Re: [Curdle] draft-ietf-curdle-ssh-kex-sha2 and diffie-hellman-group1-sha1 (1024-bit DH)

[Tero Kivinen <[e-mail]>]

Russ Housley writes:
> I agree with Tero.  However, I like the use of SHOULD+ and MUST- to
> signal to project and product planners what is coming.   

Actually it is much better to explain that kind of things in the
actual text. For example:

3.5.  diffie-hellman-group14-sha1

   This method uses [RFC3526] group14 (a 2048-bit MODP group) which has
   no concerns.  This generated key exchange group uses SHA-1 which has
   security concerns [RFC6194].  However, this group is still strong
   enough and is widely deployed.  This method is being moved from MUST
   to SHOULD- to aid in transition to stronger SHA-2 based hashes. This
   method will transition to MUST NOT when SHA-2 alternatives are more
   generally available.


Gives much better information what to expect than SHOULD- in table.
SHOULD+, SHOULD- were useful when we only had table of requirement
levels. Now when we have the text about each algorithm explain the
reasoning, those + and - characters are getting more and more
meaningless.

For example the SHOULD- is defined in section 2 to mean that this
algorithm will be deprecated to MAY in future version, but in practice
we quite often go from SHOULD- to SHOULD NOT (or even MUST NOT, as the
text above suggests).

Anyways this is separate issue than wheter we should make
implementations having code for diffie-hellman-group1-sha1
non-conforming by publishing that implementations MUST NOT implement
diffie-hellman-group1-sha1...
-- 
[e-mail]