With axml, the parsing process is driven from user-code. In more traditional SAX-like API's for XML parsing, the parser calls an implementation of an abstract interface for each node or CDATA element in the input stream. axml on the other hand is called from the user code to parse the next logical element in the input stream.
The incentive for this approach is twofold.
axml is a work in progress and not yet ready for a first release. Well, it's not even clear if the basic approach is feasible. That's why this page probable isn't very clear. I'll try to provide sample usage code as soon as possible to give an idea how working with axml will be. If you are interested feel free to mail me : bvh-cplusplus@irule.be.