On Fri, Dec 9, 2011 at 18:58, Earl Hood <address@hidden>
On Fri, Dec 9, 2011 at 5:25 PM, Jeffrey Honig <address@hidden> wrote:
> I have my fetchmail scripts ignore messages with a duplicate messageid
> (using a history of messageids I've received during the last 30 days or so)
> to avoid duplicate mail.
Depends on what you mean by "duplicate". From a mail archiving perspective,
you can have multiple messages with same message ID, but can be considered
different messages, even if the "content" is the same.
Example: Mail that is sent to individuals and mailing list(s). Mailing
lists typically add additional headers (e.g. List-*) and custom footers,
so the message that is received by list subscribers is not identical
to versions received directly from the original sender.
If the content is the same, they are the same message and would be duplicate. A message identifier identifies exactly one instantiation of a message. The changes you describe must not change the the content of the message, or a new Message-ID would need to be generated (see RFC2822 3.6.4).
When I send a message to a mailing list, I could get it back with some more headers and footer, or I might not get it back because I do not subscribe to that mailing list, or do not receive copies of my own posts. I believe that the list of where it has been and what ads the mailing list host adds are irrelevent.. Someone may redist said message to another mailing list, or it may go through a gateway to a news group and back to another mailing list. I wouldn't even see that information so do not feel that archiving duplicate copies of a message with different transport information is necessary.
There are valid reasons to keep messages with duplicate IDs, such as your example below.
My point is that we should allow for duplicate Message-Ids if we implemented a Message-Id index.
> It is probably too much to assume that everyone would not have duplicate
> copies of a message with the same messageid. Some people may have valid
> reasons for this.
Archiving messages to mailing lists is a valid reason.
> And there may be messages w/o a message-id.
This is the case for draft messages in nmh.
As an anecdote, I was not seeing some e-mail replies from my boss. I tracked it down to the Good (tm) e-mail client he was using. It would not generate a new message-id on a Reply. Thankfully they fixed it in a later update before I missed something important.
Jeffrey C. Honig <address@hidden
GnuPG ID:14E29E13 <http://www.honig.net/jch/key.shtml