Internet Host: nic.cerf.net
Directory: farnet
Subdirectory: farnet_docs
Filename: ccirnmay90-views
Last Updated: Aug 3, 1991
Preface: this file contains three further appendices to my report on the
May 1990 CCIRN meeting at Sophia/Antipolis, France. These views are personal
and in many respects tentative. They are offered here for the use of those
within the FARnet community who may find them of interest. They are not
necessarily the official position of FARnet or of any of its members.
-- Guy Almes
Contents:
Appendix C. Personal Thoughts on Protocols and International Networking
Appendix D. Personal Thoughts on the Situation in the UK
Appendix E. Personal Thoughts on Both Fat Pipes
Appendix C. Personal Thoughts on Protocols and International Networking
It's important in all this to distinguish among three important networking
technologies:
<> IP. This is the connectionless network layer of TCP/IP that we all use
within NSFnet, etc. It runs nicely over leased lines and routers.
<> CLNP. This is the connectionless network layer of OSI: TP4/CLNP. It
supports all the OSI application layers, and would also run nicely over
leased lines and routers. Favored by the American OSI community, CLNP
could be supported by the NSFnet Backbone and most regionals within a year.
<> CONS. This is the connection-oriented network layer of OSI: TP0/CONS/X.25.
It supports all the OSI application layers and makes good use of the X.25
networks common in Europe, but would not make good use of leased lines and
routers. Favored by the European PTTs, some European government GOSIPs,
and some elements within the EC.
The political battle is *not* between IP and OSI so much as between
connectionless IP+CLNP and connection-oriented CONS/X.25.
An attractive evolutionary path, particularly (but not exclusively) from an
American point of view, is dual IP+CLNP support using essentially the style of
leased lines, LANs, and routers we now employ. Then let the relative usage of
IP and OSI/CLNP balance itself out in the market. A small but growing number
of European countries are allowing CLNP within their GOSIPs.
The path favored by the PTTs and major elements within some European
governments and the EC is to mandate support for CONS/X.25. This would make
good use of the existing investment the European PTTs have made in X.25.
The real danger here (from an American point of view) is that some of the
Europeans may be pressed to adopt the CONS/X.25 style in order to conform to
their national GOSIPs and EC policies. This would require application- or (at
best) transport-level gateways to connect the connectionless American/Pacific
world with the connection-oriented European world.
The official policies of European networking on these protocol issues have
implications for the efficient use of the fat pipes. For example, British
policy does not permit use of IP or even (I think) CLNP within the wide-area
networks of the UK. Again, German (DFN) policy permits IP, but gives
preferential treatment to users of OSI protocols. This problem will have to
be worked out with sensitivity for European views and for the needs of the US
federal people to be diplomatic vis a vis their European government
counterparts.
Appendix D. Personal Thoughts on the Situation in the UK
To their credit, the JAnet people in England were pioneers in the use of
computer networking in the academic context, and they had made significant
progress by the time that the central importance of the Internet was clear.
The result is JAnet, an X.25-based network spanning all the academic and
research labs of the UK. Layered over X.25 are the so-called Coloured Book
protocols, named after the color of the documents that define them. Thus
their file transfer protocol is officially NIFTP, but more commonly `Blue
Book', and their electronic mail protocol (which, by the way, uses RFC 822
format and almost uses domain naming) is officially JNT Mail, but more
commonly `Grey Book'. These protocols are quite usable protocols based on
X.25 and providing reasonable email, ftp, and remote login service within
the UK.
The problem, of course, is with interoperability. Mail between the Internet
and JAnet has to traverse an application-level gateway in London. This
gateway, nsfnet-relay.ac.uk, is itself a performance and reliability bottleneck,
despite the dramatic improvements made during the last year. Moreover, since
it is a mail gateway, there is no hope for applications other than email between
the Internet and the UK.
So, to a first-order approximation, it might appear that the UK fat pipe will
only provide us with email service. This is almost true. There are two
immediate and one potential exceptions that I will mention.
First, the London end of the fat pipe will support a small number of campuses
that will appear directly on the Internet. UCL and ULCC are historic examples,
and a *small* number of defense, NASA-related, and academic sites will be
added.
Second, the London end will support video conferencing using the ST protocol,
just as is now available only in Los Angeles, San Francisco, Boston, and
Washington. Its use in support of trans-Atlantic conferencing will be of
interest.
Third, and currently only of potential importance, is a (to my knowledge)
little-known fact about the use of TCP/IP within the UK. At many sites in
the UK, the common workstations support TCP/IP (actually no surprise).
Moreover, TCP/IP is heavily used within campus LANs. The pattern emerges of
heavy IP use *within* each campus together with exclusive use of the Coloured
Book protocols *between* campuses. It seems likely that a majority of the
academic computers (ignoring small PCs) in the UK are capable of running IP,
have IP configured, and in fact use it day-by-day *but only within their
campus LAN*. Further, this indicates that there is a substantial user community
that would be able to fully use the Internet (in the strict IP sense) if only
IP were supported between the campus LAN and the nsfnet-relay in London. It
is amazing to me that the UK authorities refuse to support IP for these
purposes.
The only explanation I got for the lack of support for IP between campuses
was that the official policy is to use Coloured Book for the present, and to
use OSI when it materialises. To my knowledge, this is the most rigid policy
of any in western Europe, and it is particularly regrettable given the close
cultural affinity between the American and British academic community.
My hope is that, once the UK fat pipe is in place, diplomatic efforts can be
made to correct this situation. In the meantime, email service should be
fairly good. Also, despite its interoperability problems, JAnet is quite a
serious technical effort, and the people involved with it deserve our respect
and cooperation (along with some brotherly nudging!).
Appendix E. Personal Thoughts on Both Fat Pipes
The intention of American policy is to encourage the Europeans to take the
initiative in making intra-European networking work well, then to work with
them to build high-quality bridges across the Atlantic. Further, we do not
wish to micro-manage the Europeans, either in fact or in perception.
Thus, the German end of the fat pipe comes to the DFN, a German organization
with some German federal government support. DFN, it turns out, is inclined
to encourage the use of CONS/X.25 and to permit the use of IP/X.25. Thus,
if you send IP packets over their X.25 network (called WIN) straight to the
IP router at the German end of the fat pipe, then you will be charged for
this use. If, on the other hand, you use CONS/X.25 to get to an application-
level gateway at the German end, and this gateway then uses TCP/IP protocols
and sends IP packets to that same IP router at the German end, then you will
not be charged. Thus IP is permitted, but discouraged.
Now, apart from whether you find this situation agreeable, think about the
intention of American policy, and consider the interests of Germans or other
Europeans who want to use IP to get to America at high performance. The RIPE
folks, for example, might like to place one of their routers at the German end
of the fat pipe; they may not, however, for the German end is for the exclusive
use of DFN. The problem is that, while intending *not* to micro-manage the
European situation, we have given *one* European player (DFN, in this case)
control over access to a key trans-Atlantic bridge. We thus de-facto give DFN
a negative control they would not normally have.
Consider the following alternative. Suppose the German end of the fat pipe
had an IP/CLNS router with an ethernet interface, and that DFN *and others*
were permitted to place their routers/gateways on this ethernet. This provision
would be inclusive, not exclusive. With this alternative scenario, we would
not be influencing, either intentionally or unintentionally, the various players
in European networking. This suggestion could, perhaps, be implemented with
the current UK and German fat pipes.
By the way, the development of this example makes the German situation seem
worse than it is. The RIPE people, despite some criticisms of the fat pipe
arrangement, are enthusiastic about the effects of the support, albeit limited,
for IP throughout the very widespread WIN network.
In the future, I can imagine the following approach by the FNC to the planning
of trans-Atlantic pipes. If the Europeans are slow to do the thorough
cooperative intra-European networking that we wish they would, and if we are
to refrain from micro-managing while maximizing the technical quality and
effectiveness of internation infrastructure, then an open competitive approach
might be useful. The FNC could announce a call for proposals from the various
European networking groups. Each proposal would propose a specific fat pipe
and explain how it would be used to help reach a broad European community in
an inclusive, high-performance, way. Among the required sections of such
proposals might be:
proposed pipe: bandwidth, cost, etc.
proposed routing at European end: IP, CLNP, application/transport gates
proposed engineering: performance and reliability within Europe
proposed cost sharing: American and European contributions
proposed European community served: which European net(s) would be
served
quoted costs of leased lines from various European sites to the proposed
European end of the pipe
stated willingness to permit others to use the pipe
Thus the Americans would not have to guess who the correct players are. This
knowledge of the best players would be based on the best proposals. There are
surely political problems with implementing this suggestion, but it should be
explored with the FNC.
