Hi,

I'm new to parsing/using xml but this project seemed reasonable to cut
my teeth on.  I have a few dozen "articles" that are local
announcements of interest for my group's customers.  They have a
simple format of a title, zero or more static or hyper-linked images
and one or more paragraphs of text.

A "Title" will hold plain text.  The "Text"s will hold plain or mixed
content.  The "Image"s will need to know about the hyper-link URL (if
any); the image source URL and possibly "height" and "width"
attributes.

I have made a stab at making a DTD

<!ELEMENT article (title, image*, text+)>
<!ELEMENT title   (#PCDATA)>
<!ELEMENT image   (src, width?, height?, link?)>
    <!ATTLIST src    CDATA #REQUIRED>
    <!ATTLIST link   CDATA #IMPLIED>
    <!ATTLIST width  PCDATA #IMPLIED>
    <!ATTLIST height PCDATA #IMPLIED>
<!ELEMENT text    (#CDATA)>

A sample xml doc looks like

<?xml version="1.0" ?>
<!DOCTYPE article SYSTEM "http://www.itg.uiuc.edu/publications/news/
news.dtd">
<article>
  <title> Applied Physics Letters Features ITG Image on Cover </title>
  <image link="http://scitation.aip.org/dbt/dbt.jsp?
KEY=APPLAB&Volume=90&Issue=21"
    src="/images/apl_cover-130.jpg" />
  <text> The cover for the <a href="http://scitation.aip.org/dbt/
dbt.jsp?KEY=APPLAB&Volume=90&Issue=21">May
21, 2007 edition of Applied Physics Letters</a>features an image
produced in the ...
</text>
</article>

Now, I have spent time searching this group and a couple others
related to the scripting language and the XML parser i am using.  I
*know* what my problem is... what i don't know is why I have it.

My XML parser chokes on the first "&" (ampersand) in the "link"
attribute of the "image" tag.  I know that being "well-formed" means
the amps should be "quoted" but I thought that the "CDATA bits in the
DTD meant that *ALL* characters are accepted in this context.

Is my DTD wrong for the xml I have?  Is my parser/validator not
picking up on the DTD?

I know that I can pre-process the incoming xml file and change the
amps to the html entity version but that feels wastefull if CDATA is
doing what i thought it should do.

other than a clue :), what am I missing?