Bruce Momjian <pgman@candle.pha.pa.us> writes:
> Well, we had the discussion when Jan was adding TOAST, and Jan was
> saying we still need large objects for I/O purposes and for very large
> items.
Yeah. I hadn't realized until this conversation that TOAST couldn't
support partial updates of huge values. That is a definite problem for
supporting lo_read/lo_write-type operations.
regards, tom lane