Best approach for file filters

Hello,

I'm a bit new in OOo development. My project consists in creating a kind 
of "framework" to import / edit / export adapted documents for visually 
impaired people using OOo. The file format we use is XML dtbook (see 
http://www.loc.gov/nls/z3986/v100/), a specific DTD to describe documents.

So I want first to create import and export filters for this format. I'd 
like to use XSLT because it's a language I know well and I think the 
filters will be easier to maintain and extend. We also already have 
different XSLT for this document type. I also would like to develop in 
Java and reuse my filters as a command line tool if it is possible. The 
second step of this project is to create macros to guide users in 
creating well structured adapted documents. This work will be part of 
the OpenAIP (Open Accessible Information Processing) sourceforge project 
we are currently setting up.

So my question: I have seen that there are at least 2 different 
approaches to integrate filters in OOo:
- using the OOo SDK,
- using xmerge,
And it seems that both are used, which one should I prefer ?

I had a look to writer2latex, which uses xmerge. But I was surprised to 
note that it actually relies on a patched version of xmerge. In 
addition, its author says in his roadmap that he considers giving up 
xmerge (to fix some problems and install writer2latex as an uno 
package). It make me wonder if xmerger is still my solution ...

Thanks to help me to make the best choice.

---
Benoit Guillon
Software designer
BrailleNet Association
Paris, France