[88513] in North American Network Operators' Group
Re: ml hacks for goodmail
daemon@ATHENA.MIT.EDU (Simon Waters)
Wed Feb 8 05:26:27 2006
From: Simon Waters <simonw@zynet.net>
To: nanog@nanog.org
Date: Wed, 8 Feb 2006 10:25:52 +0000
In-Reply-To: <878xsmg6ux.fsf@mid.deneb.enyo.de>
Errors-To: owner-nanog@merit.edu
On Tuesday 07 Feb 2006 22:08, Florian Weimer wrote:
>
> As far as I can tell, the filters at AOL are far less problematic than
> crude filters at smaller sites which simply use SORBS or
> bl.spamcop.net.
Not here, no one cares if some small bit player has stupid filters, but when a
significant volume of your email goes somewhere stupid filters hurt, queues
build, users complain, and we are a bit player in the email world.
We have a regular email to a customer rejected weekly by AOL because it
contains a "banned URL". Wouldn't be so bad, but it contains web referer
stats, so is nothing but URLs. We've no idea which URL it is, and I'm not
doing a binary filter approach to work around their broken filters.
Simplistic content only based rejection of email is just a broken model, as is
using end-user input in too simplistic a fashion (end users make too many
mistakes), AOL do both. I manage to filter all my personal email with no
content inspection over and above "no Windows executable attachments here -
thank you", no end user interaction, no silly places where falsely classified
email stagnates, it really isn't difficult to deploy filters like this.
But I thought the whole thing looked like a marketing campaign for Goodmail,
and nothing more.
