Re: draft-ietf-ipsec-ikev2-06.txt

[Paul Koning <pkoning@equallogic.com>]

>>>>> "Andrew" == Andrew Krywaniuk <askrywan@hotmail.com> writes:

 Andrew> As Paul discussed, the MUST clauses can only refer to things
 Andrew> that affect compliance. Therefore, it sounds to me like the
 Andrew> only MUST we need is the following:

 Andrew> "Implementations that provide an interface for the user to
 Andrew> enter a purely alphanumeric shared secret (i.e. a password),
 Andrew> must allow that value to be a minimum of 64 bytes(*) long."

That sounds like a fine requirement.

 Andrew> Then there can be the usual warning about poorly-chosen
 Andrew> passwords elsewhere.

Right.

 Andrew> (*) Alphanumeric characters comprise 1/4th of the available
 Andrew> 256 bits. Therefore, for 128 bits of key strength, you need a
 Andrew> minimum of 64 bytes.

I think there's a glitch in your arithmetic.

If the password is just alphanumeric, that gives 62 possible
characters, so just about 6 bits per character.  So for 128 bits of
key strength you need at least 12 characters.

If you assume that a halfway reasonable passphrase has 2 bits of
entropy per character, then you do arrive at the minimum of 64
characters.  

I would suggest not putting that footnote into the spec; simply state
the requirement, which is a perfectly sensible one.

    paul