Ah, I found an old project I had even forgotten I made.
Tasty Search - Google Custom Search seeded with my delicious bookmarks, of course I dropped using delicious about 2 years ago when I made my own bookmarking service but probably I should add it back in to my workflow.
One of the tricky things in building the index was having sites that were too big only index their absolute urls whereas sites that were small would index their absolute urls and their relative urls. So for example if I bookmarked [http://sheldonbrown.com/diy] both that url and sheldonbrown.com itself should be indexed, whereas if I bookmarked [http://lifehacker.com/376204/nerdcycle-mounted-monitor-exercise-bike] then only that absolute url should be indexed, otherwise I would end up with a search engine with a lot of lifehacker content drowning out poor old sheldonbrown.
And if we extend the idea behind tf/idf to develop a metric for the worth of sites we would probably have to conclude that the site of sheldonbrown was likely to be more informative and important than that of lifehacker.
Of course there are some problems that an expanding web brings to that idea. Many corporate sites and blogs when I made Tasty were much smaller, thus they did not crowd the index overmuch. That has changed. Knowing what I know today I guess I would have to make a tool to periodically clean the index of the built up cruft.
Given that Google has added capabilities to its custom search over the years, maybe I should resuscitate the project.