Dear all, I'm sending the China version of my free-shape text breaker in case you wanna see what it does. If you really wanna compile and run it, then you need to install 2 other packages, but read till the end before you do it. (1) Objective Caml: http://caml.inria.fr/ It's a standard package in linux distros (at least in Debian), otherwise you can take binaries from the Download section at that site (2) ADvi: http://pauillac.inria.fr/advi/ My prototype links with the advi modules (v 1.6.0). Follow the download and build instructions at the site. (3) Dyntyp: dyntyp-src-050823-1121.tar.gz attached This is my prototype The actual algorithm is inside xbreak.ml. There is also break.ml which contains earlier version of the algorithm which doesn't really work. In order to build it, you need to install (1) and to untar and build (2). Then untar my package (3), preferably with the following directory layout: <SOME DIRECTORY> +--adk | +--advi-1.6.0 | +--camlimages-2.2 +--<SOME DIRECTORY> (dyntyp, but I should think of something better) +--src because (2) is referenced from (3) in Makefile.config and it expects adk in ../../adk. If you need different structure, change Makefile.config in src. Oops, I think, advi doesn't work on Windows, so the whole prototype is only for linux (unix) at the moment (but it will change soon). OK, so inside src (of (3)), do make, it should create dyntyp.opt. The shellscript show.sh runs a demo. Well, linking with ADvi was a quick hack for getting something visible. My next plan is to change it for a TeX plugin as it was announced. For that I need: * parsing of \showlists output, not very difficult * accessing tfm, it will take some time, I have to rewrite my NTS module * outputing TeX input So it will all take some time, but then it won't be dependent on ADvi and it will work under Windows too. So it makes more sense not to install ADvi and to wait few weeks. Cheers, --ksk