ietf-mxcomp
[Top] [All Lists]

Re: The real reason MARID won't work, RE: testing of MARID proposals

2004-08-06 20:30:32

"Murray S. Kucherawy" <msk(_at_)sendmail(_dot_)com> wrote:
In fact I have heard (admittedly, not seen) evidence to the contrary,
namely that even the large ISPs receive the vast majority of their e-mail
from only a few domains.

  The following stats are for how real email is distributed across
source IP's.  It's not what you're looking for, but it's indicative:

    http://www.striker.ottawa.on.ca/~aland/spam/sent.png

    http://www.striker.ottawa.on.ca/~aland/spam/sent-logscale.png

  That's more interesting.  log-log across 4 orders of magnitude.

    http://www.striker.ottawa.on.ca/~aland/spam/cumulative.png

  Any successful system getting deployed to the *right* 70k IP's will
touch 90% of the non-spam mail.

 It follows that the DNS impact would quite possibly not be as
severe as we all assume.

  The data I posted above is for *all* recipient domains in the data
set.  My bet is that for most recipient domains, incoming messages are
distributed exponentially across sending IP's.

  While the second doesn't follow from the first, it is often true.
Also, DNS caching wil probably mean that non-cached lookups will most
likely depend on the number of unique source IP's, and not on the
message volume.

   Alan DeKok.


<Prev in Thread] Current Thread [Next in Thread>