Re: [PATCH] wip: Add forward-message-refs function.

On Thu, Aug 27, 2026 at 14:38 +0200, Arnt Gulbrandsen wrote:

>JFYI, this was discussed in the IETF working group during the work on 
>either 2822 or 5322, I think it may have been in 2003 or 2004 but don't 
>hold me to that.
>
>A mail thread is not necessarily a tree. Many people want it to be, 
>much software expects it to be, but that relies on cooperation by 
>everyone, and some email software doesn't cooperate.

Well, sure. Email is very old, is implemented almost everywhere, and 
has changed a lot; it's inevitably imperfect. Mutt itself allows the 
user to generate a reply to multiple messages. I think that's OK; I 
just think it should not generate References: for multiple parents.

>The conclusion at the time was twofold. ① In-Reply-To should name all 
>the messages to which this message is a direct reply. ② References 
>should contain the References of the message being replied to and 
>Message-Id of that message, if this message is a reply to a single 
>message. ③ There is no specified algorithm for constructing References 
>for a message that is a reply to several messages. I think someone 
>mentioned a MUA that can construct such replies, one with a four-letter 
>abbreviation as name, but nobody wanted to discuss this in detail or 
>standardise it. Yes, this is the third point in a two-point list, and 
>it's a no-op.

That's what I got from reading RFC 5322.