Peter Eisentraut wrote:
> On ons, 2011-04-06 at 11:01 -0700, Jeff Davis wrote:
>> On Tue, 2011-04-05 at 14:57 +0000, Roy Hann wrote:
>> > Nikolas Everett wrote:
>> >
>> > > I think it is saying something that one of the most exciting features in 9.1
>> > > would be lost on a huge portion of the DB using public. I've interviewed
>> > > DBAs who don't know what a transaction isolation level is. Most of the
>> > > developers I've interviewed don't know.
>> >
>> > A DBA doesn't need to know what an isolation level is. Programmers need
>> > to know. (But they don't.)
>>
>> One of the use cases of SSI is to enforce constraints, which are
>> certainly of interest to DBAs.
>
> Well, but users can freely change the isolation level, so it would not
> really be an effective constraint mechanism.
We're getting O/T, but...
The checking of constraints is expected to be done using
serializable isolation regardless of the isolation level of the
transation that triggered it. See section 4.35.4 of the latest SQL
standard:
"Regardless of the isolation level of the SQL-transaction, phenomena
P1, P2, and P3 shall not occur during the implied reading of schema
definitions performed on behalf of executing an SQL-statement, the
checking of integrity constraints, and the execution of referential actions
associated with referential constraints."
A friend pointed this section out to me just a couple of days ago.
--
Roy