Andy Dingley <dingbat@c...> wrote:
> What specifies the permitted root element(s) for a document ?   HTML,
> SGML, XHTML or XML ?

When using a DTD, any declared element type can be the root element.
It must be specified in the Document Type Declaration in the XML file.
The same is true for SGML, HTML, XHTML eg

    <!DOCTYPE table PUBLIC "-//W3C//DTD HTML 4.01 Strict//EN">

specifies a document starting with <table> and containing anything
valid in HTML 4.01 tables.

Warning: *browsers* are not SGML conforming applications, so they won't
understand this. They *will* understand if you use XML or XHTML, but
I don't know what their reaction to a XHTML fragment would be.

> My question is, is there any way to author a non-HTML DTD (SGML or XML)
> so as to restrict valid documents to only allow a certain subset of
> their elements to be used as the root element? 

Yep, just use the element type name of your choice in the Document
Type Declaration. This is required to be supported by all conforming
editors using a DTD. If you use a Schema, all bets are off, as the
specification of a root element type is done quite differently there.

> Can this restriction be
> expressed _entirely_ within a DTD?  

No, not at all. *Any* element type of a DTD can be used as the root
element type.

But conforming applications (eg editors) usually make a good guess
if they are worth anything, when they parse the DTD -- it's not
hard for them to spot that at least one element type is never used
in the content model of any other element type, and is therefore a
good choice for a default root element type. Oddly, some otherwise
very good editors fail to do this, possibly because their programmers
simply didn't grok XML markup.

> Is this used within the HTML DTDs ?
>  (i.e. not just in the doctypedecl)

Not explicitly.

> Is this fragment a valid HTML document ?  

Yes, perfectly.

> If not, why isn't it? Just
> which part of its definition is forbidding this fragmentary use?
> <!DOCTYPE div PUBLIC "-//W3C//DTD HTML 4.01//EN"
> "http://www.w3.org/TR/html4/strict.dtd">
> <div>
>     <p>Foo</p>
> </div>

You can test this by running it through any SGML validating parser
(eg nsgmls).

> Good tutorial refs on DTDs are also welcome. I don't know anything like
> enough on DTD innards.

The best by far is still Eve Maler and Jeanne El Andaloussi, "Developing
SGML DTDs -- from text to model to markup", Prentice Hall, 1996. You
just have to skip the bits which refer to those parts of SGML which were
dropped in the XML Specification (see the list in the FAQ on converting
DTDs to XML at http://xml.silmaril.ie/developers/dtdconv/).

But you should also bone up on Relax NG, which is a schema language with
a short (DTD-like) syntax as well as a verbose syntax, from which you
can generate DTDs, W3C Schemas, and more. This may be an easier way into
document modelling.

///Peter
-- 
XML FAQ: http://xml.silmaril.ie/