Tom Lane wrote:
>> http://candle.pha.pa.us/main/writings/pgsql/sgml/sql-set-transaction.html
>> http://candle.pha.pa.us/main/writings/pgsql/sgml/transaction-iso.html
>>
>
> It's a bit amusing that this person is dissing us for not having
> REPEATABLE READ, when what he actually seems to want is SERIALIZABLE
> (which we've had since 1999). Certainly REPEATABLE READ does *not*
> guarantee a "stable view of data during one transaction" --- see the
> discussion of phantom reads in the second link given above.
>
> regards, tom lane
>
>
Tom,
This is what the firebird guy said:
> Serializable is stricter and somehwat unusable in a multi-user, loaded> database, because only one transaction can
runat any time. Let's say
you> would have one long running serializable transaction encapsulating a> reporting query, this will cause other
transactionsto wait.>> There is a pretty good paper on discussing why it was a somewhat bad
idea to> describe transaction isolation levels in terms of phenomena in the SQL> standard. This paper also describes
transactionisolation levels for
MVCC> databases. The paper is from 1995.
http://www.cs.duke.edu/~junyang/courses/cps216-2003-spring/papers/berenson-etal-1995.pdf>> SNAPSHOT in Firebird isn't a
SQLstandard compliant REPEATBLE READ
either.> SNAPSHOT in Firebird is between REPEATABLE READ and SERIALIZABLE, but> without blocking other transactions.
Is this true? will SERIALIZABLE block all transactions on the whole
server, or just on that one connection?
Thanks,
Tony
