Re: [secdir] SECDIR review of draft-ietf-hokey-erp-aak

[Qin Wu <[e-mail]>]

Hi, Radia:
Thank for your valuable review. please see my replies inline.

Regards!
-Qin
----- Original Message ----- 
From: "Radia Perlman" <[e-mail]>
To: "The IESG" <[e-mail]>; <[e-mail]>; <[e-mail]>
Sent: Thursday, February 09, 2012 2:38 PM
Subject: SECDIR review of draft-ietf-hokey-erp-aak


> have reviewed this document as part of the security directorate's
> ongoing effort to review all IETF documents being processed by the
> IESG.  These comments were written primarily for the benefit of the
> security area directors.  Document editors and WG chairs should treat
> these comments just like any other last call comments.
> 
> This is mostly OK technically, but sloppily written.  It's about giving a MH
> the keys to a new access point after a handoff.
> 
> Starting from section 3..."Peer" is sometimes called "MH" (mobile host). It's
> preferable to use a single term, and I'd prefer MH (since it's not clear who
> the MH is peer with in this protocol).

[Qin]: Okay. We can use MH for terminology consistency.


> 
> I found it surprising that the SAP knows where the MH will roam to.  I'd think
> that the MH would need to see a new CAP and tell its SAP who it wants to connect
> with, or that the MH finds the CAP and tells the CAP who its SAP is/was.

[Qin]: SAP doesn't know where the MH will roam to. But SAP knows which CAP is adjacent
 to itself and therefore SAP can provide the MH a list of CAPs which are all neighbor of that 
SAP via EAP-Initiate/Re-auth-Start message. Then MH can override CAP list carried in the 
EAP-Initiate/Re-auth-Start message and send a new CAP list using EAP-Initiate/Re-auth
 message to backend server through SAP. We can add some text to make this clear if you 
believe this addresses your concern.

> Under figure 6, it says "The x flag is reserved. It MUST be set to 0".
> Shouldn't the document say "and ignored on receipt"?

[Qin]: Agree, will fix this by following your suggestion. Thanks.


> There are various values that say TBD in the spec, but in the IANA
> considerations,
> there are values.  So shouldn't they be copied into the spec instead of TBD?
> Also, not all the TBD values are listed in the IANA considerations.

[Qin]: Okay.

> In the 4th paragraph before section 5.3, it talks about computing an integrity
> checksum over the ERP/AAK packet, excluding the authentication tag field.
> Does this mean that the authentication tag field is zeroed out for the
> computation,
> or that the message is truncated to exclude that field?

[Qin]: I think it is the latter case. The integrity checksum over ERP/AAK packet is computed 
from the first bit of code field of packet to the last bit of cryptosuit field of the packet which
 is placed before authenticator tag field.

> In the paragraph before section 5.3, it talks about silently discarding the
> EAP-Initiate/Re-auth packet if it's not supported by the SAP.  On a silent
> discard, how long do you wait for a response before assuming there won't be any,
> or perhaps that it got lost? (does it run over a reliable Transport?  Even so,
> it might be silently discarded).

[Qin]: That's a good point. In the case of SAP initiating ERP/AAK, this is not a issue since SAP has 
already support ERP/AAK. In the case of Peer initiating ERP/AAK, the peer should maintain 
retransmission timer and maximum retransmission times. Based on retransmission timeout estimation,
 the peer can know there won't be any response. Does this address your comment?

> There's also lots of minor typos.  Should be proof-read.

[Qin]: Okay, will fix this in the update. Thanks.


> Radia