Which messages to feed to Bayes?

[Jan-Peter Koopmann]

> The most important thing is that the messageID is preserved 
> as SA uses that as a key.  If the messageID is not preserved 
> salearn won't realise that it has already learned from the 
> message.  This means it will learn the message twice (once as 
> ham and once as spam) rather than relearning it.

Good to know. AFAIK the message-ID is preserved. But I will check.

Regards,
  JP