Hi Ryan,

If spamfilters use machine learning, like the ones at Google, Microsoft,
Yahoo, Proofpoint (and Cloudmark) then they tend to have a lot of inputs.
Including "reputation" on AS and IP which may be dependent on changes in
routing. Because that is one of the tricks that spammers use. This can
cause weird false-positives (good email being filtered). You may not even
know that you are doing something wrong, but you may still end up on the
bad side of filtering because of that.

Thing is, these ML engines are hard to "introspect" (its rules are not laid
out, but are the result of self tuning internal parameters). It decides
itself what is bad and not bad. Not even the people managing the filter may
be able to tell exactly unless they have an example. Or tune it, for that

So watch those... I would not worry about the rest.



On 6 April 2018 at 12:21, Ken O'Driscoll via mailop <mailop@mailop.org>

> On Thu, 2018-04-05 at 12:21 -0600, Ryan Harris via mailop wrote:
> > Could this cause other issues I'm not thinking of?
> I think you just need to make sure that whatever you're doing wouldn't look
> like hijacking to a (moderately intelligent) machine learning algorithm.
> And if you're keeping it all under the same AS then it probably wouldn't.
> I've never personally encountered a problem which was purely caused by
> reassigning netblocks under the same org.
> Ken.
> --
> Ken O'Driscoll / We Monitor Email
> t: +353 1 254 9400 | w: www.wemonitoremail.com
> Need to understand deliverability? Now there's a book:
> www.wemonitoremail.com/book
> _______________________________________________
> mailop mailing list
> mailop@mailop.org
> https://chilli.nosignal.org/cgi-bin/mailman/listinfo/mailop

My opinion is mine.
mailop mailing list

Reply via email to