On Sun, 4 May 2008, pete wrote:
Well, in my case the user's copy might be very different from the copy in the queue, because the mail procesing chain starts with spamdrop, then other stages come (for example attachment stripping, removal of the html part of mime multipart messages, rearrangement of headers, addition of more headers). I was concerned that if such a message is fed to spamdrop, then it might learn bogus things, so it is good news for me that if such a processed message is fed to spamdrop, then it will use the original copy for learning.
Yes, it's true if the processed message contains the spamdrop related email headers.
Later, i'll report back if any issue arise, and i now start a measurement of the precision.
I would really like to see your accuracy. Digitally yours, SJ. --
