Re: [802.21] Comments on 21-06-0698-01-0000_IP_configuration_methods_IE.doc
Hi Yoshi,
my comments are inline.
On 8/5/06, Yoshihiro Ohba <yohba@tari.toshiba.com> wrote:
> Hi Daniel,
>
> Please see my comments below.
>
> On Sat, Aug 05, 2006 at 10:11:04AM +0900, Soohong Daniel Park wrote:
> > Good discussion and effort Yoshi,
> >
> > For further details, we may have to clarify each facility like below:
> >
> > - IP Address Configuration Method IE
> > - DHCP Server Address IE
> > - HA/FA Address IE
> >
> > Let me answer on the first IE among IP_Configuration_Methods_IE.
> >
> > As I explained in SD, the existing flag of PoA Capability IE is not
> > sufficient since there is no any explicit information in order to
> > prepare what kind of IP address configuration will happen while
> > attaching a new PoA. For example, even the mobile station has a PoA
> > Capability IE=IPv6, the mobile station based on the existing IP
> > stack/RFC will trigger Router Solicitation message (after listening
> > Router Advertisement for a while) in every time regardless of DHCPv6
> > availability (It means there is no RS/RA capability). That causes lots
> > of delays for the IP configuration when in motion. Simple extension of
> > 21 IE can resolve these problems in mobile environment. If the mobile
> > station has a PoA Capability IE=IPv6 & DHCPv6, then the mobile station
> > will trigger DHCPv6 Solicit message immediately instead of RA or RS
> > message for its own IP configuration method. I guess it's clear
> > concept. Isn't it ?
>
> First, it is not clear whether it is allowed to run DHCPv6 before
> running router discovery.
I don't think we need a deep implication regarding the proposed IE
since it is a simple hint to support more efficient mobility than now.
Anyway, it is most likely L2 trigger concept (especially Link UP) that
can trigger upper layer (IP layer) operation before completing its own
L2 operation. This benefit is an explicit consensus in IETF and
world-wide deployment now in mobile environment. That can reduce
reduncent delays sort of caused by hierarchical host operations
(L1->L2->L3->upper layers). Given this utility, the mobile station can
run its IP operation in parallel while L2 attachment. Likewise, 802.21
can trigger DHCPv6 before runing router discovery based on 802.21
conceptual model (especially Event Service).
>
> Second, if it is allowed to run DHCPv6 before running router
> discovery, then it should be possible to run DHCPv6 and router
> discovery in parallel regardless of the use of proposed IE or even
> regardless of the use of 802.21.
No, To run DHCPv6 on the host, it should get a related information
from router (Router Advertisement [Managed flag] = set), then the
mobile station can start sending DHCPv6 Solicit message. If no RA
(host timeout), the mobile station can also start initiating DHCPv6
operation. From the real experience, it spends at lease a couple of
seconds or more. Yes, it is an existing limitation from the IETF
perspective, but we can live with 802.21, and I don't see any trouble
to make use of the facilities of 802.21 to resolve those limitations.
> The only benefit to use the proposed
> IE would be to avoid running DHCPv6 when DHCPv6 is not used at all for
> configuring IP address and other information, but I think DHCPv6 is
> used in most deployment at least for configuring other information.
It is also required the host to get a related information from the
router (Router Advertisement [Other flag] = set). Host timeout is
applied for this case too. But, it is not a critical part from the
handover perspective since other information will take place after
completing IP configuration. So, it is not scope of 21.
> So the benefit of the IP_Configuration_Methods_IE is still not clear
> to me.
I hope the mentions above help. Further details are highly welcome.
--
Daniel (Soohong Daniel Park)
Mobile Convergence Laboratory, SAMSUNG Electronics.