steveb - 1:37 am on May 21, 2006 (gmt 0)

"'determining relevence' for a simple keyword-based query is easy-peasy"

Of course. Basic relevance is mostly a 1990s concept, but that isn't the point.

"But what you certainly cannot do, is use any algorithmic logic to reliably determine that one website is NOT relevent to another."

A strawman, and a bad strawman at that. An algorithm can and should determine what is relevant, and what likely is not relevant. As always, desiring black and white answers is just not helpful. They aren't going to be perfect, and they will sometimes make poor judgments, but they should make judgments. Pages and websites as a whole that link to many other sites/pages that appear non-relevant to the content of the original page should hoist a red flag and get a scoring demerit, whereas pages that link to relevant, QUALITY content should get a scoring boost. This wouldn't be remotely controversial is there weren't yarn websites out there exchanging links with motor oil sites strictly to decieve the engines.

Relevant and related should be extremely easy for a site owner to judge with reasonable accuracy. I link to a friend of mine's wildly offtopic site because he is a friend of mine and that makes it relevant, but I don't expect a search engine to figure that out. I do expect that they judge this link an exception to the rule of me linking to high quality, relevant sites.