Rod Taylor wrote:
> Are there missing NOT NULL constraints on the sl_event table?
>
> Seems that ev_timestamp and ev_type, at very least, should be NOT NULL.
>
>   
That seems to make sense, certainly for those columns.

Anywhere that events are created, a timestamp is inserted, and it
certainly doesn't make sense to allow a NULL event type...

As for other columns:

- ev_origin and ev_seqno are implicitly NOT NULL because they
participate in the primary key.

- ev_data1 thru 8 should all permit NULL because they are all optional
"argument values."

- The min/max XID fields are always populated, so probably *could* be
NOT NULL.

- ev_xip can be NULL; there may be no transactions in progress.

I think the only ones I'd be keen on modifying are the two columns you
"poked at."

Any other opinions on this???