Design approach to MASS (was Re: [ietf-dkim] On per-user-keying)

[Earl Hood]

On August 10, 2005 at 07:32, Dave Crocker wrote:

> >  DKIM violates basic software design principles.
> 
> As noted, DKIM is a protocol specification, not a software design.  

Same engineering patterns apply to both.

> >  For example, computing a cryptographic hash of mail message data (includin > g
> >  canonicalization methods) in itself is a useful capability.
> 
> Since DKIM has a number of parametric components, including canonicalization 
> and 
> signature algorithm choices, I do not understand what additional factoring yo
> u 
> are concerned about.  

Why should the generation of a message-header-based signature be
tied to a key management system?  The cryptographic standards do not
do this.  The method for creating digital signatures is independent
of any key management system.

RFC-1847 is good example of what I am talking about, and I
think any header-based signature effort could follow the same
pattern.

> It sounds as if the main concern is about splitting things into separate 
> documents, rather than changing the architecture or specification.

Splitting things up is a nicety, but not essential (as I noted in my
previous message).

> Resolving the charter and producing a threat analysis are our tasks right now

A diversion, necessary diversion, but a diversion from the arguments
I am attempting to make.

--ewh