RE: additional requirement: cooperation

["Michel Py" <michel@arneill-py.sacramento.ca.us>]

>> I think your modified version is more restrictive than the original.
>> Multihoming today does not require cooperation at all between
>> transit providers in the general case (although it can be
>> necessary to achieve certain goals, e.g. load sharing between
>> providers without PI address space).

The purpose of my posting was that I think the original text has some
restrictions in the following sense:

A new multihoming solution might require a new protocol, let's call it
"xyz". The original text allows "xyz" to be configured between a site
and its transit providers, but not between transit providers. If "xyz"
needs to flow over the entire Internet, which is a reasonable scenario
for global multihoming, "xyz" would likely need to be configured between
transit providers.


>> How about:
>>  Two sites are "direct neighbours" if they are able to exchange
>>  traffic betwen their sites directly, without IP datagrams being
>>  processed by a router located at some third site.

>> That avoids the routing connotations, but accommodates traffic
>> carried over tunnels (since the encapsulated packets' IP datagrams
>> are unmolested by the routers involved in providing tunnel
transport).

You might want to be more specific, because it could be argued that IP
datagrams crossing routers providing tunnel transport are "unmolested"
but are "processed", indeed.
 
Otherwise the sentence is a lot better than mine ;-)

> "A multihoming strategy MAY require cooperation between a site
> and its transit providers, but MUST NOT require cooperation
> directly between the transit providers except for direct neighbors".