Re: Parameter entity references

>>Digitome Ltd. wrote:
>> This semantic parse would naturally subsume the structural checks
>> we *could* have performed with a validating parse but what would
>> be the point? We need our own layer of semantics anyway. It seems
>> natural to combine the two.
[Paul Prescod]
>Where are you going to find an editor that restricts your content
>models?
For this app we don't want restricted content models in an editor. We want
users to be able to add element types on the fly. We just want well formed
XML. (Aside:many authors I know *hate* the top down  authoring model that
on-the-fly validators enforce. For some document types it is simply 
not appropriate).

>How are you going to describe your semantic restrictions to
>someone at another site using different software than you? The answers
>to these questions are important to many of us.
>
These questions are very important to us as well. I am a believer in DTDs!
However they are not the be all and end all of validation techniques. There
are problems for which they are not appropriate.

We may well have DTD's that describe *aspects* of the semantic
restrictions either as DTDs or meta-DTDs. Other aspects will need to
be captured some other way. Here is a "for instance":-

   "Paragraphs in a section can be numbered. If any paragraph is
   numbered, they must all be numbered. Numbers must be unique 
   and must increase in value but not necessarily in single unit increments."

JSP diagrams, pseudo-code and good old English will be used for this
sort of thing. What I obviously loose by not using SGML validation is the
link between my documentation of the semantic restrictions and the restrictions
themselves - they can get out of sync. This is one of the big wins of
SGML validation. For this app we have decided to live with its loss
because of the amount of non-DTD validation required.

Don't get me wrong. I believe in DTDs. I believe in SGML validation. I am 
simply pointing to the fact that for some apps, the validation that SGML
can provide is not enough. Sometimes so much extra semantic validation
is required as to make "WF XML + custom code" the appropriate solution.

Regards,
Sean

Sean Mc Grath

sean@digitome.com
Digitome Electronic Publishing
http://www.digitome.com

Received on Tuesday, 3 June 1997 04:03:34 UTC