On 11/29/2010 4:14 PM, Ann W. Harrison wrote:

> On 11/29/2010 4:08 PM, Dimitry Sibiryakov wrote:
>
>> Can't say about PG and MySQL, but Oracle doesn't just wait till commit, but in most
>> cases also restart the statement after that. IMHO, it is quite unpleasant process,
>> overcomplicated and AFAIK, Oracle has some bugs in this place.
>>
> With PG and InnoDB, the statement waits until the most
> recent record version becomes available and then proceeds.
> It's very simple and works well. Jim implemented it in
> Falcon and NimbusDB.

In both Falcon and NimbusDB, a select for update (either implicit or
explicit) in "write committed" mode cause a dummy update similar to an
Interbase/Firebird deleted record stub, but with a distinctive record
number indicating "record lock."

--
Jim Starkey
Founder, NimbusDB, Inc.
978 526-1376