Sorry, Richard, that is not "MUST"; it's "SHOULD".  We use MUST when
things won't interoperate otherwise or when not doing it actually
harms security.  "To help the sender" merits SHOULD.

Barry

On Mon, Jul 21, 2025 at 12:01 AM Richard Clayton <[email protected]> wrote:
>
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> In message <[email protected]
> il.com>, Barry Leiba <[email protected]> writes
>
> >> I think what we want is:
> >>
> >> The verifier MUST support at least one of the signature algorithms.
> >> The verifier MUST check all the algorithms it supports.
> >> The signature MUST be valid for all signatures.
> >
> >I think this is closer to right, but...
> >
> >> The verifier MUST check all the algorithms it supports.
> >
> >Why?
>
> to assist the (purported) sender ...  the reason for having a scheme for
> adding a new algorithm is the notion that at some point down the line
> one of the existing algorithms will be felt to be insecure.
>
> the scenario envisaged is not that an algorithm is broken in such a way
> that anyone can forge it and no one can trust it -- but that it can be
> broken at great expense by specialist devices
>
> in such a world continuing to use the (slightly-)broken algorithm may
> make sense for a while because not all verifiers will have implemented
> the new, replacement, algorithm and a sender will wish to use the new
> one (for those who understand it) alongside the old one (for those who
> have yet to catch up)
>
> yes the sender is taking a risk, but they may prefer verifiers using the
> old algorithm to considering their email not to be signed at all
>
> hence what is essential is verifiers checking every signature they
> understand -- and then rejecting the email if ANY signature that they
> understand fails
>
> >Perhaps I want to retain support for algorithm Q in case I get
> >messages with it, but I'm really done with it and prefer algorithm T.
> >What I want to do is only check Q if that's the best I have.  And
> >perhaps a sender is sending Q to support verifiers that haven't added
> >support for T yet, but they are also sending T.
>
> I am expressing it as "understand" not in terms of "best" ... we may end
> up in a world where post-quantum algorithm has weaknesses to assuming
> that you can order algorithms is not necessarily going to be the case
>
> >What value is there either to me or the sender to tell me I MUST check
> >the Q sig?  How does it harm anything if I just check the one with T?
>
> the harm is that if the sender has provided two signatures AND you know
> how to check both of them AND one fails then there is a sophisticated
> attack occurring and you should refuse the email
>
> now some people seem to think that during a changeover period people may
> screw up and make a mess of the new algorithm and all their signatures
> fail -- best then in my view to reject their mail. It will concentrate
> their minds wonderfully.
>
> They should not be experimenting with their new code by sending email to
> strangers -- they should be debugging their system and sending email to
> friends who can tell them if they can interwork ..
>
> >What's wrong with something like this:
> >The verifier MUST support at least one of the signature algorithms.
>
> no ... we want people to support RSA (because that's the type of key
> that most people have at the moment) AND we want people to support
> elliptic curve (because that is widely believed to be a better way of
> signing at present) ....
>
> THEN when someone writes an RFC for a new post-quantum scheme they
> should write MUST in that RFC if they feel that is appropriate at the
> time; they may feel SHOULD is better if the existing algorithms are
> still working fine at that time, I do not prejudge that.
>
> >The verifier SHOULD check all the algorithms it supports.
>
> no ... MUST --- as explained above, to help the sender
>
> >The signature MUST be valid for all signatures that are checked.
>
> yes, absolutely
>
> >...and we add an explanation for the SHOULD.
>
> an explanation is certainly appropriate -- that explanation may also say
> that by local policy you can accept anything you like (your server your
> rules) but equally that's not necessarily going to be very neighbourly
> of you !
>
> - --
> richard @ highwayman . com                       "Nothing seems the same
>                           Still you never see the change from day to day
>                                 And no-one notices the customs slip away"
>
> -----BEGIN PGP SIGNATURE-----
> Version: PGPsdk version 1.7.1
>
> iQA/AwUBaH1nEmHfC/FfW545EQIINQCfat4ekqD3i72kfOfYhwhCsGGnwWUAoLMA
> gewwetdHgs6n/k4kl2PeHG/s
> =2Z5l
> -----END PGP SIGNATURE-----
>
> _______________________________________________
> Ietf-dkim mailing list -- [email protected]
> To unsubscribe send an email to [email protected]

_______________________________________________
Ietf-dkim mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to