Architecture [Re: Agenda for Vienna]

[Brian E Carpenter <brian@hursley.ibm.com>]

Tony Li wrote:
...
> This implies that we should not be looking at particular
> proposals.  We should not even be looking at 'categories'
> of proposals.  Instead, we should be talking about possible
> architectures.  That's one layer farther up.

This is *very* important. In the proposed second session, please
let us concentrate on the architectural model of each of the
two solution classes you want to explore.

For example, whether we achieve identifier/locator separation
by encapsulation, rewriting, 2-addresses-in-1-header, or
2-halves-in-1-address should be an implementation detail in
the first level of discussion.

Tony Li also wrote:

> |    Some notions that seem relatively undisputed:
> |    
> |    - we separate the identifier and locator functions so we can have 
> |    multiple locators and still have a single identifier
> |    - IP in transit works with locators, everything above the 
> |    IP layer with 
> |    the identifier
> |    
> 
> Do we really have consensus on these points?  Does this imply that
> tunneling is no longer a consideration?

As I just said, I don't think that follows. Map-and-encap is just an
implementation of id/loc separation, viewed from high enough.

   Brian