Re: [sasl] MOGGIES Proposed Charter

[Nicolas Williams <[e-mail]>]

On Mon, May 24, 2010 at 03:07:35PM -0400, Jeffrey Hutzelman wrote:
> --On Monday, May 24, 2010 11:41:34 AM -0500 Nicolas Williams
> <[e-mail]> wrote:
> 
> >I'd like to see a strong argument for using absolute cipher suite /
> >mechanism strength numbers.  About the only argument for that that I can
> >think of is that users are used to such measures of strength, therefore
> >we should stick to what users know.  I think that's a fairly strong
> >argument for having numeric strength measures.
> 
> I don't.  [...]

I didn't say it was strong enough...  But since we've seen arguments for
numeric strength measures I'd like to see the rationales, and the above
one is the only one that I could think of.

Perhaps there's another possible rationale: a) there's some strong need
for ordered preference sets and b) we explicitly do not want local
policy.  I can't imagine (b) being true; (a), of course, is, but crypto
strength numbers needn't be the only way to get (a).

>           To the extent that users are used to numeric measures of
> crypto strength, I expect they either (1) have only seen values "0"
> and "56", or (2) are likely to compare them in unmeaningful ways
> (1024 is bigger than 256, so a product advertised as using "1024-bit
> encryption" (1024-bit RSA plus 40-bit DES) is better than a product
> advertised as using "256 bit AES encryption").

I was thinking more of 0, 56, 128 and 256.  But yes, I find this
argument relatively weak, as in weaker than my arguments for symbolic
policies instead :)

> The fact is that the "strength" of a complex authentication
> mechanism using multiple cryptographic algorithms in various ways is
> simply not a linear quantity.

My view exactly.  (Not linear, not constant, not sufficiently useful.
Policy is needed no matter what, therefore I'd like to skip straight to
that answer.)

Nico
--