is DocBook adapted to this document generation schem ?

[Carl <multivac@altern.org>]

Hi,

I'm looking for a way to generate good looking documents from a data 
tree. After a look on google, here is my view :

My data tree will be a (java) object tree certainly mapped on a DB.

I will use LaTeX as a type setting formater, then export LaTeX files in PDF.

The process to generate document will be this one :

Data tree & XML template will be proceed to generate an XML document.
I've looked on freeMarker & Velocity to to this, any recommandations, 
other options ?

This XML document will be transformed by an XSLT to produce LaTeX file. 
the project DB2LaTeX seems to solve this part, any recommandations, 
other options ?

Then the LaTeX document will be compiled and a PDF file produced.

Now the difficult part (at least for me) :
The documents I would to produce are : bills, contracts, letters, 
administrative papers, etc. All these documents are strongly typed. So 
for each type, I will need to write a :
- adapted data tree
- XML template
- custumized XSLT to produce LaTeX file
- may be a custom LaTeX style

The last decision is : whitch DTD use for the XML template ?
Still on google, I've founded :
- DocBook
- tBook
- TEI

Is it a good option for me to use one of these or should I defined my 
own DTD for each type of documents ? I'm asking this question in terms 
of good separation layer and design pattern. Should this DTD reflet the 
form or the content ? If I use custom DTD, it will be the DTD of an XML 
representation of the data tree. What is the good way ?

Sorry if this post is a little OT but I don't know how to know if 
docBook is a good option for me.

Thanks for all direction and answers,

Carl