Re: 2822 section 3.6.4 (Message-ID generation)

["Charles Lindsey" <[e-mail]>]

In <[e-mail]> Arnt Gulbrandsen <[e-mail]> writes:


>Hi,

>2822 suggests (section 3.6.4, page 25) using IP address, system time and 
>something like a unix pid for message-ids. I'm not entirely happy with 
>that advice: It says the msg-id MUST be unique, but goes on to suggest 
>an algorithm that does not generate unique results any more.

Interestingly, we are having that same debate on Usefor at the moment (for
at least the second time this year).

Currently, Usefor says the same. The msg-id MUST be unique. Reason?
Because RFC 2822 says so, and we didn't want to be different, even though
what is demanded is technically impossible.

>The observation that "no two hosts use the same domain name or IP 
>address at the same time" is regrettably not true. Addresses like 
>192.168.100.100 are in use by countless machines, and that two of them 
>might send mail at the same time, from the same "pid", isn't 
>far-fetched. Granted, it's not something that'll happen every five 
>minutes, but the net is big and keeps growing.

But actually, that is a good reason for that MUST. It says that people who
own multiple servers with the same domain need to be aware that the
suggested algorithm is insufficient in their case (though it is fine for
most people). The "MUST" says it is _their_ problem and _their_
responsibility to fix it.

Usefor got as far as writing a draft for an informational RFC on the
subject. We may well get it out and dust it down once our main document is
out of the way.

-- 
Charles H. Lindsey ---------At Home, doing my own thing------------------------
Tel: [phone] Fax: [phone]   Web: http://www.cs.man.ac.uk/~chl
Email: [e-mail]      Snail: 5 Clerewood Ave, CHEADLE, SK8 3JU, U.K.
PGP: 2C15F1A9      Fingerprint: 73 6D C2 51 93 A0 01 E7 65 E8 64 7E 14 A4 AB A5