I have an unwavering commitment
to improving the semantic circuitry of the enterprise, so I admit to
zealotry in that sense. But I do not see any resemblance to the
unfortunate fellow in the horrible example. I like to see things work; I
like to make things work; it saddens me when they don't. (013)
Would you be as dismissive if I had put LISPish "DSSSL" in place of
"RDF" in my stories? Twelve years ago I would have made exactly such
statements, based on real working applications. (014)
>
> Please note that the schema.org group (Google, Microsoft, Yahoo!
> and others) have some very intelligent R & D people. They looked
> at RDF, but they adopted schema.org instead. (015)
Those great minds are being applied to problems mostly unrelated to my
work, and often towards ends with which I do not sympathize: how to
maximize ad click rates; how to collect and use crowd behavior data to
usurp freedom of choice and prejudice information-finding; in short, how
to bend the semantic circuitry of the WWW to profit the lords of the
cloud (to use Jaron Lanier's term). I aim to make straight the ways of
communication in the enterprise. (Nevertheless, I do appreciate the
remarkable non-semantic innovations that some big web companies have
developed and contributed to open source.) (016)
> They use an approach
> that Guha and Bray could have adopted back in 1998: put very simple
> HTML tags in web pages and gather all the triples (or N-tuples) in
> JSON notation. (And JSON is just LISP with curly braces.) (017)
The first part of that sounds like RDFa. (018)
JSON is JavaScript Object Notation, a serialization format for
JavaScript objects. When it gets a hefty specification to interpret it
as something other than JavaScript objects, it will be LISP with curly
braces. (But why would we need that?) (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)
|