To start with, here's why I'm coding torrentbarf - I've found that there are a number of problems with the most popular private bittorrent trackers out there. Among them:

• They're all written in PHP, which is scalable only if designed correctly (torrentbits scripts include a number of library files on every page load, including tracker announces, including stuff that isn't necessary).

• Many are based on the torrentbits source code, which is several years out of date and forked into oblivion (so they cannot share any code improvements, and improvements are difficult to create).

• They are poorly templated (using a hand-crafted templating system).

Torrentbarf is written using django, a Python web framework that runs using fastcgi. Not only does it run faster (the framework is optimized to run high-volume sites on cheap hardware), it is modular (and thus easier to add features to). It is immune to common errors that plague PHP apps (namely, SQL injections and register_globals).

At the moment, every relative statement in the previous paragraph is a bold-faced lie. Yes, I'm working with django, and yes, I have a pretty good idea of how trackers work, but no, it is not working yet, and thus no, it is not optimized (although if Psyco is installed on the local machine, support for it has been added to django's manage.py for this project).