I've long supported a proof-of-work concept augmented with whitelisted keys for email spam filtering. There are basically two cases here: 1. The sender is sen

I've long supported a proof-of-work concept augmented with whitelisted keys for email spam filtering. There are basically two cases here: 1. The sender is sending one email to one receiver. The sender computes the proof of work and sends it along with the message. This takes some time, but typically it can be done in the background. Waiting few seconds between sending and receiving an email is typically not an issue, and in fact many email clients wait for some time before sending anyway to allow "undo send" functionality. Rather that doing this computation on the mail server, it should be done on the sender's client, so that the server doesn't get overloaded with proof-of-work computations. 2. The sender is sending many emails to many receivers (a mailing list). In this case when the receiver signs up for the mailing list, the sender sends a request for a whitelist token to the receiver's mail server. IF the user accepts the request, the receiver's mail server returns the token and then the sender sends all further communication with the token instead of with a proof-of-work. This solves a few problems: a) Receivers opt in rather than opting out of receiving communications. The ubiquitous pre-checked "send me spam" checkbox loses its effectiveness. b) Receivers can revoke tokens at any time. c) Senders who are sending large amounts of legitimate mail don't have to compute a proof-of-work for every email they send. Receivers simply drop email which doesn't come with either a proof-of-work or a whitelist token. This drives up the cost of sending large numbers of spam emails because each spam email requires a large amount of computation. And even in cases where a spammer has a large amount of computation at their disposal (botnets are a common case) it makes it easier for servers to distinguish between mailing lists and spam: a large number of identical emails could be either, but mailing list mailings should come with a whitelist token. Doing things this way would mean we can drop these terrible DKIM and SPF systems that both fail to prevent spam and make it difficult to send legitimate mail.