[Top] [All Lists]

[ontolog-forum] fitness of XML for ontology (WAS: [ontology-summit] The

To: ontolog-forum@xxxxxxxxxxxxxxxx
From: Paul Tyson <phtyson@xxxxxxxxxxxxx>
Date: Sun, 02 Feb 2014 17:28:02 -0600
Message-id: <1391383682.6034.78.camel@tristan>
The summit proprietor requested we move this thread to ontolog-forum.    (01)

I changed the subject because I'm weary of the derogatory harangues and
cheap pot-shots about XML that crop up whenever a topic touches on
tools, data formats, or W3C standards. Let's have a frank and focused
discussion, preferably derived from in-depth first-hand experience,
about whether and how XML can benefit ontology practice.    (02)

Executive summary of my position:    (03)

Those who don't use XML in this area only have one oar in the water.    (04)

Furthermore, for the competent DIYer, all the standards and open-source
tools are available to build useful, and even breakthrough,
ontology-driven applications today in any enterprise using RDF, XML, and
associated standards around each (SPARQL, SKOS, OWL, RIF, RDB2RDF, XSLT,
XPath, XProc, etc).    (05)

Responses to John included below....    (06)

On Sat, 2014-02-01 at 20:46 -0500, John F Sowa wrote:
> Paul,
> Your note of Jan 24 was buried in a ton of notes in this thread.  But I
> just wanted to comment on the following exchange:
> >> But many voters in the W3C were in the grip of an untested ideology
> >> -- the use of XML as a wrapper for knowledge representation languages.
> PT
> > Untested? 30+ years of experience, starting with IBM's internal use
> > and transitioning to an ISO standard (SGML) in widespread use for
> > large-scale publishing, and then universal deployment on the WWW in
> > the guise of HTML? What would you consider "tested"? Maybe ASCII?
> I am *very* familiar with the *ML family of languages.  In the early
> 1970s, I was working at IBM, just down the hall from Charlie Goldfarb.
> He introduced me to GML for word processing.  Today, I still prefer
> to edit HTML tags for word processing instead of using MS Word.
> (I use OpenOffice or LibreOffice to convert HTML to .doc or .pdf.)
> In the late 1980s, I was working on standards for computational
> linguistics, and I *recommended* SGML for annotating texts.  I
> still recommend the *ML notations for that purpose.
> But there is a *huge* difference between annotating texts and doing
> knowledge representation.      (07)

Your use of "annotating texts" sounds dismissive in comparison to the
holy grail of "knowledge representation". I don't get the distinction
when it comes to machine-readable data. In both cases you must feed a
sequence of characters--some subsequences of which look like words to
humans--to a computer process. To make the sequences of characters
intelligible--to humans or computers--we use markup (or a degenerate
form of it in fixed-field-length records). As far as I know, there's no
getting along without markup if you want to get useful data in and out
of a computer process. So your only choice is whether to use
specialized or generalized markup. (As you know, but others mightn't,
the "G" in SGML was for Generalized. And the "S" for Standard, so you
have a standard generalized language, not a proprietary one. Perhaps the lack of
these reminders in the XML acronym contributes to the prevalent lack of
understanding.)    (08)

> In the 1990s, XML was *untested* technology
> as a notation for kn. rep.  The exercise with RDF/XML and OWL/XML
> notation has demonstrated that using XML for that purpose was a
> major *blunder*.    (09)

The only "blunder" was (in retrospect) undue emphasis on RDF/XML as the
only valid RDF exchange syntax, which has long since disappeared,
leaving only a strawman target for churlish criticism. RDF/XML itself
is--conceptually--an elegant solution to the problem of representing
graph data structures in a thoroughly hierarchical markup model. (It is
certainly not the only possible solution.) The fatal limitation of
RDF/XML is that the universe of URIs that people want to put in RDF
cannot be mapped easily to the range of allowable XML element names. I'm
not going to judge whether the designers of RDF/XML should have foreseen
this, or whether they should even have adopted a system that required
such mapping. It's a moot point. Here and now, you can choose to use
RDF/XML if it works for you, or ignore it if it doesn't. But don't throw
marbles on the floor to trip up people who are or aim to be productive
with XML and ontologies.    (010)

> Google went back to the lightweight RDFa as a bridge between the
> annotation and JSON for knowledge representation.     (011)

As above, I don't get the distinction. They chose a standard specialized
markup language for getting their data in and out of computer processes.
It works for them now. By the time they build it out to do all XML can
do, they will have reimplemented a generalized markup language.    (012)

>  And guess what?
> JSON is just LISP with brackets and curly braces.    (013)

LISP and JSON are both fine specialized markup languages with varying
degrees of usefulness in programming, "annotating text", and "knowledge
representation". They are *not* generalized markup languages.    (014)

--Paul    (015)

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    (016)

<Prev in Thread] Current Thread [Next in Thread>