🚀 go-pugleaf

In article <23254A@erik.naggum.no>, Erik Naggum <erik@naggum.no> writes:
> 
>Especially harmful is ignoring things that turn out to be more essential
>than the researcher thought.  This happens over and over, and I'm very
>skeptical of anyone who argues that X is non-essential and therefore not
>worth considering.  

  Thank you, Erik.  I'll repeat this to anyone that considers my problems 
with SGML as being non-essential.  :->

>I'm willing to concede the point that the structure is the
>essence, but SGML is in fact concerned with the linear representation,
>and not with the structure per se.

  This is the communications-protocol view of SGML.  Would it be correct 
to say, then, that the structure per se is captured in the DTD?  Or in
both the DTD and some external programs?  Or just the external programs?
Or just not at all? 
    
>I use SGML to define the syntax, and embed the semantics in
>the application 

  Ok. This is not what a data model does, unfortunately for those who 
thought that SGML was a data model (and perhaps that was only me).  A 
data model defines semantics and is mostly uninterested in syntax.  The 
relational model, for instance, defines what happens when you apply 
operators to a relation (i.e., the semantics of the operators), but could 
care less how you choose to represent the relation or the operators.

  What I hear Erik saying is that "the marked-up document is portable,
but the meaning of it is anybody's guess".  This is directly due to
the embedding of semantics in applications, and is the primary reason 
that, without support from some data model, SGML will result in 
non-portable, application-dependent documents.  

-Darrell.