I'm building yet another keyword/bookmarking application and there's
one kind of query that's eluding me. For brevity, let's say I have a
table with three fields... 'tag' and 'site' and 'id', There can be
duplicate entries for either 'tag' and 'site', but 'id' would be a
unique autoincrement value.

I'd like to have a query that extracts the last ten records, BUT with
the restriction that there can be no more duplicated values for
'tag'.

In use, I would be 'bookmarking' sites as I surfed the web, and in
practice I usually am interested in one particular topic for a while,
so in a day where I might bookmark 50 or 100 sites, the 'tag' values
will go into the database in bunches. So a query for say, the last ten
records, they might all have the same tag. I want to get the last ten
tags from the query.

Have I stated the problem clearly enough? I'm thinking it's probably
a "select distinct tag..." ??