> If I'm storing data that may contain encoded versions of <>&" and ', do 
> I need to store that data in CDATA sections or am I misunderstanding 
> the role of CDATA?

I'm not sure what you mean, but don't (try to) use entities _amd_ CDATA.
that is you can do
<foo> a &lt; b </foo>
or equivalently
<foo><![CDATA[ a < b ]]></foo>

in either of those forms the content of the element is
the 7 characters " a < b " if you do

<foo><![CDATA[ a &lt; b ]]></foo>

The &lt; is not an entity refertence (as & is not a  markup character
in a CDATA section) so it is the 10 characters " a &lt; b "

The answer to the question in the subjext line is "nothing" you _never_
need to use CDATA. The only thing CDATA does is to make < act like &lt;
and & act like &amp; within its scope so if you have a section which
would require a lot of these characters to be escaped it's often more
convenient to use CDATA, but it never _need_ be use, you can always
remove the CDATA section markup and replace & and < by the entities, and
get an equivalent document.

David


________________________________________________________________________
This e-mail has been scanned for all viruses by Star. The
service is powered by MessageLabs. For more information on a proactive
anti-virus service working around the clock, around the globe, visit:
http://www.star.net.uk
________________________________________________________________________