Re: Move forward

[Iljitsch van Beijnum <iljitsch@muada.com>]

On Wed, 12 Mar 2003, Pekka Savola wrote:

> > > > 3. Strong identifier/locator separation: the identifer isn't an address
> > > >    usable for routing, so the end host must implement the solution
> > > >    (basic multiaddressing as we know it today, SCTP, HIP)

> > > This is ignoring the short-term solutions with multiple addresses.

> > No, these are type 3.

> Well, not by the classification: there is no strong identifier/locator
> separation in e.g. host-centric multihoming approach.

Well, let's not start a flame war over this. If we're going to have a
few more classes then it would make sense to make this a separate class.

But my reasoning is that those solutions really use an implicit
locator/identifier separation: an email domain has several MXes. The
email domain is the identifier, the MX addresses are the locators. SCTP
doesn't even seem to use an identifier at all, so I guess the socket or
control block is the identifier there. It's certainly not any of the
individual addresses. And this must be done on the host so it's the
"strong" variety.

> > Note that they fail to meet some basic multihoming
> > requirements, though, so they shouldn't be considered viable solutions
> > as-is, except for some specific application types such as the DNS.

> Which realistic solutions _do_ meet those requirements?

> Right..

Address agile TCP, to name just one?

> Mobility could be useful, IMO, if and only if you could secure a construct
> like Binding Update with like, IPsec.  And this is a non-starter at the
> moment.

IPsec isn't as evil as people think it is. I got it to work in a few
hours. As long as we can have the IPsec just between the mobile host and
the home agent this shouldn't really be a show stopper.