Anyone knows why the PostgreSQLDialect uses sequences instead of PostgreSQL "serial" or "bigserial" data type for the native id mapping ?

PostgreSQL 8 supports identity generation via serial and bigserial data types, and there is no reason to use sequences when mapping id as "native".

The documentation states that native picks identity, sequence or hilo depending on the capability of the DB. Since PostgreSQL 8 is able to handle identity, why is sequence used instead ?

Thanks !

Well sounds like it also supports sequence, no?

Sequences, in general, are preferred since the inserts can be delayed until flush time. That is not the case with identity.