Re: CDATA by any other name... (was The raw and the cooked)

Richard Tobin (richard@cogsci.ed.ac.uk)
Mon, 2 Nov 1998 16:28:30 GMT

One reason to regard a CDATA section as equivalent to the characters
in it is that it is prefectly reasonable for a processor to transform
a CDATA section into plain character data with character entities
where required. A processor that outputs canonical XML will do this.

Suppose we regard <![CDATA[ ]]> as invalid in element-only content.
If the processor is non-validating, it will not check this, and
will produce valid output from invalid input.

You might for example run a document through such a processor merely
to change its character encoding. It would be unfortunate if this
process changed the document's validity.

Slightly less plausibly, a processor might decide to output all character
data as CDATA sections to avoid using character entities. This process
could make a valid document invalid.

-- Richard