++1

DTDs are at about the right cost and complexity for what schemas do best:  protection from and of the humans.   The XML designers were good programmers but terrible sociologists.

The web is a design delimited by programmer-centrism in a society where programmers are required and loathed.  The paranoia is still quite evident.  XML Schema is a stellar example of how far things can go off the rails by trying to do too much with a single piece of technology in a multiverse of technologies applied in a pipeline.   Prescription is good where contracts have to enforce quality over several universes of technology and humans with multiple agendas both real and imagined.

Data?  What is that?

Texts:  I work with one of the more if not most complex DTDs.  The organizing unit is the work package.  Of the approximately 69 work packages specified, 27 are used with a noticeable frequency.  Of the 27, two are used most frequently by an order of 2.  Of these two, they share roughly 70 percent of the same markup types.   As a result, in systems where the style sheet and not the DTD control the real-time presentation and navigation, tagging to the wrong part of the subtree is easy and won’t be noticed unless someone (a human) actually inspects the tags because the determinant data is in the titles.

Far too often in the past the employer/employee (say personnel record) is used in these examples.   This is a trivial dataset being used to justify a very expensive decision.

The diseases are often ignorance, ego and self-importance.   No size fits all.  Some sizes fit most.

len

-----Original Message-----
From: Sean McGrath [mailto:sean.mcgrath@p...]
Sent: Wednesday, April 10, 2013 6:11 PM
To: xml-dev@l...
Subject: Re: Probabilism (was Re: Fwd: Not using mixed content? Then don't use XML)

>>[Sean McGrath]

Schemas ain't probabilistic.

>[Simon St. Laurent]
>How does "perpetually renegotiable" compare to probabilistic?

>That's another key piece of human conversation I'm bringing to this story.

Yes. For the probabilistic parts, Sausserre. For the "perpetually renogotiable" parts, John Searle's concepts of Social Reality and the difference between what he calls "brute facts" and "institutional facts":
http://en.wikipedia.org/wiki/John_Searle#Social_reality

Meaning is not something that is "out there", ready to be bound into our content. As content moves between producers and consumers - and especially across institutional boundaries - the meanings that are superimposed on the markup and the content change.

As time passes, the "semantic drift" can also be witnessed within a single institution where the "markup of our forefathers" is reinterpreted by a newer generation.

So it goes.

-- 

Sean McGrath, CTO, Propylon Inc.

http://www.propylon.com

http://seanmcgrath.blogspot.com

@propylonsean