Network Working Group B. Leiner
Request for Comments: 1560 USRA
Category: Informational Y. Rekhter
IBM
December 1993
The MultiProtocol Internet
Status of this Memo
This memo provides information for the Internet community. This memo
does not specify an Internet standard of any kind. Distribution of
this memo is unlimited.
Abstract
This document was prepared by the authors on behalf of the Internet
Architecture Board (IAB). It is offered by the IAB to stimulate
discussion.
There has recently been considerable discussion on two topics:
MultiProtocol approaches in the Internet and the selection of a next
generation Internet Protocol. This document suggests a strawman
position for goals and approaches for the IETF/IESG/IAB in these
areas. It takes the view that these two topics are related, and
proposes directions for the IETF/IESG/IAB to pursue.
In particular, it recommends that the IETF/IESG/IAB should continue
to be a force for consensus on a single protocol suite and internet
layer protocol. The IETF/IESG/IAB should:
- maintain its focus on the TCP/IP protocol suite,
- work to select a single next-generation internet protocol and
develop mechanisms to aid in transition from the current IPv4,
and
- continue to eXPlore mechanisms to interoperate and share
resources with other protocol suites within the Internet.
1. IntrodUCtion
The major purpose of the Internet is to enable ubiquitous
communication services between endpoints. In a very real way, the
Internet IS inter-enterprise networking. Therefore, the issue of
multiprotocol Internet is not just the issue of multiple network
layers, but the issue of multiple comparable services implemented
over different protocols.
The issue of multiprotocol Internet is multidimensional and should be
analyzed with respect to two simultaneous principles:
- It is desirable to have a single protocol stack. The community
should try to avoid unconstrained proliferation of various
protocol stacks.
- In reality there will always be more than one protocol stack.
Presence of multiple network layers is just one of the
corollaries of this observation, as even within a single
protocol stack, forces of evolution of that stack will lead
to periods of multiple protocols. We need to develop
mechanisms that maximize the services that can be provided
across all the protocol stacks (multiprotocol Internet).
2. Background and Context
2.1. The MultiProtocol Evolutionary Process
In an IAB architectural retreat held in 1991 [Cla91], a dynamic view
of the process of multiprotocol integration and accommodation was
described, based on the figure below.
--------------- --------------
! ! ! !
! ! ! Interop- !
! Primary ! >>>>>>>>>>> ! erability !>>>>>
! Protocol ! ! ! v
! Suite ! -------------- v
! ! v
! ! v
! ! -------------- v
! ! ! ! v
! ! >>>>>>>>>>> ! Resource ! v
! ! ! Sharing !>>>>v
! ! ! ! v
--------------- -------------- v
^ v
^ -------------- v
^ ! ! v
<<<<<<<! Harmonize !<<<<<<<<<<<<<<<<<<<<
! !
! !
--------------
Figure 1: MultiProtocol Evolution Process
The figure describes the process from the perspective of a community
working on a single primary protocol suite (such as the IETF/IESG/IAB
working on the TCP/IP protocol suite.) (Note: It must be kept in mind
throughout this paper that, while the discussion is oriented from the
perspective of the IETF/IESG/IAB and the TCP/IP protocol suite, there
is a complementary viewpoint from the perspective of each of the
communities whose primary focus is on one of the other protocol
suites.) There are other protocol suites (for example, IPX, OSI,
SNA). Although the primary emphasis of the community is developing a
system based on a single set of protocols (protocol suite), the
existence of other protocol suites demands that the community deal
with two ASPects of multiprotocolism. The first is interoperability
between the primary protocol suite and other protocol suites. The
second is resource sharing between the primary protocol suite and
other protocol suites. Both interoperability and sharing may happen
at multiple levels in the protocol suites.
Achieving interoperability and resource sharing is difficult, and
often unanticipated interactions occur. Interoperability can be
difficult for reasons such as lack of common semantics. Resource
sharing can run into problems due to lack of common operational
paradigms. For example, sharing bandwidth on a link may not work
effectively if one protocol suite backs off in its demands and the
other does not. Interoperability and resource sharing both require
cooperation between the developers/users of the different protocol
suites. The challenge in this area, then, is to develop mechanisms
for interoperability and resource sharing that have minimal negative
affect on the primary protocol suite.
The very attempts to achieve interoperability and resource sharing
therefore lead to an attempt to bring the multiple protocol suites
into some level of harmonization, even if it is just to simplify the
problems of interoperability and sharing. Furthermore, the
communications between the communities also leads to a level of
harmonization. These processes, together with the normal process of
evolution, lead to changes in the primary protocol suite, as well as
the other suites.
Thus, the need for new technologies and the need to accommodate
multiple protocols leads to a natural process of diversion. The
process of harmonization leads to conversion.
While this discussion was oriented around the relation between
multiple protocol suites, it can also be applied somewhat to the
process of evolution within the primary protocol suite. So, for
example, as new technologies develop, multiple approaches for
exploiting those technologies will also develop. The process then
hopefully leads to a process of harmonization of those different
approaches.
2.2. The Basis of the Internet
The rapid growth of the Internet has resulted from several forces.
Some of them are "practical", such as the bundling of TCP/IP with
Berkeley Unix and the early decision to base NSFNet on TCP/IP.
However, we believe that there is a more fundamental reason for this
growth. The Internet (and the TCP/IP protocol suite) were targeted at
Inter-Enterprise Networking. Although the availability of TCP/IP on
workstations and the desire to have a single environment serve both
intra- and inter-enterprise networking led to the use of TCP/IP
within organizations, the major contribution of the Internet and
TCP/IP was to provide to user communities the ability to communicate
with other organizations/communities in a straightforward manner
using a set of common and basic services.
Fundamental to this ability was the fact that the Internet was based
on a single, common, virtual network service (IP) with a supporting
administrative infrastructure. This allowed a ubiquitous underlying
communication infrastructure to develop serving the global community,
upon which a set of services could be provided to the user
communities. This also allowed for a large market to develop for
application services that were built upon the underlying
communications.
An important corollary to having a single common virtual network
service available to the end user (open network service) is that the
selection of applications becomes the province of the end-user
community rather than the intermediate network provider. By having
this common underlying infrastructure, user communities are able to
select their desired/required application services based on their
unique needs, with assurance that the intermediate networking service
will support their communication requirements. We believe that this
has been of considerable importance in the success of the Internet.
In addition to providing network layer services for TCP/IP transport
layer and applications, IP may be used to provide network layer
services for non-TCP/IP transport layer and applications. Such use is
clearly beneficial, since it allows preservation of all the benefits
of a single, common, virtual network service (IP), while at the same
time widening the set of applications available to the end users.
3. Directions for Multiprotocolism
Over the past few years, with the increasing scope of the Internet,
has come an increasing need to develop mechanisms for accommodating
other protocol suites. Most techniques have fallen into the regime of
either interoperability (techniques that allow for communications
between users of different protocol suites) or resource sharing
(allowing common resources such as links or switches to jointly
service communities using different protocol suites.) It must be
noted that such techniques have been quite limited, with
interoperability happening primarily at application layers and
resource sharing happening to limited extent.
This need to deal with multiple protocol suites has led to discussion
within the community concerning the role of the IETF/IESG/IAB
regarding the TCP/IP protocol suite versus other protocol suites.
Questions are asked as to whether the TCP/IP protocol suite is the
sole domain of interest of the IETF/IESG/IAB or if the community
needs also to deal with other protocol suites, and if so, in what
manner, given these other protocol suites have their own communities
of interest pursuing their development and evolution.
The answer to this question lies in understanding the role of the
IETF/IESG/IAB with respect to the process described above (Figure 1).
The continued success of the Internet relies on a continued strong
force for convergence, making sure that the primary protocol suite
(TCP/IP) is successful through an evolutionary process in
accommodating both the changing user requirements and emerging
technologies.
Since this process requires a continued effort to accommodate other
protocol suites within the overall Internet, efforts at
interoperability and sharing must continue. Thus, we can summarize
the directions for the IETF/IESG/IAB as two-fold:
- Have as a primary focus the evolution of the primary protocol
suite (TCP/IP), acting as a force for convergence at all times
towards a single set of protocols, and
- Make provision for other protocol suites within the global
Internet through mechanisms for interoperability and resource
sharing.
4. Next Generation Internet Protocol
The principles described above for multiprotocolism can also be
applied to the discussions regarding the next generation internet
protocol. Currently, there are several candidates for IPng, which
raises the question of how to deal with multiple protocols at that
level. We note that even if just one is selected, there is an issue
involved in transitioning from IPv4 to IPng.
Selection of a single Internet protocol is not the only way of
dealing with this issue. Even if a layer of ubiquity is required
(such as that provided currently by IP), we might consider providing
ubiquity at a different layer. For example, we could imagine having a
common transport protocol running over multiple internet protocols.
We also could imagine achieving interoperability by use of common
application services (such as Directory services) running over
diverse communication services (both transport and network layers).
These alternatives do not provide the considerable benefits of a
single internet protocol, and therefore would be undesirable. Having
a single internet protocol provides a common communication
infrastructure across the various networks, thereby achieving the
following:
- Communities of end users can select their desired applications,
independent of the technologies used to support the intermediate
networks.
- The common underlying infrastructure provides a common
marketplace upon which application developers can create new and
exciting applications. Installation of these applications does
not require end users to select a corresponding network protocol
(although some advanced applications may require enhancements,
such as high-bandwidth approaches).
Thus, the community (IETF/IESG/IAB) should continue to act as a force
for convergence by selecting a single next generation Internet
protocol and developing methods to ease the transition from IPv4 to
IPng. Specifically, at the applications layer, it is desirable to
promote different approaches and "let the marketplace decide."
However, it is unacceptable to treat the internet protocol layer in
the same way.
5. Conclusion
Historically, the IETF/IESG/IAB has acted as a strong force for the
development of the Internet by acting as a force for convergence on
and evolution of a single primary protocol suite. This has served
the community well, and this approach should be continued for the
future. In particular, the IETF/IESG/IAB should:
- maintain its focus on the TCP/IP protocol suite,
- work to select a single next-generation internet protocol and
develop mechanisms to aid in transition from the current IPv4,
and
- continue to explore mechanisms to interoperate and share
resources with other protocol suites within the Internet.
6. References
[Cla91] Clark, D., Chapin, L., Cerf, V., Braden, R., and
R. Hobby, "Towards the Future Internet Architecture",
RFC1287, MIT, BBN, CNRI, ISI, UC Davis, December 1991.
Security Considerations
Security issues are not discussed in this memo.
Authors" Addresses
Dr. Barry M. Leiner
Senior Scientist
Universities Space Research Association
625 Ellis Street, Suite 205
Mountain View, CA 94043
Phone: (415) 390-0317
Fax: (415) 390-0318
EMail: leiner@nsipo.nasa.gov
Yakov Rekhter
T.J. Watson Research Center, IBM Corp.
P.O. Box 218,
Yorktown Heights, NY 10598
Phone: (914) 945-3896
EMail: yakov@watson.ibm.com