First of all, my kudos to Michael for discussing this so openly.
On 10 Jun 2016, at 12:05, Hugo Slabbert wrote:
I think everyone gets that the preferred behaviour is to reject at SMTP time, that it gets difficult/impossible to do the more tests you try and stuff into the filtering decision making, and that we don't want backscatter. But what options are there for working within those parameters while still honouring that a 25x response means that the user will have *some* indication of the message arriving, be that in their inbox, junk folder, or even just a report page/link of "these things were so horrible we did not even bother putting them in your spam folder"?
I don’t think these are “so horrible that we didn’t even put them in the spam folder”. There are obviously lightweight checks you can do before 25x. I would presume Hotmail is doing as much as they can there, because this is cheaper for them — this is the message you did not have to store, serve and eventually delete.
They then run their second tier tests, which are progressively more expensive. In some cases the messages get to the spam folder, in others the messages get silently discarded. We can only guess at the reason for the different outcomes. The point is that they can get away with silently discarding because their users do not complain about this in enough numbers, so there’s no argument for fixing this brokenness. Probably we arrived at this scenario because the anti-spam filters we’re talking about are the product of an evolutionary process years in the making, where incremental changes were used to adapt to the times — the current solution is *very* different to what was in place say, 10 years ago.
More modern implementations in use at other large providers were designed from the ground up to avoid this entirely, which is why they are more cost effective over the long run. I think this could be better leveraged as a business reason to rally the required resources to fix this “weakness”. Of course, this depends in the business model to really support this kind of thing, which I find dubious for Hotmail.
Best regards -lem
_______________________________________________ mailop mailing list mailop@mailop.org https://chilli.nosignal.org/cgi-bin/mailman/listinfo/mailop
