Question

I have query regarding specifying sub-parts of an element while defining a DTD for XML. I want to have an element titled “Description”, which may have any inter-leaved sequence of a BookRef and PCDATA. I’m using following statement in my XML DTD.

<!ELEMENT Description (#PCDATA|Courseref)* >

However, I want to enforce a more strict constraint than *. I want to use +, which should mandate the having of at least one PCDATA or Courseref. However, when I use + instead of *, I get a parse error using xmllint.

I’m new to DTD and I want to know, if it is illegal as per XML DTD Specs to specify a + operator.

Answer 1

Yes, the XML spec requires that content models of the form

(#PCDATA | foo | bar | baz)*

list #PCDATA first and use * not + (or anything else) as the occurrence indicator (http://www.w3.org/TR/xml/#NT-Mixed).

A lot of design considerations played into this, most of them now of purely historical importance. One, however, may be worth noticing: if + were allowed and you did write

<!ELEMENT Description (#PCDATA|Courseref)+ >

the element declaration would define precisely the same set of valid element instances as the form using *: the token #PCDATA matches zero or more characters of parsed character data, so an element instance like <Description/> would be valid against either form of the element declaration (the zero-length string matches the content-model token #PCDATA once, so the requirement that a +-marked choice be satisfied at least once would be met).

You might convey your intent here by making Description contain

(p+ | Courseref)

and saying in the documentation that empty p (paragraph) elements are frowned upon. But DTDs do not provide a way of requiring that there be any minimum length content for a #PCDATA string. That’s one reason some people prefer to use XSD, or Schematron, or Relax NG.