ietf-822
[Top] [All Lists]

Re: Intent to revive "expires" header fromdraft-ietf-mailext-new-fields-15

2008-07-29 14:03:50

Bill McQuillan <McQuilWP(_at_)pobox(_dot_)com> writes:

I believe that conflating netnews and email is very misleading here.

The difference is in the concept of "ownership."

In netnews the tradition is that the originator of a message "owns" it.
Thus the author is empowered to cancel a message that he or she has
posted and to suggest the lifetime of that message by using the
"Expires:" field.

I don't believe this is a correct characterization of how netnews works or
how the netnews protocol documents present control over message stores.
It is quite well-established in the Usenet world that the operator of a
news server has complete control over that message store, including
keeping messages for an arbitrary length of time, ignoring Expires,
rejecting all cancel messages, and so forth.  The original poster has no
control over the message within the protocol once it has left their
posting client; they can only provide advisory information and requests.

As far as Expires: is concerned, a conflict may seem to arise in the
area of retention policy when a recipient is using an employer's systems
and equipment to receive the message, but an analysis of the
"ownerships" involved will usually resolve it. But notice that the
Expires: does not come from either of the "owners" of the message.

Expires is not, either in Usenet nor in the proposal here for e-mail, a
command that must be followed.  It's additional metadata about the message
that can be used by any component in the processing flow as desired, or
ignored entirely.  It is frequently ignored on Usenet, or at least limited
within particular ranges.

There is a significant difference in this area between netnews and e-mail:
in netnews, all messages are assumed to expire at some point, so Expires
only provides input as to when that should happen.  The assumption in
e-mail is that all messages are retained forever in the absence of some
explicit action (user deletion, site-wide retention policy configuration,
etc.).

However, I don't think that difference poses any real difficulties for the
definition of Expires.  Regardless of the expectation around when the
message goes away, the existing netnews definition of the field is "this
message will no longer be useful after this date, and you therefore may
want to discard it."  This meaning is just as reasonable in e-mail as it
is in netnews.  (In netnews, it also carries the additional meaning of
"this message will be useful up until this date, so you might want to keep
it for that long rather than discarding it earlier," which it will not
have in e-mail due to different assumptions, but I don't think that's an
issue.)

In both cases, the effect of the header is entirely advisory and is
determined by the local policy of the recipient, not the sender.

The proposed meaning for e-mail and the established meaning for netnews
are so similar that the only differences I see are edge-case quibbling.  I
think it would be very strange to use a different name for the header in
e-mail when basically the same meaning is intended.

-- 
Russ Allbery (rra(_at_)stanford(_dot_)edu)             
<http://www.eyrie.org/~eagle/>

<Prev in Thread] Current Thread [Next in Thread>