On Sun, Sep 29, 2019 at 2:13 PM Keith Medcalf <kmedc...@dessus.com> wrote:
> On Sunday, 29 September, 2019 01:28, Gwendal Roué <gwendal.r...@gmail.com> > wrote: > >Those N reader connections allow concurrent database reads. Those "reads" > are > >generally wrapped in a deferred transaction which provides snapshot > >isolation. > > No, it provides REPEATABLE-READ isolation. There is no actual "snapshot" > taken and no snapshot exists. You are merely not seeing data written to the WAL transaction log at a > point-in-time subsequent to the point in time at which you commenced the > "repeatable-read". > I don't see where you are going with this Keith. Repeatable-reads "in the past" *are* snapshots IMHO. The WAL file *does* contain enough information combined with the main database file pages, to logically *and* physically represent a "snapshot" of the DB at that point-in-time. So not calling it a "snapshot" is a stretch at the very least. What is a "snapshot" according to you, if that's not it? And also why do you think Richard, who knows a thing or two about databases, called these API *snaphot* then? I'm genuinely curious here. I think I disagree with you, but most time I do, I'm wrong, so I'd like to understand, really. --DD _______________________________________________ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
