Re: [Mip4] Request to consider FMIPv4 as a working group document

[Rajeev Koodli <rajeev at iprg.nokia.com>]

Hi Alex, Espen and Kent,

thank you for your comments. We can clarify these points
in the draft. I will also provide some comments below.

Regards,

-Rajeev


Alexandru Petrescu wrote:

> [I have some comments on the draft, not the request itself, please
>  disregard if already discussed or clarified.]
>
> Espen Klovning wrote:
>
> > I do have some concerns that more or less reflect what Kent Leung has
> >  voiced in previous mails on the list. 
>
> >- AR == FA
> >- IP address management
>
> I tend to agree with this.
>
> The draft explains, but is not clear about what happens to FMIP when
> CCoA/DHCP _is_ used.  If DHCP is _not_ used with FMIP then the draft 
> should say so and in a coherent manner, IMHO; otherwise the generic 
> "compatibility" statements confuse me at three points.
We decouple new DHCP address acquistion latency by binding the current 
CoA (either CCoA obtained via DHCP or FA-CoA) to the NAR's IP address. 
So, when the MN arrives on the new link, it can do
DHCP/MIP while it can still use its previous CoA to send packets, and it 
receives packets sent to Previous CoA from the NAR.

>   "[First Scenario]: the access routers are not involved in IP address
>    management for the MN. The mobile nodes acquire a new care-of address
>    upon attaching to a new subnet link as they normally do."
>
> In some deployments, MN normally acquires an address with DHCP and the
> first-hop router is a DHCP Relay, so the AR practically _is_ involved in
> IP address management for the MN, although the AR is not a DHCP Server.
>  So the question coming to mind immediately is what is meant by "access
> routers are not involved in IP address management for the MN"?  Does
> that simply mean that AR is not a DHCP Server and is a DHCP Relay?
The description is contextual. When we mentioned "access routers are not
involved in IP address management..", it is in reference to the Second and
Third scenarios where the AR/FA provides an IP address.
But, you are right that the First scenario includes an AR acting as a DHCP
relay. The protocol itself does not depend on such a functionality. This can
be clarified.

> Second point is about DHCP and additional delays.  If vanilla DHCP is 
> used then the effective DHCP delays are in the range 1s-10s with a 
> 802.11b WEP first-hop, 100Mbit Ether and GHz computers.  That is very 
> large, easily influencing real-time.  It is not clear in the draft how 
> any of these numbers influence FMIP - is FMIP slower if vanilla DHCP 
> is used?
Since packet forwarding to/from MN does not depend on DHCP completion, it
does not affect. (See above. the forwarding is set up for Previous CoA).

> Third point is the apparent incoherence between abstract and body, 
> with respect to DHCP.  The abstract says "For reducing the IP address 
> configuration, the draft provides two alternatives" and none is DHCP. 
> Or, later in the body one can read "Address allocation and 
> configuration may be supported using DHCP or any other method."
This needs to be rectified. _If_ the NAR returns a New CoA for exclusive 
use of MN (CCoA) via
DHCP, the NAR must have a pool of addresses to choose from when 
processing HI.



> Just some thoughts...
Thanks again for your comments.

> Alex


--
Mip4 mailing list: Mip4 at ietf.org
   Web interface: https://www1.ietf.org/mailman/listinfo/mip4
    Charter page: http://www.ietf.org/html.charters/mip4-charter.html
Supplemental site: http://www.mip4.org/