4 ms·
From what I've seen anti spam works almost entirely by two methods. Only accepting messages from trusted servers (which rules out having too many small servers)
by gpm 6y ago
From what I've seen anti spam works almost entirely by two methods. Only accepting messages from trusted servers (which rules out having too many small servers), and information asymmetry where the spam artists don't know what signals you are looking for (which rules out open source).
- sellyme 6y ago> information asymmetry where the spam artists don't know what signals you are looking for (which rules out open source). Machine learning mitigates this somewhat - doesn't really matter if the code's open source if even the people who wrote it have no idea what's actually happening in the black box. Of course, that method of anti-spam has an entirely separate set of problems. But we've got those either way.
- gpm 6y agoIf the model weights are available you have the same problem. If the model weights aren't available you won't get many instances of the server running.
- anticensor 6y agoWhat about open-source infrastructure with a closed-source antispam plug-in?
- gpm 6y agoClosed source probably isn't enough, you probably need to not widely distribute the binary either. At which point you can't have many server instances to federate.