Re: HIP Mobility and multihoming draft has reached the repositories

[marcelo bagnulo <marcelo@it.uc3m.es>]

Hi Pekka,

I really liked the draft, thanks for submitting it.

Some comments.

- I guess that an end to end path failure detection mechanism is
required in order to detect when the multi-homed node has to send the
REA packet. I think that a general mechanism (meaning that it works
properly with any transport layer that is being used for the
communication) is needed. A possible option would be to use some sort of
keepalives between the communicating hosts, you can even use ICMP
packets. However, if this is the case, i guess that it would be better
to define a new type of packet (perhaps a HIP packet) to avoid the ICMP
filtering issues.  

- I fail to understand the proposed transition mechanism. If i
understood it correctly, the idea is to build FAs somewhere in the
Internet. These FAs will be used for mobility support, in which case
they will located in the mobile node "home network". Also, these FAs
will be used to support backward compatibility, so that no-hip enabled
nodes can communicate with hip enabled nodes for mutli-homing support.
In this case, where do you place the FA?, and what addresses does it
provide) (meaning from where does it obtains the addresses) i guess you
cannot put it in the mutli-homed network, since it is the one that may
become unreachable. So you can place it in the ISPs network, but in this
case it would obtains addresses from the isp, so the effect is similar
to only use a single isp.
Another option is to place it on a higher level isp, and obtain
addresses from it. However, i guess that this isp will not do it for
free, since it will have to carry all the traffic generated by non hip
enabled hosts, that i guess will be pretty high in the beginning.
And, another question related to this is which addresses are announced
in the DNS? I mean suppose that a multi-homed site is using HIP and it
is using a HA, then which addresses does it announce in the DNS?

- A minor comment..
In page 9 

"The purpose of the
   interfaces is to group the addresses into collections that are likely
   to experience fate sharing.  For example, if the host needs to change
   its addresses on interface2, it is likely that both address21 and
   address22 will simultaneously become obsolete."

While this section is a general section, this comment seems to me to be
too mobility oriented. I mean if the mechanism is being used for
multi-homing support, an address can become unreachable while the other
one is not"

Besides, i am not really certain that "The interfaces are logical
objects.", pseudo or virtual interfaces perhaps, but i would say that
real physical interfaces are the most natural interfaces. 

Thanks, marcelo 




On Thu, 2003-06-19 at 15:44, Pekka Nikander wrote:
> http://www.ietf.org/internet-drafts/draft-nikander-hip-mm-00.txt
> 
> --Pekka Nikander
> 
>