[e2e] common interface: path or node?

[Micah Beck]

Joe,

> > This implementation strategy leads to a situation in which a widely
> > implemented choice of forwarding protocol, in this case a particular
flavor
> > of IPv4, becomes expensive and disruptive to change.  Cisco is often
blamed,
> > but one might also blame the implementation strategy, which encourages
> > investment in hardware, software and procedures that makes the network
layer
> > protocol difficult to change.
>
> Alternately, it is precisely the reason the Internet has become
> ubiquitous, vs., e.g., protocols with more "fluxibility" ;-)

When you say "protocols with more flexibility," you are presumably comparing
the IP approach of adopting a generic network layer protocol as the common
interface to the network to approaches that impose more specialized
transport layer protocols.  Yes, that is the reason that the Internet was
able to meet the requirements of a diverse application community, and to
become ubiquitous.  This much is history.

I am talking about the current situation, where the application community
has grown much more diverse, and where the common network protocol, IP,  is
under pressure to satisfy requirements that are not necessarily consistent
with one another, and in some cases are not consistent with ubiquitous
deployment of IP in a scalable network.  A lot of time is spent arguing over
which modifications to IP would be effective, or would be advisable given
the requirements of network scalability.  IP has become ubiquitous through
flexibility - the issue I am raising is whether a network layer protocol can
be flexible enough to continue to meet the diverse needs of the application
community, or whether we may be reaching the limits of flexibility of a
single network layer protocol.

In other words, is it possible that IP is a victim of its own success?  By
making the Internet so flexible, it has encouraged the idea that application
requirements should determine the capabilities of the network, and invited
new communities to express their requirements.  Are these competing
requirements now so diverse that even IP cannot satsify them all - that
perhaps no single network layer protocol can?  If so, we must either make
hard choices, and disappoint some communities by failing to meet their
requirements.  Or else we must look for an implementation strategy that is
even more flexible, one that allows for heterogeneity at the network layer
while preserving interoperability.  Perhaps a common model of the
intermediate node.

/micah