🚀 go-pugleaf

Darrell Raymond writes:

>  I take it, then, that you would disavow the following statements
>by Travis?
>
>>   The syntactic model of SGML, however, lends itself
>>   very well to describing virtually any type of data, both
>>   in content and in relationships.
>>
>>   So, the next time someone asks you what SGML is, instead
>>   of saying "it is a language used to markup text", say
>>   "SGML is a powerful tool that can be used to describe,
>>   model, and interchange data."

I think Brian has oversimplified.  I think the better way to
say this would be to say "SGML systems can be powerful tools
for describing, modeling, and interchanging data", where by
"SGML system" I mean an SGML-based application that defines
and understands the semantics and uses as its base data storage
form SGML documents.

I think the question you are asking, at least implicitly, is
"is it useful to consider building data modeling systems
on an SGML base?"  This is certainly an interesting question,
one worth asking and investigating.  My gut reaction, as I've
already said, is "yes, it is."  You may be able to convince
me that there are relationships that cannot be represented
practically using SGML methods.

You mention that in SGML, interchange is only enabled at
the application layer.  You are more or less correct.  SGML
never promised more.  Interchange without regard to DTDs
and the semantics behind them (not defined by SGML constructs),
is only meaningful when the semantics are not important, such
as for DTD-driven SGML editors that can provide tag menus but
not formatting or other processing without additional information
(SoftQuad's Author/Editor and ArborText's The Publisher are examples).

However, there is, as I've said, no barrier to defining standard
languages for specifying semantics such that interchange between
processors is enabled.  DSSSL is an attempt to do just that, as
is HyTime, as is the Davenport project.  There is no reason you
couldn't define an SGML-based architecture for data modeling
languages and associated processors using HyTime as a model.
Note that HyTime defines validation that a conforming HyTime
engine must perform that is semantic in nature.

Eliot Kimber                      Internet:  drmacro@ralvm13.vnet.ibm.com
Dept E14/B500                     IBMMAIL:   USIB2DK9@IBMMAIL
Network Programs Information Development     Phone: 1-919-543-7091
IBM Corporation
Research Triangle Park, NC 27709