Re: [Mipshop] AD review of draft-ietf-mipshop-pfmipv6
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Mipshop] AD review of draft-ietf-mipshop-pfmipv6



The new version looks good to me, thanks. Please submit it to the directories.

Jari

Hidetoshi Yokota wrote:
Hi Jari,

Thanks for your patience. The attached revised document is supposed to
reflect the rest of your comments. Please also see inline, where the
revised part is explained item by item. If there is no outstanding issue
there, we will upload it shortly.

Hidetoshi Yokota wrote:
Hi Jari,

Appreciate your detailed review. The editorial corrections are mostly
done and the revised version is attached to this mail. Regarding the
technical comments, only the policies for the correction are listed for
now. The complete correction will be posted as soon as possible. Please
see inline for the responses to your questions and comments:

Jari Arkko wrote:
I have reviewed this document. It was generally in good shape and easy to read. I did find a number of issues though. Please discuss them with me on this thread and/or revise the draft accordingly.

Technical:

I do not understand the IP host aspects of the handover. For an unmodified host, what kind of interfaces exist on the host, what addresses they have, and at what time are interfaces removed or added? Is this exactly the same as in RFC 5213, or does PFMIPv6 introduce some change here?
The basic principle would be that this specification does not require
any additional IP-level function to the MN running in the PMIPv6 domain.
This MN should therefore be the same as defined in RFC5213.

A typical network interface would be one with the cellular network,
where the network basically controls the movement of the MN. Different
types of interfaces could be involved such as different generations (3G
and 3.9G) or different radio access systems. Any type of IP address
could be assigned (IPv4/v6 or global/private), but the assigned address
must be preserved before and after the handover. PFMIPv6 should be able
to handle the single radio situation, where only one interface is active
at any given time. This is a tougher situation in terms of packet loss
and delay.

I would like to see a new section on manageability considerations. For instance, Section 5 talks about some configuration issues. These should be mentioned, and there should be a description of what the operator needs to configure in order to set up a PFMIPv6 network.
We will add a new section for manageability considerations to clarify
the above and to reflect your suggestion.

A new subsection is added in Section 5 describing the above aspects:

-------------------
5.  PMIPv6-related Fast Handover Issues

5.1.  Manageability Considerations

   This specification does not require any additional IP-level
   functionality on the LMA and the MN running in the PMIPv6 domain.  A
   typical network interface that the MN could be assumed to have is one
   with the cellular network, where the network controls the movement of
   the MN.  Different types of interfaces could be involved such as
   different generations (3G and 3.9G) or different radio access
   systems.  This specification supports a MN with the single radio
   mode, where only one interface is active at any given time.  The
   assigned IP address is preserved whether the physical interface
   changes or not and the MN can identify which interface should be used
   if there are multiple ones.
--------------------


If the new router's interface is configured to
respond to queries sent to link-layer addresses than its own (e.g.,
set to promiscuous mode), then it can respond to the NUD probe,
providing its link-layer address in the solicited Neighbor
Advertisement.
Is this according to RFC 5213? I seem to recall that RFC 5213 operated on the same link layer addresses.
I think your observation is correct... I'll check with the other authors
to see if it is ok to rewrite it.

The new text doesn't mention the promiscuous mode and emphasizes the
common link-layer address among all MAGs in the PMIPv6 domain:

---------------
5.2.  Expedited Packet Transmission

   [...]

   The protocol specified in this document is applicable regardless of
   whether link-layer addresses are used between a MN and its access
   router.  A MN should be able to continue sending packets on the
   uplink even when it changes link.  When link-layer addresses are
   used, the MN performs Neighbor Unreachability Detection (NUD)
   [RFC4861], after attaching to a new link, probing the reachability of
   its default router.  The new router should respond to the NUD probe,
   providing its link-layer address in the solicited Neighbor
   Advertisement, which is common in the PMIPv6 domain.  Implementations
   should allow the MN to continue to send uplink packets while it is
   performing NUD.
----------------

At least one mobility option MUST
uniquely identify the target MN (e.g., the Mobile Node Identifier
Option defined in RFC4283 <http://tools.ietf.org/html/rfc4283>) and the transferred context MUST be for
one MN per message.
I would like the required options to be specified in more detail. Which identity options are sufficient to satisfy the MUST?
The required option should be the MN Identifier in the Mobile Node
Identifier Option.

The above description is added in Section 6.1.1:

---------
   Requested option
             In order to uniquely identify the target MN, the MN
             Identifier MUST be contained in the Mobile Node Identifier
             Option.
---------

If a default set of context information is defined and
always sufficient, this option is not mandatory.  This option is more
useful to retrieve additional or dynamically selected context
information.

Context Request Option is typically used for the reactive (NAR-
initiated) fast handover mode to retrieve the context information
from the PAR.  When this option is included in the HI message, all
the requested context information SHOULD be included in the HAck
message in the corresponding mobility option(s) (e.g., HNP, LMAA or
MN-IID mobility options).
Please specify what the default set of context information is, by listing the required options when the CRO is not present.
The default context information to request is the Home Network Prefix
Option. If the Mobile Node link-layer is available and used, the Mobile
Node Link-layer Identifier Option MUST also be requested. With these two
options and the MN identifier, the NMAG can construct the PBU.

The above description is added in Section 6.2.1:

----------
   The default context information to request is the Home Network Prefix
   Option.  If the Mobile Node link-layer is available and used, the
   Mobile Node Link-layer Identifier Option MUST also be requested.
----------

Editorial:

   HO-Initiate:
        A generic signaling message, sent from the P-AN to the PMAG that
        indicates a MN handover.  While this signaling is dependent on
        the access technology, it is assumed that HO-Initiate can carry
        the information to identify the MN and to assist the PMAG
        resolve the NMAG and the new access point or the base station to
        which the MN is moving to.  The details of this message are
        outside the scope of this document.

4. Proxy-based FMIPv6 Protocol Overview
Section 4 would probably benefit from an additional paragraph at the beginning to explain what assumptions there exist about lower layer functionality.
Yes, the predictive fast handover relies on the lower layer
functionality to trigger to send the HI. A new paragraph will be added
to clarify it.

-------------
4.  Proxy-based FMIPv6 Protocol Overview

   If the MAGs can be informed of the detachment and/or attachment of
   the MN in a timely manner via e.g. the lower layer signaling, it will
   become possible to optimize the handover procedure, which involves
   establishing a connection on the new link and signaling between
   mobility agents, compared to the baseline specification of PMIPv6.

  [...]
-------------

All the other corrections pointed out below have also been reflected in
the revised document.

Regards,


Note: Messages sent to this list are the opinions of the senders and do not imply endorsement by the IETF.