Things still to do, in order of priority:

  20050627: fix problem resulting in zero tokens with some messages
  20050227: autotrain option "-u", retrains based on classification
  20050624: try stripping double/triple letters, eg "fffinancce" -> "finance"
  20050308: token for nonsense META tags (META NAME="blah blah blah")
  20050226: support MH/Maildir training folders
  20050227: environment variable for -d and -g
  20050228: more verbosity, with profiling data
  20040320: comma-separated training folders (-T spam1,spam2,spam3 nonspam1,2,3)
  20050322: generate a "%age new tokens" score (eg 90% new tokens)
  20050201: support for SQLite v3
  20050527: look at http://plg.uwaterloo.ca/~trlynam/spamjig/

Any assistance would be appreciated.
