> the rule that intercepts 'insert' works like a charm -- and the
> primary difference is that the 'update' rule (above) calls
> stage_change() where the create rule doesn't need to -- so i
> thought that might be the culprit:
>
> db> select stage_change(0,1,1);
> stage_change
> --------------
> 0
> (1 row)
>
> so it's not. or at least it's got *me* convinced, anyhow.
Well, it's not going to be referenced by name any more, it'll be
referenced by oid which is changed when you drop and create
it again. You'd need to drop and create the rule again as well.