Re: computational model of transactions

Marshall wrote:
> Erwin wrote:
> >
> > AFAICT, MVCC is only safe to use in transactions that don't update
> > anything, i.e. transactions that do *nothing but reads*.
>
> Um, if we are doing nothing but reads, we don't need any concurrency
> controls at all. Everything can read everything all the time safely, if
> there are no writes going on. So it kind of seems like above you're
> saying MVCC is a no-op.

I thought Erwin meant that only reads could proceed concurrently with say a single write under MVCC, but that multiple writes could not proceed concurrently with MVCC --- there would need to be some other locking mechanism. As I noted above, I don't think that is true either, but I don't think he was saying MVCC only works on a read-only database! Received on Fri Aug 04 2006 - 20:17:46 CEST