[Spambayes] Re: [Python-Dev] Re: some preliminary timings

[Skip Montanaro]

    Alex> This highlights another recent question: do we want to stop using
    Alex> the email package in favor of our own (presumably lightweight)
    Alex> message parser?

It's just a Simple Matter of Programming. ;-)

    Alex> Personally, I don't care if the throughput of hammie is under 10
    Alex> messages per second... my mail feed isn't that dense (and I'll be
    Alex> very disurbed if it becomes so), and people who do have denser
    Alex> feeds should probably have a daemon process for filtering instead
    Alex> of firing off a new process for each message.

My original motivation in looking at this is that here at Northwestern
University the group I'm in (which manages the four large mail servers) has
been asked to look at possible *server side* mechanisms for filtering spam.
I think it's mostly politics that client side mechanisms aren't of interest,
but in part it's because client support is less than server support in this
environment.  That being the case, getting scoring time down to where we can
process a dozen or two messages per second would be desirable, even if the
available cpu power on the thousands of clients far outweighs the cpu power
available on the servers, and also ignoring issues of training.

Skip