The idea is, sometimes, although an element contains ordinary-looking (as far as the XML processor can tell) text, that text needs to be interpreted in some particular way. Suppose for example, that it is PostScript, or VRML, or Base64-encoded image data. A good way to handle this situation (when you are validating) is to declare notations for these kinds of text, and then an attribute of type Notation to identify them when they occur.
Here's an example:
<!NOTATION ps PUBLIC "PostScript Level 3">
<ATTLIST Rendered In NOTATION (ps) #IMPLIED>
112 75 moveto 112 300 lineto
Back-link to specCopyright © 1998, Tim Bray. All rights reserved.