Yahoo and their mail filters..

Suresh Ramasubramanian ops.lists at
Thu Feb 26 01:25:03 UTC 2009

On Wed, Feb 25, 2009 at 11:28 PM, Barry Shein <bzs at> wrote:

> I realize this is easier in theory than practice but I wonder how much
> better the whole AOL (et al) spam button would get if they ignored the
> spam button unless two (to pick a number) different customers clicked
> the same sender (I know, forged sender etc but something like that) as
> spam in a reasonably short amount of time like an hour or a day at
> most.

.. and you think AOL doesnt track these?  Come on, barry - try to give
large mailops shops with massive userbases some credit for clue level.
 You have all the clue in the world but you dont even begin to guess
at the firehose AOL / Yahoo / we etc have to deal with.  Or what we
routinely do, as a matter of best practice.

I wont claim perfection, infallibility etc for any of the big 3
(hotmail / yahoo / aol) or even for us (large enough - 76 million
users we filter for, 40 million of which we host).  But a user report
based spam reporting system works quite well on the aggregate.

And yes, legitimate outfits can wind up blocked (universities because
of unfiltered machines on campus, and because of nigerians / phishers
hacking user accounts, webhosts because of hacked scripts, or because
they end up hosting a high volume spammer in part of a /24 with legit
customers near him ..)

One thing that may need to be improved at one place or the other is
false positive handling - make that faster and more efficient, and
also publish the "unblock contact path"  in block messages you issue,
and you would find a lot of the gripes getting resolved.  To some
extent anyway.

Postmaster work is a place for people with decent mailops / routing
skills, yes - but far more than that, it is for people with both soft
skills for customer service plus a finely tuned b.s detector.  It is
complex, and far too long for nanog .. took maawg three or four
brainstorming sessions over a year to discuss.

And then some others relevant to this thread -


More information about the NANOG mailing list