From: hughes@ah.com (Eric Hughes)
If the disrupter is really motivated, he could have multiple identities and give positive ratings to his messages, so they would get through.
No one says you have to believe a particular rating.
This would imply that subscribers see the source of each rating. You would have to know that in order to judge whether to believe one or not. But I think this might consume too much bandwidth. With possibly many raters, each producing a potentially multi-dimensional rating per message, this would be a lot of stuff to send along with each message. My suggestion would be to just present the union of all the subject tags produced by the raters. This is a moderate amount of information, and to the extent that raters agree on subject tags it could in many cases be a very succinct presentation. We don't want to make this too unwieldy.
Unless someone else vouches for a message, it would not appear for a subscriber to the filtered list.
The system I want to experiment with for cypherpunks is not filtration at the mailing list server but rather filtration at the user's end. The "filtered list" is whatever passes through one's own filter. I am not talking about making toad into an extropians-style list with lots of server operations.
This makes sense, but there must still be two lists: one, the "raw" list, which is seen (at least) by raters and contains messages which have not yet been rated; and the other, the "rated" list, which has the rated messages. My suggestion was that messages which did not receive any ratings by anyone would not make it into the rated list. Obviously an alternative would be to send it out tagged to show that no one cared enough to rate it.
My suggestion is that the ratings be based on subject tags.
I suggest that one kind of rating be based on subject tags, or primary topic, or keywords, or something similar. I also suggest that other kinds of ratings exist.
Hal's suggestion is to make a rating based on salience to topic. This is fine, it allows a sheaf of related topics and concerns to be unbundled according to a particular reader's viewpoint.
This could also be used for negative ratings: subject tags such as "flame", "faq", "rant", etc. could be used to give more information than just the topic of the message. People could set up their own systems to filter the message to exclude messages with certain of these tags.
a rating message would include some message identifier
There is already the right message identifier. It appears in each piece of mail in the header field Message-Id.
Message-ID is probably OK, but it is kind of long. Many mail agents will insert an "In-Reply-To" into the header which identifies the message ID, but not all will. It would be a real pain to type one in manually. Another advantage of numbering messages sent on the "raw" list would be that people would be able to tell when they have missed messages (but that is irrelevant to the ratings issue, I admit). Hal