Document: draft-ietf-netlmm-proxymip6-10.txt
Reviewer: Elwyn Davies
Review Date: 18 Feb 2008
IETF LC End Date: 20 Feb 2008
IESG Telechat date: 21 Feb 2008

Summary:
This document is well written and is in fairly good shape for submission to the IESG.
There are a number of minor issues which ought to be fixed.  I think that for a more general reader there are a number of points that are fully covered in the detail but when reading the introduction and protocol overview questions arise which aren't answered until later in the document, and I was left thinking whether the problem had been addressed until much later.  I have noted these items in the comments.  
I believe it would only take a few sentences to lay these issues to rest and make the document easier to understand for those who are not immersed in netlmm.

Comments:

s3, paras just after fig 2:  The para after fig 2 claims that the LMA sets up the bidir tunnel; then the next para claims that the MAG 'establishes' the (same) tunnel.  Be clear which one has responsibility.

s3, p12:
>
>    The local mobility anchor, being the topological anchor point for the
>    mobile node's home network prefix, receives any packets that are 
> sent by any correspondent node to the mobile node.
>
I think it should be made clear (assuming I understand correctly what is going on) that this 'correspondent' node does not require any of the
MIPv6 correspondent functionality and will not see any specialized MIPv6 messages.  It is difficult to think of an alternative term, but this could be confusing.

s5.3.4, item 2:  Whether the tunnel is deleted is surely an implementation issue.  The LMA and MAG could agree to maintain the tunnel even if there are no MNs active on the MAG to save on setup costs.  I think this could be a MAY, leaving it up to implementers to optimize if desired. (This is actually discussed later in s5.6.1.. a forward pointer is needed here)

s6.1, bullet 2:  Does this make any assumptions about commonality of IID between addresses used by the interface?

s6.5: I thought it was said earlier that checking that the MN is entitled to mobility services was a MUST.  Does the SHOULD refer to the means of authenticating?

s8.2: The (M) flag is not added to the Binding Ack message by RFC 4140 (only to the Binding Update msg).

s10: The new registry for the Handoff Indicator values is not specified in the IANA Considerations.

Areas where some earlier explanation would make life easier for the general reader:

3:  I think it would be useful to explain how the LMA knows that the MN that has moved from p-MAG to n-MAG is the same MN.. and hence gets the same prefix somewhere here (the MN_identity, auth and policy I think).

s5.1, last para:  (it turns out that s5.2 and s6.3 give most of the answers but I was left worrying) States that the mobile node's address prefix would normally be the key for the binding cache entry.  This implies that it will be a unique key and hence every MN requires a different address prefix.  I think this is sort of implied earlier, but I think it could be stressed at the outset.  This raises a couple of more significant issues in my mind:
- A MAG using stateless address config will be advertising one prefix per attached MN:  if there are lots of them the RA's will be very large.  Is this an issue? (not if everything is a pt2pt link, but
- How does the MN know which prefix to use if it isn't a pt2pt link?
[s5.2 does not address these issues although it clarifies that the prefix per MN mode is used]  s6.3 states that it is assumed that links will be pt2pt and hence only one prefix on link.  This should be made clear earlier, probably in the protocol overview.

I think some additional notes on the Shared_prefix model case and what happens if the links are *not* pt2pt would be helpful:  There is a significant risk of the RAs getting very large if multiple prefixes need to be advertised - and it requires slightly different policy to make sure a MN can pick the right prefix if there is more than one prefix advertised.

s6.4: I think that explaining where the DHCPv6 server would be and how it would know what prefix to take the address out of would be useful.  
(A few words plus a pointer to s6.11 would do.)

Editorial:

(idnits is happy with the document)

s2.2, MN-Identifier: Expand NAI and MAC acronyms.

s4.1: Associate PAD with Peer Authorization Database.

Fig 5: Provide key to various acronyms.

s5.1, bullet 1: 'enabled'/'turned off':   This reads as if the flag is 
sometimes present and sometimes not present.  Suggest using 'set'/'reset' or 'true'/'false'.

s5.1, bullet 3: Define ALL_ZERO.

s5.2: The terms Per-MN-Prefix and Shared-Prefix model are not defined in this doc.  Are they imported from some other netlmm doc?

s5.3: Would be useful to put in a pointer to the message format descriptions here.

s5.4.1.1: use of == and != ... use of words is better than C programming language .