On Tue, Oct 8, 2019 at 12:51 AM Alessandro Vesely via mailop < mailop@mailop.org> wrote:
> On Mon 07/Oct/2019 23:38:23 +0200 Brandon Long via mailop wrote: > > Also, it's hard to optimize for the servers that send us one message a > day. > > > If it sends a message a day, it cannot be spam (by the B in UBE). > We use the overall spam label for things that often aren't that bulk, including 419 scams and spearphishing, for example. There are specific rules for those, but the overall system still protects against them. Brandon > > > > I've argued before that we should have better handling for the smallest > > servers (whitelist the first 5 messages/day for low volume IPs, for > > example), > > > Yes, please :-) > > There don't seem to be big advantages for Google in killing small > servers. It > might more convenient to support them, for the sake of the mail ecosystem. > > > > but the total volume compared to the effort against the major spam > > campaigns, it's hard to get that high enough on the priority list. We > did > > make some changes for that for smtp time blocking, but it doesn't move > any > > of our numbers because the number of messages affected is tiny... and > when > > you're talking about IPv6, even small numbers like that can result in > large > > enough holes for spam campaigns. > > > IPv4 allows to store a single DB record per IP address, even on a PC disk. > As I stated, we have the data, it's just not useful data. And a transactional global quota system for every address is more expensive than a single disk. Given we've had unknown or low volume hosts ramp to millions of message attempts in under 30s, none of this stuff ends up being that trivial. Anyways, we have the system, but there are resource limits even for us. Brandon
_______________________________________________ mailop mailing list mailop@mailop.org https://chilli.nosignal.org/cgi-bin/mailman/listinfo/mailop
