Re: [IPsec] RFC4869 bis submitted

[Bill Sommerfeld <[e-mail]>]

On Thu, 2009-11-19 at 15:08 -0800, Paul Hoffman wrote:
> The text says:
>   IKEv1 implementations MUST
>   support pre-shared key authentication [RFC2409] for interoperability.
>   The authentication method used with IKEv1 MUST be either pre-shared
>   key [RFC2409] or ECDSA-256 [RFC4754].
> To me, that sounds like preshared keys are just fine for IKEv1 in this
> profile, but I might be misunderstanding what you mean by "usefully".

I make a distinction between "optional to use" and "optional to
implement".  There are many things which are optional to use but
effectively mandatory to implement -- an implementation is not viewed as
complete unless it allows the use of the option.

As you pointed out earlier in this thread, this profile is an individual
submission describing a particular organization's requirements.  I'm
reluctant to guess or rely on the guess of someone who isn't part of
this organization when the document itself could be clarified.

> >As a practical matter, the ECDSA piece of this spec is likely to be the
> >largest and last piece built -- given a working elliptic curve codebase,
> >plugging ephemeral ECDH into an IKE implementation is a much smaller
> >problem than building ECDSA into both an IKE implementation and the PKI
> >client codebase, tools, and keystores it relies on.
> 
> Probably true, but ECDSA is far from impossible, as the OpenSSL people
> have shown for a while now.

There is a large gap between "can deliver next week" and "impossible".
(I'll also note in passing that changing which PKI client codebase you
use in an IKE implementation is also a sizeable task).

					- Bill