Re: [ietf-822] Expires header field

[Russ Allbery <[e-mail]>]

"John Levine" <[e-mail]> writes:
> It appears that Michael Richardson  <[e-mail]> said:

>> A place where it would be useful is when dealing with reports from cron
>> jobs or other monitoring systems where one really would like a frequent
>> report of a fault, but don't need to keep all thousand previous ones.
>> A supersedes would also help, but meh.

> Usenet has both.  Expires is widely implemented since usenet expires
> everything anyway, and this just provides advice to adjust the
> expiration date.  Supersedes has incredible security problems so I don't
> know anyone who implements it.

INN implements it, but as with unauthenticated cancels it's frequently
disabled by the local administrator.  (This may be what you meant.)

Supersedes is not a good idea without message authentication, a nut that
we've never been able to truly crack for either Netnews or email.  The
latest attempt in Netnews is RFC 8315, which is specified for Supersedes
as well as cancels.  It's probably too early to tell if this will catch on
and work in practice.

Expires is less of a challenge since one can just reject nonsensical
values and it otherwise doesn't matter that much.

(That said, a lot of modern news servers use cyclic buffers for storing
news instead of files in the file system, and INN's implementation of that
concept only supports shortening the expiration, not lengthening it.)

-- 
Russ Allbery ([e-mail])             <https://www.eyrie.org/~eagle/>