On Thu, May 13, 2021 at 1:31 PM Pól Ua Laoínecháin <linehanp@tcd.ie> wrote:
I hope that clears up what I meant - and also that what I have written above demonstrates a reasonable understanding of the difference between the potential outcomes of an RC transaction and an S one?
It does. I see now you are trying to describe snapshots of data available to statements in terms of database timestamps as opposed to referring to transactional boundaries.
I've never seen isolation levels discussed in those terms before and do not know whether it is accurate or missing some important nuances. With locks, snapshots, and valid interleaving versus invalid interleaving (for serializable) all coming into play the notion of timestamp isn't really in scope. But if it helps for basic comprehension then it does have merit and indeed the links you describe seem correct at first glance.
