[Barry Warsaw]
On Sun, 2003-09-28 at 05:13, Harald Meland wrote:
- Whenever Mailman receives a message whose message-id is already present in the archives, the original Message-Id: header is renamed to e.g. X-Original-Message-Id:, and Mailman generates a fresh (as in "not yet present in the archives") message-id before the message is either archived or sent to the list members.
This is what I was thinking about. Alternatively we could rewrite all message-id headers when we accept the message. That would guarantee uniqueness but it would break the correlation of message-ids between list copies and direct copies. Is that bad?
I don't think the RFCs speak clearly of this either way; however, it would break things for people who use message-id-based techniques to correlate received duplicates.
On the other hand, such message-id-based techniques are IMHO workarounds for the ((still) very common) problem of people/programs not respecting the various (more-or-less standard) headers for directing where replies should go.
The less aggressive approach would surely be to only generate new Message-Id:s for messages that already exist in a list's archive.
(note that we already do this for NNTP posted messages, and there has been some off-list discussion about that).
The Message-Id: field is (very) much more significant in NNTP than it is in SMTP.
Harald