Re: Thoughts about layering multi-addressing

[Tim Shepard <shep@alum.mit.edu>]

Pekka Nikander said:

> > Could this be something passed down from the ULP (prefer a given  
> > locator pair) for ULPs that are aware of multiple locators? Or a  
> > "locator pair preference sub-layer" for other ULPs?
> 
> Something like the following?
> 
>                   +----------------+  +--------------+
>                   | SHIM unaware UL|  | SHIM aware UL|
>                   +----------------+  +--------------+
>                           ^                   ^
>                           | ULIDs only        |
>                           v                   |
>       Admin       +----------------+          |
>       interface-->| "Policy" subL  |          | ULIDs+locators
>                   +----------------+          |
>                           ^                   |
>                           | ULIDs+locators    |
>                           v                   v
>                   +----------------------------------+
>                   |               SHIM               |
>                   +----------------------------------+
>                                    ^
>                                    | locators only
> 
> 
> The admin interface could then used to implement policy negotiation
> on the top of the basic SHIM, at some later point, if needed.  SHIM
> aware ULPs (like SCTP++?) could implement similar functionality
> themselves.


Nice picture!  It reminds me of a thought that's been bouncing around
in my head for the last 36 hours.....  At the ACM Turing Award Lecture
on Monday night (streamed live on the net, but video on line somewhere
I think), Bob Kahn mentioned that while they were thinking of a
layered system while designing the Internet protocols, the picture
they had in mind was not the usual vertical stack of protocols, but
rather stair steps.  That way, the interface of each layer is exposed
allowing others to build on top using whatever interfaces they need.
Pluggability and extensibility were more important to them than
layering.

Your picture makes me think you may have seen the lecture.

BTW, some of the discussion between Cerf and Kahn was very relevant to
the issues facing the net today regarding the mobile, multihomed, and
less-than-always-connected natures of the machines connected to the
net today.  (Cerf said that when they were designing the Internet
protocols in the 1970s, the machines connected to the net were rather
large and tended to stay put.)

			-Tim Shepard
			 shep@alum.mit.edu