Date: Mon, 4 Aug 2025 16:23:55 -0500
From: Nathan Bossart <nathandbossart@gmail.com>
To: Jacob Champion <jacob.champion@enterprisedb.com>
Cc: =?iso-8859-1?Q?=C1lvaro?= Herrera <alvherre@kurilemu.de>,
	Dave Cramer <davecramer@gmail.com>,
	PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Subject: Re: More protocol.h replacements this time into walsender.c
Message-ID: <aJEk695hVaEc4ve3@nathan>
References: 
 <CADK3HHLs-8fTiYikJBU45bkMYABVifkYPQU9s3vwO4EUcuOo=w@mail.gmail.com>
 <202508041231.myxuzx2udlze@alvherre.pgsql>
 <aJEQUTYLkS836qu5@nathan>
 <CAOYmi+nQ75D+OcyDa0cH3q5Hh_vioErtDgXHMaNGK4xfeG2EPA@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: 
 <CAOYmi+nQ75D+OcyDa0cH3q5Hh_vioErtDgXHMaNGK4xfeG2EPA@mail.gmail.com>
Archived-At: <https://www.postgresql.org/message-id/aJEk695hVaEc4ve3%40nathan>
Precedence: bulk

On Mon, Aug 04, 2025 at 02:11:26PM -0700, Jacob Champion wrote:
> On Mon, Aug 4, 2025 at 12:56 PM Nathan Bossart <nathandbossart@gmail.com> wrote:
>> The replication protocol uses many of the
>> existing PqMsg macros already, so it would be a little strange if only a
>> subset of the replication protocol messages used the special prefix.
> 
> May I ask why? These messages are legitimately different; they're
> tunneled through CopyData, so their reservations don't collide with
> the top-level codes.

Ah, I missed that finer detail.  IIUC the codes at hands are _only_ used in
these tunneled messages, in which case they belong to a distinct category.

>> There's also backups, which use the replication protocol but
>> have their own special characters [0].  If we're going the prefix route,
>> would we add another prefix for those, or use the replication one?
> 
> My vote would be to add another. 'p' is a password message in the
> top-level protocol (one of many, actually), a progress message in a
> backup stream, and a status request in a replication stream, so I
> think they deserve their own namespaces.

These also seem to use the same tunneling mechanism.  I retract my
objection.

-- 
nathan