In the Postgres version of OpenACS, the acs_object_id_seq is created like so:
create sequence t_acs_object_id_seq;
create view acs_object_id_seq as
select nextval('t_acs_object_id_seq') as nextval;
which seemed to me to be a strange way to implement it. Is this a result of the automated porting from Oracle?
Dirk did some tests on this, and it's about twice as slow as just a straightforward create sequence acs_object_id_seq.
Is there a reason this was done, and why we shouldn't speed up object creation by moving over to a straightforward sequence?
(incidentally, Dirk initially thought caching sequence values would have a speed increase as well, but it turns out this is not the case. Right, Dirk?)