From time to time it occurs that two people answer a mail in the same way where one would do – closing an unblock request, for example.
When this almost happened on debian-release the other day I amused myself by dreaming up an SMTP header that would prevent such embarrassment. I wasn’t being serious in the slightest, but nevertheless X-RaceProtection was born (and it turns out at least one resident of a certain IRC channel thought I was).
X-RaceProtectioncan be a message identifier or the simple value ‘yes’ and is intended to prevent duplicate replies to, for example, mailing lists. When set as a mail ID, list software should silently drop the message being delivered if the identified message has already received a reply – that is, another message quoting that ID inIn-Reply-To. IfX-RaceProtectionis simply ‘yes’, the mail ID ofIn-Reply-Tofor the message being delivered is used, providing a shortcut.
This means you can set X-RaceProtection when replying to a mail where there is a chance of collision. If someone beat you to it, there is no embarrassment at your mail arriving with a later timestamp.
If someone fancies implementing this for smartlist/debbugs, please be my guest!
The idea makes sense in principle, but the name of the header doesn’t do a good job of conveying its meaning. You only want to use this when making the obvious reply, such as “done” or “processed” or “handled”, as opposed to any form of substantive response. Also, RFC6648 deprecated the use of the “X-” prefix. Thus, I’d suggest a header name like “ObviousReply” instead, with the same semantics.
You might also consider a couple of additional flags, space-separated from the “yes” or Message-Id. First, if either the earlier or later mail has the keyword “checksum-unquoted” in ObviousReply, the system processing the mail should only drop later mails in favor of earlier ones if the unquoted (non-‘>’-prefixed) text of both messages matches exactly, modulo whitespace and with standard signature dropping (“– “). Second, if ObviousReply includes the keyword “symmetric”, both mails must include the ObviousReply header to get dropped; that seems like a reasonable quid-pro-quo for using the header, enforcing that either party believes that they’ve written the obvious message and wouldn’t mind having their message dropped in favor of another.
Good idea, but does the X- prefix for experimental stuff not have been deprecated?