RE: CDATA sections in W3C XML Infoset

[Gavin Thomas Nicol <gtn@ebt.com>]

> After I get the tree, I call normalize() on it. Suppose my processing
> is to count people whose first name is "Karl", in a document like
>
> <people><person><first>Karl</first><last>Käfer</last></person></people>
>
> Then I'd do
>
>   karls = 0
>   root.normalize()
>   for person in root.getElementsByTagName("person"):
>       first = person.getElementsByTagName("first")[0]
>       for n in first.childNodes:
>           if n.nodeType == Node.TEXT_NODE and n.data == "Karl":
>               break
>       else:
>           continue
>       karl += 1
>
> Now, if there are CDATA sections in there, this algorithm will break:
> the string Karl could still be split across several nodes.

As I just noted, normalize() shouldn't be guaranteed to succeed.
Even so, the example above is simplistic... if I choose to, I
can break the code above any number of ways.

CDATA sections are usually used to simplify the markup of
largish chunks of text that contain markup characters, so you'd
be less likely to find them in the above than you'd probably
than entity references.

  <?xml version="1.0" encoding="US-ASCII"?>
  <!DOCTYPE people [
  <!ENTITY ad "&#x00E4;">
  ]>
  <people>
    <person>
      <first>Karl</first>
      <last>K&ad;fer</last>
    </person>
  </people>

At the end of the day, almost every application has some number
of constraints on the actual use of XML. If that is not the case,
it's better to normalize, then process.