[Avcheck] avcache announcement

[Michael Tokarev]

Ralf Hildebrandt wrote:
> 
> On Thu, Mar 14, 2002 at 06:00:58PM +0300, Michael Tokarev wrote:
> 
> > Ralf, this may be a better idea/solution for your exact problem.  How
> > about using Message-ID field instead of checksum as a key?
> 
> The message-id can be forged -- specified by the sender.

Forged != specified by the sender.  Legitimate senders generates unique
Message-IDs -- even br0ken outgluck.

> I don't know how qmail generates (if at all!) message-id's for mail
> that was intended to be multi-recipient mail.

Qmail will insert message-id once before sending mail out.  If at all.
Well, this is interesting point -- qmail may NOT insert missing msgid
at all.  But for most MUAs, message-id generated at MUA level, not
by MTA.  I mean MUAs used by "regular" posters, not by some robot etc.

> It should be a checksum of the mailbody. Fefe's opinion was that there
> could be collisions (which is true), but given the fact that the cache
> is very short lived (10 minutes), this should not be a problem.

It is very difficult to forge such short-lived msgid.  Well yes, one
may send a clean message, triggering "clean state on" for this msgid,
and short after that, send another with the same msgid but infected.
Alternatively, once someone received message from fefe user, he may
send back infected message with the same msgid.  The question is --
WHY this may be needed?

Regards,
 Michael.