http://docs.python.org/library/xml.etree.elementtree.html
http://effbot.org/zone/element.htm

This is a nice library, new in Python 2.5, to save data into a tree, similar to 
DOM.

It would be nice if the txt2tags processing evolute from the current:

- read source line by line
- for each line, match blocks
- if matched:
  - inside the block lines, match inline marks
  - escape everything
  - dump tagged block
- repeat until EOF

to:

- read source line by line (or char by char)
- match blocks and save to etree
- inside the block lines, match inline marks and save to etree
- repeat until EOF
- close input file, we won't need it anymore
- now scan the etree from the start to do the escaping and tagging

In other words, separate the two process that today are mixed together:
- read, parse and close input file
- convert data to target (escape and add tags)

Than, in the middle, we can dump a XML representation of the etree to spot bug 
and find exactly how txt2tags parsed the source file. Each target could be a 
class that just knows about the etree, not txt2tags markup.

Ideas and suggestion are very welcome.