Also I think it's a good idea to put a Bayesian spam filter in Friendica sooner rather than later. We're much better positioned to deal with unwanted posts than some of our federation partners, but it's much better to have this in place and not be used than not to have it when it's needed.

This isn't difficult, however I'm faced with a choice and wanted to take a show of hands. Whether the spam filter training data should be global for a site, or if it is per-person.

If it is global, it will work instantly for new members without requiring their own corpus of spam decisions. It will also catch a post that several other people on the site flag.

The downside of a global filter is somebody who has legitimate conversations about penis enlargment and Rolex watches could be seriously inconvenienced. Such people do exist. It's also likely that MafiaWars posts won't be visible in anybody's stream.

I'm leaning towards personal so that you can tailor it to your own likes/dislikes, but the obvious benefits of a global filter make it a tough call.