Re: Identifier/locator recap

["J. Noel Chiappa" <jnc@ginger.lcs.mit.edu>]

    > From: Iljitsch van Beijnum <iljitsch@muada.com>

I've been meaning to send in a minor rant for some time now, and this message
gives me a good hook..

    > Traditional multihoming as is done in IPv4 will not scale.

That is to say, multihoming which is supported entirely by the routing.

    > An alternative is to give each host in a multihomed site an address for
    > each ISP the site is connected to. .. However, current transport
    > protocols are unable to jump to new addresses in mid-session. Solution:
    > separate the identifier and locator functions of the IP address.

Here's my rant:

This group is spending a lot of its time discussing the identity/location
separation. I wish to make it crystal clear that this is a *secondary* issue
when it comes to supporting multi-homing (the expressed purpose of this
group :-).

The way in which one does multi-homing that scales is use of multiple
routing-names (a.k.a. addresses)#. Not as much attention has been given to
this part of the problem (e.g. how do you decide which address to use, how do
you decide when to switch, etc) although I have recently seen some traffic
about it.

But please everyone, keep in mind that separation of location and identity is
*not* what supports multi-homing - what makes multi-homing work is use of
multiple addresses. Separation of location and identity comes in because of
*other* concerns that people have.

	Noel


Note #: I think it's possible to do multi-homing with only a single
routing-name (a.k.a. address) - but to do so would require that the addresses
be asigned automatically by a system which was looking at the actual physical
connectivity. In other words, addresses would not be assigned by a registry,
and would change as the topology changed. In terms of the location/identity
issue, you'd still wind up with the same problem, that an address was not a
useful long-term idenitifer of *who* you were talking to - it would only say
*where* it is.