However, far more importantly then bitcoin I want to know why the hell we don't have proof of work as an email spam filter yet!!!!!!
If email clients maintained the wallets you wouldn't need to even worry about the cost if you had a balance of in/out mail. You could still white list anyone you want to email for free. A cost of $0.001 per message would have very little impact on regular users. A botnet would still need to get the funds somewhere to bulk send, if they cracked the mail programs internal wallet they could drain that and send a small number of emails, but beyond that they'd need funds, and quite frankly if they could get at the wallets they'd just take the money direct.
It raises an interesting question. what is the value to a spammer of an individual email? The return from advertising/scams/phishing/whatever divided by the number of emails sent. It's got to be tiny.
I wonder about mailing lists - a couple of thousand members isn't too big a list, but $1 a message adds up. Who pays it? What keeps someone from adding fake accounts to run up the costs of lists they dislike?
It's still a deterrent and makes it non trivial to be a spammer. Also even on botnets each node is large numbers of emails, so it increases the odds that the owner of the infected computer would notice all the work being done.
Makers of specialized 'spam' hardware could potentially be targeted by law enforcement easier then dispirit spammers (especially if the hashing algorithm could be designed to be specific to email systems).
I can see some arguments as to why it's not 100% effective, but I don't think it needs to be 100% effective to add additional value. That's why I'm amazed it's still not being done. Just like I'm amazed that we still don't have easy to use email encryption, yet demand it for the web (email traffic is far easier to eavesdrop on, and often far more sensitive).
Because it's not very useful until nearly everyone supports it.
If a non-trivial amount of people don't implement it (which is highly likely when it's first released), then you have to accept non-proof-of-work-emails, or you'll miss the emails. That is the hole in the system that spam will get in through. So why bother supporting it yourself if it won't stop spam?
Secondly it is not that spam is a binary thing. You can use proof of work as another feature in your baysian system.