Re: [Proposal] Fully WAL logged CREATE DATABASE - No Checkpoints

On Fri, Feb 11, 2022 at 12:11 PM Andrew Dunstan <andrew(at)dunslane(dot)net> wrote:
> The last one at least has the advantage that it doesn't invent yet
> another keyword.

We don't need a new keyword for this as long as it lexes as one token,
because createdb_opt_name accepts IDENT. So I think we should focus on
trying to come up with something that is as clear as we know how to
make it.

What I find difficult about doing that is that this is all a bunch of
technical details that users may have difficulty understanding. If we
say WAL_LOG or WAL_LOG_DATA, a reasonably but not incredibly
well-informed user will assume that skipping WAL is not really an
option. If we say CHECKPOINT, a reasonably but not incredibly
well-informed user will presume they don't want one (I think).
CHECKPOINT also seems like it's naming the switch by the unwanted side
effect, which doesn't seem too flattering to the existing method.

How about something like LOG_AS_CLONE? That makes it clear, I hope,
that we're logging it a different way, but that method of logging it
is different in each case. You'd still have to read the documentation
to find out what it really means, but at least it seems like it points
you more in the right direction. To me, anyway.

> I can live with the new method being the default. I'm sure it would be
> highlighted in the release notes too.

That would make sense.

--
Robert Haas
EDB: http://www.enterprisedb.com