[Cfrg] Re: universal MACs

[RJ Atkinson <rja at extremenetworks.com>]

David,

My protocol preference in an IETF context is for algorithm-independent 
protocols,
so that one does not need to change the protocol to add support for 
additional
algorithms in future.  This property holds for at least OSPFv2 
authentication and
RIPv2 authentication (and yes, I am working on documentation updates to 
each
of those specifications to document how SHA-1 and its relatives are 
implemented).

My algorithm preference in an IETF context is for algorithms that are 
acceptable
to the largest number of end users.  Right now, many governments (from 
multiple
governments, including several in Europe and several in Asia, NOT just 
the US)
and many non-governmental customers are insisting upon the same 
algorithm set,
which is the set of currently NIST-approved algorithms and modes.  This 
set of
algorithms appears to be the preferred set of algorithms and modes 
right now --
by a VERY wide margin.

In practice, that means there is a strong buyer preference from many 
countries
for SHA-1 (and its NIST-documented relatives) and for AES (in 
NIST-acceptable
modes).  Several countries, including several in Europe and several in 
Asia, are
also insisting that cryptographic products they buy have NIST FIPS 
140-2 approvals.
This also tends to reinforce the selection of algorithms and modes 
since only the
NIST-approved algorithms and modes can get a FIPS 140-2 approval.

I would even suggest that for the moment, any new security protocols 
developed
in an IETF context ought to include documentation on how NIST-approved 
algorithms
and modes are used with those protocols -- even if the IETF were to 
decide that
it preferred some non-NIST algorithm or mode for a particular security 
protocol
as the "default" algorithm/mode choice.

Yours,

Ran Atkinson
rja at extremenetworks.com

PS:  Nitpickers should please note well that all of my comments above 
are about
the end-user/marketplace desires, and there is no comment above about 
the quality
of any particular algorithm or mode.


On Oct 28, 2004, at 07:51, David A. McGrew wrote:
>     I suspect that the group preference would be towards hash 
> functions that are portable (e.g. fast on a wide variety of CPUs) and 
> that require minimal per-key state.  There may be other points of 
> view; others please chime in if you have other priorities.
>
> David


_______________________________________________
Cfrg mailing list
Cfrg at ietf.org
https://www1.ietf.org/mailman/listinfo/cfrg