On Wed, 12 Feb 2003 11:02:26 +0000, David Woodhouse wrote:
>On Wed, 2003-02-12 at 09:56, Ralf G. R. Bergs wrote:
>> >Using an RFC1918 IP literal wouldn't be too clever, I suppose -- but
>> >private hostnames are permitted, as long as they're unique. Isn't your
>>
>> I think that private hostnames ARE NOT permitted. As message ID needs to
have
>> a FQDN as the part after the "@," and a private hostname doesn't qualify as
a
>> FQDN (because it is, by definition, not unique.)
>>
>> >own internal domain nomenclature sufficiently unique? Is there really a
>> >chance of someone else naming their internal machines 'ralf.wg', etc.?
>>
>> Absolutely.
>
>RFC2822 §3.6.4:
> The message identifier (msg-id) itself MUST be a globally unique
> identifier for a message. The generator of the message identifier
This is basically (at least by principle) what I said. If you choose private
hostnames you can't guarantee uniqueness.
>So you're probably right in saying it's invalid because you can't
>_guarantee_ that nobody else will use it, and you MUST do that if you
>include a Message-ID.
Right.
>OTOH you can't actually _guarantee_ that nobody else is going to use
>@ADSL-Bergs.RZ.RWTH-Aachen.DE for their message-ids either, although you
>might reasonably expect them not to. So technically, you're screwed
Strictly speaking that's right. Of course I can't defend against folks
"maliciously" generating non-unique ids that look like the come from my
machine, but since we don't have a central message id registry we simply can't
guarantee that no-one tries to duplicate message ids.
>either way -- since RFC2822 says that inclusion of a Message-ID is only
>a 'SHOULD' not a 'MUST', I suppose the logical conclusion is that you
>MUST NOT include a Message-ID in any mail? :)
Hehehe, you almost sound like a mathematician. ;-)
Seriously, message ids no doubt have their raison d'etre so let's just
continue to use them. :-)
>To be realistic, the respective probabilities of someone else out there
>using _either_ '@ADSL-Bergs.RZ.RWTH-Aachen.DE' or '@ralf.wq', although
>non-equal, are _both_ sufficiently small that they're acceptable in
>practice. Especially when combined with the already vanishingly-small
>chances of the left-hand-side matching.
You're probably right but using your own domain name is a de-facto standard so
I will continue that practice. :-)
Cheers,
Ralf
--
L I N U X .~.
The Choice /V\
of a GNU /( )\
Generation ^^-^^
