Re: Final words, I think, on error handling

On Wed, 7 May 1997 23:03:35 -0400 Arjun Ray said:
>The basic point against the Draconian case is that a single
>(monolithic?) policy towards error handling is a recipe for failure.

You have a point.  It may be useful to point out that the draconian
policy adopted by the ERB is not, in fact, a monolithic policy toward
all error handling, but a monolithic policy as regards one class
of error (the class that prevents documents from being even
minimally XML conformant).

>In fact, I believe the real stumbling block here is SGML's monolithic
>conception of "Error" itself. Dare I say ontologically, the

In both SGML and XML, moreover, the notion of error is not wholly
monolithic, in the sense that SGML -- like some programming-language
specs I've seen -- distinguishes errors (with results not defined by the
spec) from reportable errors (which a validating parser is required to
detect).  XML has errors, reportable errors, well-formedness violations
(which are by definition reportable), and violations of validity
constraints.

If you mean by 'monolithic' that the notion of error in SGML and XML
does not allow for quantitative qualification (mild, medium, serious,
for example), then I think you may be right, but until we get a
renovated formal language theory based on fuzzy rather than Boolean
logic, and parsers based on that new formal language theory, I am not
sure how we are going to get away from the fundamentally monolithic
nature of 'error' in the sense of 'ungrammaticality'.  All the notion of
'error' does in XML is identify certain specific classes of strings
which are explicitly not members of the language being defined.

But the theory of formal languages as applied to document grammar is a
topic perhaps better aired on comp.text.sgml, so I'll stop now.

-C. M. Sperberg-McQueen

Received on Thursday, 8 May 1997 11:18:55 UTC