Re: [sasl] MOGGIES Proposed Charter

[Nicolas Williams <[e-mail]>]

On Fri, May 21, 2010 at 12:38:22AM +0200, Martin Rex wrote:
> Nicolas Williams wrote:
> > I'd word that differently:
> > 
> >  * Specify an interface for enforcing security strength of GSS-API mechanisms.
> > 
> > The reason is that "reporting the security strength" of something
> > implies [to me] an absolute measure of security strength, and I don't
> > think it's possible to design a good, _stable_, absolute measure of
> > security strength.
> 
> I think that the comparable strength measured in "Bits of security"
> as used by NIST SP800-57 section 5.6.1 should be sufficiently stable.

I don't.  I'd much rather be able to ask for a credential or context
that meets a given security profile, or to check what profiles a given
credential or context meets.  The profiles should be named

Possible profile names and semantics:

 - FIPS-140-x        -> obvious semantics (FIPS-140-x compliant cipher suites)
 - <other standards> -> obvious semantics
 - Strong            -> obvious-but-local semantics
 - Fast              -> obvious-but-local semantics
 - Interoperable     -> standard, per-mechanism semantics
 - Weak              -> standard, per-mechanism semantics (basically:
                        interoperable cipher suites + obsolete suites)

 - Local-*           -> locally-defined profile names and semantics

> What changes over time is the amount of "strength" that one considers
> secure.  

Not only.  Cryptanalysis progresses and the relative strengths of
various algorithms can vary.

I abhor anything remotely like a quantification of cryptographic
strength, and will for the forseeable future.

Nico
--