At 04:14 AM 01-02-2002 +0000, Thomas Lockhart wrote:
>...
>> The oid correctly reflects the order of the insertion of the rows...
>> but look at the timestamp - the last row has a timestamp _2 minutes
>> before_ the previous row. How could this be happening? We know row
>> 69719 was inserted _after_ 69718, by probably about 30 seconds.
>
>The timestamp provided as a result of evaluating 'now' is the time of
>the start of the transaction, not the instantaneous wall clock time (if
>you want the latter there is a function to provide it).
>
>So, the times will reflect the time the transaction was started, while
>the OID will reflect the order in which the insert/update actually
>happened within the transaction.
Do postgresql backends still preallocate ranges of OIDs?
Link.