Hi, thanks for the response. Please keep in mind that my comments are no more
binding than any other, so if everyone else is convinced I am wrong, that's
okay :-)
Specific comments inline:
On Aug 26, 2011, at 6:24 PM, Mizikovsky, Semyon B (Simon) wrote:
Ben, thank you for your comment.
We do not specify the protocol between IKEv2 Peer and HAAA. We specify the
protocol between Diameter Client and Diameter Server. The Diameter Server
happens to be the HAAA. The Diameter Client happens to be the IKEv2 Server.
The protocol we specify simply delivers the key material to the Diameter
Client. Nothing else. We also allow the HAAA the benefit of using the
freshness elements that became available from the IKEv2 layer. But it is
entirely up-to the HAAA how these elements, namely Ni and Nr, are used,
although we provide the hint as a RECOMMENDATION.
I would concur that, if this draft was being produced in conjunction with, or
normatively references other drafts that describe the relationship between the
IKEv2 peer and the HAAA, then it would make sense to defer this to them. But if
you want an arbitrary IKEv2 peer to be able to interop with an arbitrary IKEv2
server/HAAA pair, then this sort of thing needs to be defined _somewhere_.
Use of a MUST in the draft that does not specify the end-point affected by
this MUST is entirely incorrect.
I recognize that this draft can't specify the IKEv2 peer behavior, so maybe it
can't solve the problem by itself. Is it defined _normatively_ somewhere else?
Now, allow me to examine your assumption that the Client and HAAA somehow may
use different rules for choosing or creating the PSK:
The Client is surely known to the HAAA, otherwise we would have to assume
that there is no prior relations between them (no shared secret) and
therefore the PSK scheme can't work by definition. For instance, this could
happen if the IKEv2 Server can not resolve the IDi reported by the IKEv2 Peer
into a correct User Identity, and therefore can not associate it with
specific proper Diameter Server. In such case, the default AAA will be
accessed, and due to the lack of a common shared secret a failure is the only
possible result. Any attempt of negotiations will also fail.
On the other hand, if the IDi is resolvable, and the proper HAAA is located,
than it is obvious that HAAA surely knows about this Client with proper
subscription and has the shared key with it. It obviously knows also which
PSK computation procedure is pre-configured in the Client, so no negotiations
are necessary either.
But my point is: all these procedures and assumptions are completely and
totally outside of scope of the draft which simply describes DELIVERY of the
already computed PSK from the Diameter Server to the Diameter Client.
You are conflating administrative relationships with implementation
relationships. If the IKEv2 client _implementation_ does not share a key
generation mechanism with the HAAA, then no amount of prior relationships and
configuration are going to make them work together. Basically, you get a high
chance that an IKE client from Vendor A cannot, under any circumstances,
interoperate with a peer that depends on an HAAA from Vendor B.
Yes, we specify the default PSK computation procedure in this draft on
insistence of reviewers. It is specified for those external applications
which do not specify their own, so "here is the right way to do it". But it
is entirely up-to the external application how to use it, configure it,
benefit from it, or even ignore it.
Can you remind me again what you mean by "external application"? From most of
the draft, I assume the external application is "IKEv2-PSK". Are we talking
about something along the lines of a 3GPP named interface, where
implementations for that context would not be expected to work with
implementations from some other context?
Simon.
Semyon (Simon) MIZIKOVSKY
600 Mountain Avenue
ALCATEL-LUCENT
DIRECTOR
WIRELESS SECURITY STANDARDS
600/700 Mountain Ave., Rm. 3C-506
Murray Hill, NJ 07974-0636 USA
T: +1 908 582 0729
M: +1 732 239 7533
F: +1 908 743 4361
Simon(_dot_)Mizikovsky(_at_)alcatel-lucent(_dot_)com
Note: This message, and any attachments, is intended only for the
recipient(s) identified above. The information contained in this message may
be privileged, confidential or proprietary, and its use or disclosure by
other than intended recipient(s) is prohibited and may be unlawful. If you
have received this message in error, please delete it, and do not distribute
or retain a copy of it.
-----Original Message-----
From: Ben Campbell [mailto:ben(_at_)nostrum(_dot_)com]
Sent: Friday, August 26, 2011 5:41 PM
To:
draft-ietf-dime-ikev2-psk-diameter(_dot_)all(_at_)tools(_dot_)ietf(_dot_)org
Cc: gen-art(_at_)ietf(_dot_)org Review Team; The IETF
Subject: Followup on Gen-ART Telechat Review of
draft-ietf-dime-ikev2-psk-diameter-08
This is a followup on my previous gen-art reviews of
draft-ietf-dime-ikev2-psk-diameter based on version 09. I realize that came
out a couple of weeks ago, and this followup may be overtaken by events. I am
sending it anyway on the off chance it's still meaningful.
This version partially addresses my concerns (below), in that it now
specifies a default mechanism for PSK generation. If I understand correctly,
use of this mechanism is completely optional. That is, the actual mechanism
chosen is still a matter of out-of-band agreement between the HAAA and the
IKEv2 peer. And if I read correctly, there is still no way for either to
declare or negotiate what mechanism they plan to use.
I realize the draft assumes that this will be used in the context of a
"protocol leveraging this diameter application", and that said protocol
should specify the key derivation mechanism to be used. I interpret (perhaps
incorrectly) to mean that a given IKEv2 peer and a given HAAA are expected to
be implemented for a particular context, and that the authors do not expect
an IKEv2 peer from one context to work with an HAAA from another.
Furthermore, it appears to me that if 2 such mismatched peers tried to
communicate, the only way they could determine they were incompatible would
be through authentication failures due to a key mismatch. I'm not sure that's
an appropriate assumption for an IETF proposed standard.
Ideally, I think things would be improved if the included key derivation
procedure was promoted to MUST implement, and a mechanism were added where
the peers can declare or negotiate the intent to use some other procedure if
they choose to do so. At the minimum, it would be good to have a way where
two peers could detect a key derivation mismatch early in the process.
On Jun 17, 2011, at 5:10 PM, Ben Campbell wrote:
I am the assigned Gen-ART reviewer for this draft. For background on
Gen-ART, please see the FAQ at
< http://wiki.tools.ietf.org/area/gen/trac/wiki/GenArtfaq>.
Please wait for direction from your document shepherd
or AD before posting a new version of the draft.
Document: draft-ietf-dime-ikev2-psk-diameter-08
Reviewer: Ben Campbell
Review Date: 2011-06-17
IESG Telechat date: 2011-06-23
Summary: This draft is almost ready for publication as a proposed standard.
I still have a concern about the generation of the PSK.
Major issues:
In my initial Gen-ART review, I made the following comment: The draft says
that the procedure that the HAAA follows to generate the PSK is out of
scope. But doesn't the IKE2 initiator need to understand the procedure? If
the procedure is not defined somewhere, how you achieve any degree of
interoperability?
The author responded that the PSK generation was in fact important for
interoperability, but that the specific procedures have been intentionally
left to other specifications. It is up to specifications that use this
Diameter application to define the PSK generation mechanism. Further, the
author indicated 2 3GPP2 specs that have done this.
I am still concerned that this means that this specification cannot be
implemented in an interoperable way without effectively profiling it. There
is no apparent coordination on how such profiling may be done, and by whom.
I think this is likely to result in implementation islands that can't talk
to each other. I recognize that there is precedent for doing this, but I
think it is something that should not be done without careful consideration,
particularly in a standards track RFC. I leave it to the IESG to confirm
whether it is appropriate in this circumstance.
I further note that there is no apparent way to negotiate or declare what
PSK generation mechanism might be used, if an implementation supports more
than one.
Minor Issues: None
Editorial Comments: None
_______________________________________________
Gen-art mailing list
Gen-art(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/gen-art
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf