Definition: Internetwork Layer: To avoid confusion with multiple meanings of ``network'' layer, we will use the term ``Internetwork'' layer to unambiguously refer to that layer at which IP runs. This is the layer at which IP routing functions. This is also the layer at which CLNP, DECnet, etc. run.
Large cloud: A collection of ``end-points'' be they routers or hosts, connected over a fabric such that communication can be established, in the absence of policy restrictions, between any two such entities. This communication within a cloud takes place using addressing and capabilities below the ``Internetwork'' layer.
The connectivity may or may not require circuit setup before communication. Such a collection is considered large if it is infeasible for all routing entities on such a ``cloud'' to maintain ``adjacencies'' with all others. Examples include, but are not limited to, ATM, Frame Relay, SMDS, and X.25 public services.
The group will investigate the operation of IP routing protocols and services over ``Large Clouds.'' Whenever possible, solutions shall be applicable to a range of ``cloud'' services. That is, the goal is a single solution applicable to multiple kinds of large ``clouds'' be they public or private, and independent of the specific technology used to realize the ``cloud'' (even a very large bridged Ethernet). It is also an objective that solutions, where possible, apply to network layer protocols other than IP.
The problems the group will cover are:
A) The architectural implications of allowing direct communication between entities which do not share a common IP network number. The group will also entertain proposals on the use of a common IP network number. If (as many believe) it is infeasible, an effort to document the difficulties will be made.
B) The routing/information protocol required to allow direct communication between two entities which were not directly exchanging routing information. This will include address resolution. The solution must couple closely to routing. It must take into account realistic connectivity policies.
C) Operation of existing protocols between peers on such clouds. Are any changes necessary or desirable? If changes are required, they will be proposed to the relevant working group.
D) Consideration of how policy restrictions and constraints (such as access control and policy-based routing paths) affect A, B, and C.
The group will also review the applicability of the work to ISDN and POTS. These technologies have a prima-facia difference, in that the number of simultaneous connections is much smaller. The implications of this for routing and relaying at the Internetwork layer will need to be explored further.