I didn't realize Ben was the source of that :)
I've translated outer joins all along with "not exists", but this after all was a fairly recent (6.5.*) addition to PG.
Suppose I should sit down and read the manuals more closely. The reality, though, is that the UNION constructs are going to be relatively inefficient, the "not exists" wins for those rows returned in the first clause, but all the others cost as much as doing "count(*)" since the db has to search all row candidates in order to return "false".
Still, it is a win ...
