I really think you could just follow a few levels of link indirection.

My original intent, detecting a “primary source”, is a little fuzzy. It would be pretty hard, for example, to distinguish between a follow-up (with new, useful information) and a simple “me too” type of post. It would be subjective, as well.

But, really, you don’t have to do this. Just notice the link structure and remind me, “This post links to X, which you have already read.”

Oh, and I think your Kevin Bacon idea is a good one too. I’d appreciate something like, “This post is 2 links away from X, which you have already read.”

Maybe the easiest way to do this could use explicit tagging. Unfortunately, I doubt that many sites provide a useful level of tagging (if any). So relying on the pages to supply tags is probably out.

URLs could act as a unique marker indicating that several posts are the same. Three posts containing the same link to a LOLCATS! photo can safely get grouped together. Same thing for posts that link to each other.

Some sort of implicit tagging might work, but how do you decide which words are important? Include lists (e.g. we call the phrase “Unicorn Chaser” valid categorization data)? Exclude lists (e.g we say the word “the” should be ignored when looking for keywords)?

Actually, the notion of implicit tagging feels similar to search engine functionality. Perhaps some google API can indicate the degree of Kevin Bacon between two links? That would outsource a lot of the difficult magic.

I wouldn’t be surprised if somebody was working on this problem (if it doesn’t already exist). The kind of RSS feeder I want is one that also filters out all posts that aren’t both practical and heavy in informational content. That sounds more difficult to me.