Even if email worked like that, I don't think it would solve the spam problem. Spam emails are rarely personalized and the same exact message is sent out to many different addresses.

As such, the spammer would only have to store each unique message and the list of recipients associated with each unique message.

Even with single copy + list, with a poll-and-request delivery model (sender polls recipient's system, which then requests message on some schedule):

1. Sender canot simply mass-cram messages to recipients, as many spam engines today do. (Qmail for a time was notorious for doing this).

2. There's good chance that any bulk spammer will accrue a negative reputation by the time most recipients bother to request messages. So they won't.

3. Spammer must sustain a stable online presence --- not be knocked offline or blocked by ISP or hosting provider.

4. Spammer has to track successful deliveries vs. messages that are complete.

Upshot: the spammer's workfactor increases.

Hrm. I wonder if this is an actual use case for blockchain and proof of work.

If you want to send me an email, you have to send me an SMTPCoin. If my user doesn't mark it as spam, I give you the coin back. If my user marks it as spam, I keep the coin.

Preventing email spam was in fact the original use case for proof of work.

https://en.wikipedia.org/wiki/Hashcash

This was the inspirstion for nano's use of proof of work too. They don't use PoW for the chain itself but do require every transaction to include a small proof of work to avoid spamming the network, almost identical to hashcash.