Re: Parameter entities vs. GI name groups

David G. Durand wrote:
> 
> Many strings are not legal values in parameter entities, because, they are
> not, e.g. complete content model groups. (this makes modular DTDs harder to
> write, because of the need to wrap things the right way). You can't build
> an element name by concatenation of PEs and regular characters (this
> probably is a good thing, but the given the benefits of simple
> implementation, is not worth it). There are other restrictions -- but I can
> never remember exactly what they are -- same problem I _used_ to have with
> whitespace rules.
> 
> If every string-substitution effect were legal for a PE, then
> implementation would be so easy it's not an issue. Only WG8 could enable
> this kind of processing. for valid documents, however.

I understand this.  They have to respect SGML boundaries for 
legal combinations (to express myself loosely but with fewer
characters).

Because 

1.  Other than entities, no other form of string substitution 
is legal within SGML (discounting pre-processors)

2.  Too many constructs that occur *naturally* in extant 
artifacts are verbose in a regular element model

3.  Some sets of existing complex schemata will be 
unavailable and unusable without PEs

I concede the usefulness of PEs in XML.

Len Bullard

Received on Monday, 23 June 1997 22:55:59 UTC