Hello,

A new version of Xmlm is available. It is incompatible with the
previous one. The callback interface was dropped in favour of a
streaming (pull) interface. The tree interface was dropped in
favour of a custom tree construction interface tightly integrated
with the streaming interface. This has the following advantages :

* The input and output processes become symmetric.
* Reclaim your recursion ! There no inversion of control, the client  
drives
   the parsing process.
* Makes it easy to create tag level parsers and use parser combinator
   approaches.
* Allows to use the custom tree api on specific parts of the
   document only.

You can get a feel of the new interface by looking at the examples at
http://erratique.ch/software/xmlm/doc/Xmlm#ex

Other significant changes are :

* Namespace support, all names are now expanded names (attributes  
pertaining
   to namespaces are preserved in the result).
* Whitespace stripping respects the xml:space attribute.
* Xmlm is functorized on string and internal buffer types. Among
   other things this can be used to perform hash-consing, to
   overcome caml string limitations or to process the character
   stream, e.g. to normalize unicode characters or to convert to a
   custom encoding.
* UTF-8 documents can start with an UTF-8 encoded BOM.
* The file test/xhtml.ml couples each XHTML character entity with
   its corresponding UTF-8 encoded character string. You can use
   it to program a suitable entity callback to parse xhtml.

(New) project homepage :  http://erratique.ch/software/xmlm

As always, your (possibly negative yet constructive) feedback is  
welcome,

Daniel