RE: WG next steps

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

On Thu, 14 Nov 2002, Michel Py wrote:

> > There are lots of ways to find hosts that aren't
> > typically considered routing, e.g. DNS.

> I agree with Shane. Although I would prefer a routing solution (it
> appears to me that host solutions are a way to get around a network
> structure that does not preserve end-to-end) I think that part of the
> puzzle might be non-routing host solutions.

Unless I missed more sleep than I thought, nobody proposed a pure
routing solution which will do what we want in the long term. All other
solutions are multi-address ones.

So what is the best way to do this? In my opinion:

1. Explicit identifier/locator separation. Locators should look like
IPv6 (or IPv4) addresses = no changes to IP or routing. Identifiers come
in many shapes, including but not limited to IPv4 and IPv6 compatible
ones.

2. Applications that do not require lower layer network access only work
with identifiers. Identifiers are mapped to "care of" addresses.

3. Explicit identifier + locators + compatibility + security/privacy
options are negotiated with the system present at one of the care of
addresses before engaging in transport layer communication. (Say hello
to layer 5.)

4a. New/improved transport protocols change locators when necessary.

4b. Existing transport protocols interact with IP through an extra
sublayer that handles locator agility. Alternatively, this function (as
well as the negotiation phase) are handled by a proxy agent external to
the host.

Software engineering 101: when the going gets though, the SE adds a
layer of indirection.

I'll write it up in more detail in two weeks.