Re: [Curdle] comments on draft-ietf-curdle-rsa-sha2-03.txt

[Peter Gutmann <[e-mail]>]

denis bider <[e-mail]> writes:

>The objection with additional variants is that if a fraction of deployments
>are set up to require them, this puts pressure on others to implement them as
>well. For Peter, this is a cost because his usage cases are extremely
>constrained.

It's also because I'm inherently lazy :-).  The smaller the number of
combinations of algorithms and mechanisms I need to implement, debug, and
test, the better.  SSL is an extreme example of this, there are such a vast
number of combinations of cipher suites, parameters, and message flows that
(a) no-one has ever tested them all and (b) when someone does come along and
test just one small corner of the whole mess, e.g. message flows, they
inevitably find vulns all over the place because composing all the different
bits and pieces is unsafe (there are now at least four, probably more,
conference papers published just on messing with SSL message flows, leading to
things like all-zero keys used and other issues).

Adding redundant ways of doing more or less the same thing isn't a good idea.
So my general objection is a combination of { code size minimisation, attack
surface reduction, work/effort reduction }.

Peter.