Re: [PATCH] Clarify the behavior of the system when approaching XID wraparound

On Wed, Oct 4, 2023 at 8:07 AM Peter Geoghegan <pg(at)bowt(dot)ie> wrote:
> If you're willing to take over as committer here, I'll let the issue
> of backpatching go.
>
> I only ask that you note why you've not backpatched in the commit message.

Will do, but see also the last point below.

I have looked over these patches in some detail and here are my thoughts:

- I find the use of the word "generate" in error messages slightly
odd. I think it's reasonable given the existing precedent, but the
word I would have picked is "assign", which I see is what Aleksander
actually had in v1. How would people feel about changing the two
existing messages that say "database is not accepting commands that
generate new MultiXactIds to avoid wraparound data loss ..." to use
"assign" instead, and then make the new messages match that?

- I think that 0002 needs a bit of wordsmithing. I will work on that.
In particular, I don't like this sentence: "It increases downtime,
makes monitoring impossible, disables replication, bypasses safeguards
against wraparound, etc." While there's nothing untrue there, it feels
more like a sentence from a pgsql-hackers email where most people
participating in the discussion understand the general contours of the
problem already than like polished documentation that really lays
things out methodically.

- I'm somewhat inclined to have a go at restructuring these patches a
bit so that some of the documentation changes can potentially be
back-patched without back-patching the message changes. Even if we
eventually decide to back-patch everything or nothing, there are
wording adjustments spread across all 3 patches that seem somewhat
independent of the changes to the server messages. I think it would be
clearer to have one patch that is mostly about documentation wording
changes, and a second one that is about changing the server messages
and then making documentation changes that are directly dependent on
those message changes. And I might also be inclined to back-patch the
former patch as far as it makes sense to do so, while leaving the
latter one master-only.

Comments?

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