Re: [exim] Has anyone done this?

At 8:38 am +0000 2005/03/06, Matt Fretwell wrote:
>Peter Bowyer wrote:
>
>> I'm not usually quick to come to Marc's defence, but as long as he's
>> not using this as a block/noblock test but as an input to a decision
>> tree or a bayesian system, it could be quite useful.
>>
>> He says that if he finds 587,465 and 783 open, he considers this a
>> non-spam indicator. I agree. What he doesn't say is that he will
>> consider the message likely spam if those ports are not open. Which
>> would be wrong.
>
>
> The above, if that is the way it was intended, I will digress and concur
>with, (to some degree). However, with it being tagged onto his existing
>thread, the implementation scenario is the questionable part.

I am sorry to say
but logics has only one way.
I too don't usually come
to Marc defences as one.
This time I think say one might
that his statement was (almost)right.

Ok, I am not a poet, and of course his proposed negative test (as in
non-spam test) will probably add nothing to the usual criteria, but
the real problem with his test is not with its validity, it is its
non-cacheability (ouch!).

g