I support in principle some sort of experiment, since it seems to be
agreed that it wouldn't have any practical effects: we'd just be trying
it out to see for ourselves whether the trust metrics will really end up
corresponding to what we think are reasonable values for various people.
I think the simple "yes/no" scheme is just a little too simplistic, but
I agree that too complex is not a good idea. The main problem is that
it simply allows lots of new accounts to add trust. I think instead we
should use one of the "trust flow" type systems, where getting a
thumbs-up from a well-respected contributor helps out your rating a lot
more than getting a thumbs-up from someone with no reputation does.
This could be jump-started by giving Jimbo an initial level of respect
and then having it flow out to whoever he approves of, and etc. Lots of
details to work out, of course.
-Mark