Re: [xml-dev] What is declarative XML? (And what's not)

[Rick Jelliffe <rjelliffe@allette.com.au>]

Costello, Roger L. wrote:
> I assert that documents that have no inherent processing semantics have substantial benefits in terms of reusing them and mining them and mashing them up.
>   
This is the old SGML idea of taking out *all* processing semantics and 
then having another layer to add them back again. Even if you didn't 
need to.

In the particular case of rich text, this approach has just about lost. 
Consider five ways of marking up a warning section:

 <warning>...

 <warning render="section">...

  <div class="warning">...
 
  <section> ...

  <br />...

It is the middle one that has won, in effect. 

I think the principle is that where there is a mass technology or 
well-accepted public vocabulary which represents the base or default 
semantic (i.e. the semantic which any reasonable use would establish) it 
is better to subclass this. There is also a simplicity principle 
involved too, that you don't strip out what you must immediately add 
again, I think. What is easy to process is an important consideration 
that swings both ways: if the data is textual then subclassing HTML may 
be the simplest to process, but if the data is very rich and structured, 
then marking it up as HTML might increase the difficulty of processing 
it. (Insert GRDDL story here.)

Let look at the case of definition lists. You could reformat them as two 
column tables. But it does no harm to mark them up as lists.

One of the benefits of pure descriptive markup is that, if it is 
complete enough, it prevents the user from having different ways to do 
the same thing: the constraint "Does every address have a postcode?" can 
be checked more easily in pure descriptive markup.  (Though this is 
because grammar-based schema languages are terrible at checking 
"architectural" structures such as those marked up using html:*/@class.)

Cheers
Rick Jelliffe