I just thought of a truly evil, twisted thing to program. I may be evil and weird for thinking of this, but if you implement it, my hat is off to you...

Your task: Use Jakarta's custom tag libraries to interpret TeX documents into valid XHTML (CSS is kosher) that look like the documents would as postscript. Bonus points if you can do it without CSS while still having it look decently close. It's kosher to use java graphics packages to do the equations you'd have trouble doing in HTML, although if you can fake that out in CSS too, even more bonus points. If request to render time is under 20 seconds for a typical scientific paper, my hat is entirely off to you. If this takes you less than 2 months, I'll be stunned. Note that no use of JNI or other tools to talk to a real TeX interpreter is permitted. Start when ready..