On Tue, Jun 20, 2017 at 2:58 PM, Merlin Moncure <mmoncure@gmail.com> wrote:
> On Tue, Jun 20, 2017 at 2:34 PM, David G. Johnston
> <david.g.johnston@gmail.com> wrote:
>> On Tue, Jun 20, 2017 at 12:22 PM, Chapman Flack <chap@anastigmatix.net> wrote:
>>> I get the reported result (DELETE 0 and a table containing 2 and 3)
>>> in both 'read committed' and 'read uncommitted'.
>>
>> Practically speaking those are a single transaction isolation mode.
>>
>> https://www.postgresql.org/docs/10/static/transaction-iso.html
>>
>> I think Merlin has mis-read the article he linked to. The example
>> being used there never claims to be done under serialization and seems
>> to describe an example of the perils of relying on the default
>> isolation level.
>
> oops -- could be operator error :-)
yep, I made the rookie mistake of setting transaction isolation level
(which immediately evaporated since it wasn't bracketed by the
transaction), but not for the default. Sorry for the noise,
serialization failures are raised and that is acceptable behavior per
spec AIUI.
merlin