On Thu, May 14, 2020 at 06:36:28PM +0200, Oswald Buddenhagen wrote:
On Thu, May 14, 2020 at 07:18:24AM -0700, Kevin J. McCarthy wrote:
On Thu, May 14, 2020 at 11:51:08AM +0200, Oswald Buddenhagen wrote:
i think a reasoanble heuristic would be to just accept anything as long as it's a single string with no whitespace.

Well, it's not hard to pull the '@' check out. I was just worried we might miss the real message-id if someone started generating garbage outside of a comment:
 Mesage-ID: lkjsdflk <1...@example.com>

If you all think that's unlikely, I'll just accept the first non-comment token, and for references/in-reply-to accept all the non-comment tokens.

my implied context was that this is a fallback, so the "anything" would be used only if no message-id between angle brackets is found. that requires extending the parser state machine by a flag indicating a overridable low-quality id.

Okay, that seems reasonable.

I currently have it set to retry only for the Message-ID header. I think it's safer to leave References and In-Reply-To in the stricter mode. Dropping non-bracketed values in them won't affect your ability to reconstruct threading in the UI and reduces the chance of sending out junk from Mutt.

--
Kevin J. McCarthy
GPG Fingerprint: 8975 A9B3 3AA3 7910 385C  5308 ADEF 7684 8031 6BDA

Attachment: signature.asc
Description: PGP signature

Reply via email to