Hi Yaron,

Thanks for the detailed review. Just responding to those that need more
discussion or where I have comments.

On 4/11/09 1:02 AM, "Yaron Sheffer" wrote:

> - 3: I suggest to start this section with an exhaustive list of locations
> where the Redirect payload can occur (IKA_SA_INIT response, IKE_AUTH last
> response, Informational from gateway).

Section 1 already talks about all three.

I think a better option would be re-name section 3 as "IKEv2 Initial
Exchange with Redirect". The sections that come later are already called
"Gateway Initiated Redirect" and "Redirect During IKE_AUTH Exchange".

> - 3: "The VPN client indicates support for the IKEv2 redirect mechanism by
> including..." -> By including ..., the VPN client indicates that it supports
> the IKEv2 redirect mechanism and that it is willing to be redirected.

Hmm... The client can always exclude the REDIRECT_SUPPORTED payload based on
its configuration if it does not want to be redirected. So do we need to
mention this? 

> - 3: The last paragraph (re: MIP6) is out of context here, and I was unable
> to understand it. I suggest to clarify and extend it into a separate
> section.

When redirect is used with Mobile IPv6, there is a separate Home Agent
discovery mechanism. So the Mobile IPv6 stack would have a home agent
configured. If the IKEv2 redirect mechanism changes the gateway information,
then the state becomes inconsistent. The original intent was talk about
that. I replaced the paragraph with the following.

   When this mechanism is used with Mobile IPv6, care must be taken to
   ensure that the home agent information is consistent with the IKEv2
   gateway information.  The Mobile IPv6 home agent discovery mechanisms
   (for instance, RFC 5026 [4]) would have a configured the mobile node
   with a particular home agent.  When the mobile node initiates an
   IKEv2 exchange with the home agent and is redirected to another
   gateway, the home agent information should also be updated, subject
   to the policy on the mobile node.

> - 6.2: Now that we've appended the nonce, we should signal the length of the
> Identity field explicitly (possibly by stealing 1 octet from the Identity
> Type). Even though the client can theoretically compute it using the length
> of the expected nonce.

Good point. I added a 'GW Identity Length' field to both REDIRECT and
REDIRECTED_FROM payloads. I added the field to the REDIRECTED_FROM payload
also for consistency.

Vijay

_______________________________________________
IPsec mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/ipsec

Reply via email to