Comments on draft-templin-autoconf-netlmm-dhcp-00

I read this document. The draft looks good and simple to me.
And I have the following comment.

1. This draft may not fit into NETLMM as it involves the
host modifications.

2. It is not clear to me the criterion for the MAP to choose
between  route
     establishment and tunnel establishment.

3. How is the DAD procedures executed? Is it always
confirmed through DHCP?

4. How are the link local addresses are configured? I think
the whole LMMD
    should be considered as a single IP link. I think this
issue
exists for Network
    based solutions as well.

5. Section 4.2 says, advertising prefixes with L-bit zero,
assuming
that the hosts
    respect this, host always send the packet to default
router which
in turn may
    send to MAP even when the CN resides on the same
physical link. And how
    is the hop limit filed processed, decremented/unchanged?


Thanks,
Syam

_______________________________________________
netlmm mailing list
netlmmngnet.it
https://vesuvio.ipv6.cselt.it/mailman/listinfo/netlmm

Syam,

Thanks for the comments, and see below for responses: 

> I read this document. The draft looks good and simple
to me.

Good; simpler is better.

> And I have the following comment.
>
> 1. This draft may not fit into NETLMM as it involves
the host
modifications.

NETLMM also involves host modifications to implement (at
least)
DNA. There are also fundamental issues with an approach in
which
the network performs mobility management with no involvement
of
the host which have been described here:

https://vesuvio.ipv6.cselt.it/pipermail/netlm
m/2006-May/001175.html
https://vesuvio.ipv6.cselt.it/pipermail/netlm
m/2006-May/001249.html

In summary, the intent of the present draft is to show that
at
least one solution alternative exists that solves the NETLMM
problem space in a simple fashion while avoiding the issues
brought on by a network-centric approach.

> 2. It is not clear to me the criterion for the MAP to
choose between
route
>      establishment and tunnel establishment.

The intent is to have the MAP create a tunnel interface and
at the same time also create a host route for the MN that
points to the tunnel interface. It is implementor's choice
as to whether to create a single tunnel interface per AR or
a tunnel interface for every MN. If you can suggest text
changes to make this more clear, we will take them under
consideration for the -01 version. 

> 3. How is the DAD procedures executed? Is it always
confirmed through
DHCP?

Exactly the same as has been articulated in the NETLMM
doc's.
I believe that means confirmed on the local link as well as
confirmed through the DHCP server.

> 4. How are the link local addresses are configured? I
think the whole
LMMD
>     should be considered as a single IP link. I think
this issue
> exists for Network
>     based solutions as well.

Another view would be that the LMMD is a multi-link subnet,
with
possibly many access network links tied together via the MAP
(LMA).
Link local addresses would have to be configured and tested
on the
local link as well as registered with the DHCP server. Agree
that
this issue is the same as for network-based solutions.

> 5. Section 4.2 says, advertising prefixes with L-bit
zero, assuming
> that the hosts
>     respect this, host always send the packet to
default router which
> in turn may
>     send to MAP even when the CN resides on the same
physical link.
And how
>     is the hop limit filed processed,
decremented/unchanged?

As described in Section 5.5 ("Multi-Link Subnet
Considerations"),
the TTL/hop limit will be decremented only once for packets
sent
by the MN to the Internet, but will be decremented multiple
times for packets sent from the Internet (through the MAP
(LMA))
to the MN. This can potentially be mitigated by having the
MAP
(LMA) and/or AR act as proxies for inbound traffic instead
of
normal routers. 

Fred
fred.l.templinboeing.com

Thanks,
Syam

_______________________________________________
netlmm mailing list
netlmmngnet.it
https://vesuvio.ipv6.cselt.it/mailman/listinfo/netlmm

Syam Madanapalli wrote:
> I read this document. The draft looks good and simple
to me.
> And I have the following comment.
> 
> 1. This draft may not fit into NETLMM as it involves
the host 
> modifications.

it is not possible to do NETLMM without some kind of
indication from the mobile node to the access router
to "announce" itself. it could either be at L2
or L3.
most folks on the mailing list seem to assume it will
happen at L2 and therefore there is no host
involvement. thats wrong.

Vijay
_______________________________________________
netlmm mailing list
netlmmngnet.it
https://vesuvio.ipv6.cselt.it/mailman/listinfo/netlmm