I'm not quite anal enough to benchmark it properly (though a few quick checks on a 2167462 row table showed no significant difference), but I'd have said that in theory "count(*)" gives the DBMS a pretty strong hint that the client isn't interested in table content, so if anything it should be optimised better.
that reminds me of the Oracle developer I once met that used the ROWID column as a metric of table size...ie SELECT MAX (ROWID) FROM foo = number of rows in foo.
(no subject)
Date: 2006-01-19 12:14 pm (UTC)(no subject)
Date: 2006-01-19 12:29 pm (UTC)(no subject)
Date: 2006-01-19 12:31 pm (UTC)