On Mon, 3 May 2021 at 19:00, Peter Geoghegan <pg@bowt.ie> wrote:
>
> On Mon, May 3, 2021 at 9:45 AM Robert Haas <robertmhaas@gmail.com> wrote:
> > But if you're saying those identifiers have to be fixed-width and 48
> > (or even 64) bits, I disagree that we wish to have such a requirement
> > in perpetuity.
>
> Once you require that TID-like identifiers must point to particular
> versions (as opposed to particular logical rows), you also virtually
> require that the identifiers must always be integer-like (though not
> necessarily block-based and not necessarily 6 bytes). You've
> practically ensured that clustered index tables (and indirect indexes)
> will never be possible by accepting this.
For IoT, as far as I know, one of the constraints is that there exists
some unique constraint on the table, which also defines the ordering.
Assuming that that is the case, we can use <unique key> + <inserting
transaction id> to identify tuple versions.
With regards,
Matthias van de Meent.