On Fri, Nov 03, 2006 at 07:07:43AM +0000, Peter Bowyer wrote:
> On 03/11/06, Chris Lightfoot <chris@???> wrote:
[...]
> > Therefore there must be, in the opinion of the people
> > responsible, some feature of dynamic IPs which makes it
> > impossible that mail originating from them is desirable to
> > receive.
>
> The part of my argument which you keep dismissing as irrelevant
> answers this question in its entirety. Either listen to what you're
> being told or go away.
Your argument was a statistical one -- you said that the
ratio of rates of spam to nonspam mail arrival from IPs in
this class was some large number, which isn't the same as
the implicit claim. It might or might not be true (from a
look at my own mail, it's clear that the various `black
lists' SpamAssassin consults very rarely alter the
bayesian filter's decision, which suggests that the
conditional entropy in that information is pretty small).
You also said,
[ this class of email origination points ]
> consists of people who have other ways to send their legitimate email.
> At the very least, they have their ISP's mail relay, which is free (or
> if it isn't, change provider).
which is obviously nonsense in general -- there are lots
of ISPs which do stupid things with mail (e.g. prohibiting
email being sent unless its sender or even From: header is
in a certain domain) and it is not always true that it is
practical to change provider. Essentially here your claim
is ``in every part of the world where there exists one ISP
selling internet access there also exists a second
comparable one at about the same price, and in any case
all such ISPs are paragons of competence''. Again this is
a lovely theory but it simply isn't true.
Again I think this is basically a problem of confusing
feature extraction for machine learning, though the
`kicking away the ladder' aspect is pretty ugly. I suppose
I could appeal to the end-to-end principle but apparently
nobody believes in that any more.
--
``Fill your tank with 800 gallons of water. Start yesterday. Remember, a
medium-sized hippo takes up at least 200 gallons. (Just out of curiosity,
why do you have a hippo, anyway?)'' (USPS, how to move a hippo)
