Regarding my patch, Theo Van Dinter wrote:
>you change the expiry algorithm from LRU to FIFO, which is
>very likely to cause checks to be less accurate.

(For the acronym-impaired: those are Least Recently Used and
First In, First Out.)

Does sa-learn not update the atime on the Bayes tokens that it
modifies?  My full procedure includes using sa-learn on all
non-whitelisted mail, so I would expect that the LRU aspect would be
preserved.

                          :: Jeff Makey
                             [EMAIL PROTECTED]

Reply via email to