Let me expand a little bit...
On (1) the reason why it's more efficient to use epoch is because we can save a clock scan (ec_formatted_full_date (sp?) does a [clock format [clock scan]]).
Another way to fix this without modifying the Tcl proc (which would mean the Oracle version would not need to be modified) is to modify every PG query that uses ec_formatted_full_date to become date_trunc('second', column).
However, the Oracle version is _already_ broken, so we could just fix this for PG and then later on fix it during the Oracle cleanup process. That'd make it more efficient for everybody.
