Re: [saag] Additions to RFC 3631?

[Eliot Lear <[e-mail]>]

On 5/21/12 11:52 AM, Mouse wrote:
>> How would you propose to handle the situation where an MTI algorithm
>> is found to be vulnerable?
> Disable it.
>
> That is to say, I would scrap the mandatory-to-implement part, not the
> allow-disabling part.  Besides, after technology has marched on for
> five - or ten, or twenty - years, today's best is likely to look
> utterly ludicrous.  Time was, crypt(1) - a one-rotor Enigma machine,
> basically - actually provided a useful level of security.  Today, it's
> what Schneier calls `kid sister' crypto, at best.  Mandatory-to-offer
> algorithms lead to forcing people to choose between ignoring the spec
> and running grossly insecure algorithms, while allowing disabling of
> mandatory-to-implement algorithms breaks most of the point of
> mandatory-to-implement.
>

Thanks for this explanation.  I wonder if there is a more nuanced way in
which this could be handled.  For instance, would it be possible for the
IETF to keep a simple list of algorithms in order of preference, and
deal with algorithm agility in that way?  This presupposes that there is
some uniformity in application requirements, or at least classes of
applications that can be grouped...

Eliot