Re: Documenting serializable vs snapshot isolation levels

On Mon, Dec 29, 2008 at 9:28 PM, Gregory Stark <stark(at)enterprisedb(dot)com> wrote:
> "Kevin Grittner" <Kevin(dot)Grittner(at)wicourts(dot)gov> writes:
>> (3) A finer-grained approach would be to make no-effect updates to
>> rows to lock them if they are to be read for purposes of updating
>> something else in the transaction. This could have a high cost in
>> disk access and table bloat. It has the advantage of providing a
>> simple technique which, if applied consistently, doesn't require
>> knowledge of software beyond what is under development.
>
> "no-effect updates" would be just the same as SELECT FOR UPDATE

...except that SELECT FOR UPDATE won't create table bloat, or as much
I/O... I think?

> However this has the same problem that we previously discussed where someone
> can still add new records which would have changed the results of the query.

...Robert