> I disagree. Before something is included, somebody should make a
> case for it. I think the minimim standard should be theoretically
> sound and useful. My feeling is that this feature fails both tests.

Some days ago I suggested the implementation of read uncommited, so it
doesn't affect the Oldest Active Transaction and the transaction can be
opened for a long time not affecting garbage collection.
AFAIK read commited transactions affects the Oldest Active Transaction.

In my opinion the implementation of a new lowest isolation level is very
useful to optimize garbage collection.
The case is:
1- The transaction is very long (or can be very long)
2- The transaction does not need a great isolation level.

The practical sample is an application that waits for user input to fetch
rows.
May be this case can be solved with read_commited read transactions (this is
commented in source code but I think it's not working now).