Re: Theoretical Basis for SELECT FOR UPDATE
Date: 30 Sep 2005 05:18:17 -0700
Message-ID: <1128082697.360156.127320_at_g49g2000cwa.googlegroups.com>
'SELECT FOR UPDATE' is a tool to implement repeatable reads in
multiversioning databases. Those databases, due to their nature, lack
the REAPEATABLE READ isolation level. Besides, in some scenarios,
SELECT FOR UPDATE can promote the SNAPSHOT isolation level to the
'true' SERIALIZABE.
Locking databases, such as MS SQL Server, DB2 or Sybase do not need the operator because they implement RR directly.
If TTM does not like it, it's unclear what they suggest instead. The current situation with isolation levels is a practical compromise between the degree of concurrency (and therefore performance) and a possibility of various anomalies (like phantom reads, etc). Theoretically, SERIALIZABLE is the only level needed to ensure correct concurrency, but it's not practical in every case for obvious reasons. Received on Fri Sep 30 2005 - 14:18:17 CEST