Richard Wesley wrote:

>
> What about a situation where it is guaranteed that no one is writing
> to the database (e.g. an embedded analysis application)? Is there a
> performance benefit to using this sort of transaction?

No. The advantage of a read-only read-committed transaction is
that it doesn't block garbage collection by requiring the system
to maintain old versions for its snapshot of data. If you block
concurrent updates, then you're not creating any new versions.
The bookkeeping for which data you can see is very slightly
simpler in a concurrency (snapshot) transaction, so it might
be more efficient, but probably not measurably.


Regards,


Ann