On the other hand... nobody's really got decentralised spam prevention working, and the only really effective anti-spam systems that I know of rely on hidden data, which implies centralisation.

The only serious ways of dealing with it are pay-per-use, which disproportionately affects certain subsets of the population, or web-of-trust, which nobody's got working for reputation on a grand scale yet.

I created https://hashcash.io/ to try avoid spam by forcing bots and users to "pay" with CPU cycles.

How well does such a solution scale up? You need to keep the requirements low enough that it runs on mobile CPUs, but if it becomes widespread enough, doesn't it make sense for a bot farm to pick up a Bitcoin mining ASIC to grind out the hashes for them?

so far it work great. in future issues might arise, but then i can always tweak hashing algo or switch blockchain. So these ASICs, created purposely to crunch hashes for my service will be obsolete the moment i tweak it a bit. So it have to be software...

As for mobile/desktop/etc - I would expect each community to have their own main audience to which site owner can tweak `complexity` parameter. And in V2 work will be happening in background while you browsing site, so when time come to post comment - enough work already will be done. Hope this make some sense :)

Bitmessage uses proof-of-work to prevent spam.

"When you send a message, your client must first compute a Proof of Work (POW). This POW helps mitigate spam on the network. Nodes and other clients will not process your message if it does not show sufficient POW. After the POW is complete, your message is shared to all of your connections which in turn share it with all of their connections."

https://bitmessage.org/wiki/Proof_of_work

https://bitmessage.org/wiki/FAQ#How_does_Bitmessage_work

So what does a message actually cost in terms of dollars, power, time? How does it prevent low volume spam?

And that may prevent botspam, although I'm skeptical. But I seriously doubt it will have an effect on abuse (people stalking others, posting harrassing Tweets, etc.)

This. The project homepage literally claims (as a good thing) that "...no one can censor you. No one can remove your posts. Your account cannot be blocked."

So not only does this not help with abusive behavior, it states that abuse will live forever and be impossible to block or filter.

/me shudders

Almost like the Internet. So spooky

Rather different, actually. If I decide your inbound SMTP mail is spam, I can block it. Ditto for blog comments, XMPP contact requests, or really most any protocol in use on the public Internet b/c of course there will be griefers and spammers and all kinds of bad out there and we need tools to filter and protect against it.

You have the right to say anything you want (though not without consequences). Conversely, I should have the right to literally not see/hear it once I've decided it is causing me harm.

Claiming that your platform/protocol makes this impossible just sounds naive at best and nefarious at worst to me.

Easy. You just get message from the people you "follow". So if you want to send a message to someone, you must make them add you to their "list" first.

Thunderbird's spam filter works pretty well, doesn't it? It's completely decentralized.

I do agree it's a difficult problem though.

Shared/public/open source warning lists are distributed intelligence, letting end-users pick their own feed.

Web of trust.