COA(90)12 22 January 1990 Report to RARE CoA on TCP/IP ============================ B.Carpenter, L.Backstrom, G.Pujolle (assisted by P.Kirstein) This report was produced at the request of the RARE Executive Committee for presentation and discussion at the CoA in January 1990. It was prompted by the rapid growth of wide-area TCP/IP in the European A&R community, the formation of RIPE as a forum for technical coordination of TCP/IP activities, and the obvious question of how RARE should respond to this situation. Contents: 1. Facts 2. Opinions and comments 3. Options 4. Recommendations 1. FACTS -------- This chapter is simply a catalogue of facts concerning the TCP/IP and OSI situation as it concerns the RARE community. "TCP/IP" is taken to refer to the entire suite of protocols and products built around the DARPA Internet Protocol. In fact, since the only protocol handled at sub-network and internet level is IP, it is more correct to refer simply to "IP", except that this may lead to confusion with ISO IP (IS 8473). In this report "IP" refers to DARPA IP. * TCP/IP is vendor-independent and completely open. * TCP/IP is used universally on-campus. Its wide-area use is growing fast, driven initially by Unix but increasingly by the fact that it is supported on all major operating systems. International wide-area use is also growing fast (EUnet, NORDUnet, EASInet, HEPnet) as well as use by national and regional networks. A map is available which gives a partial indication of the current situation. * High-performance IP routers are off-the-shelf and moderately priced. IP has proved its efficiency on lines up to 1.5 Mbps. * IP is the base for current and foreseen distributed computing services such as X, NFS, and RPC. OSI offers no current or imminent alternative. In fact the "COSINE" OSI services are a small subset of what TCP/IP offers today. * The US Internet policy is now "integration" of OSI and IP, rather than migration. BITNET is moving its wide-area traffic onto IP and tests of this mechanism have started within the EARN community. * Efforts have started (at least in the USA) to run OSI application protocols over a TCP/IP base. Indeed EAN/X.400 has been in production over TCP/IP for years. Efforts have also started (both in USA and Europe) to tackle the problem of interconnecting Connection Mode and Connectionless networks and resolving the transport class problem. If these various efforts succeed, integration of TCP/IP with OSI, and migration paths, will become easier. * IP is reasonably efficient over low-speed X.25 (up to 64 kbps). Its efficiency over high-speed X.25 (say 2 Mbps) is a subject of considerable debate and depends critically on technical factors such as window size and transmission delay. * There is at present no mechanism to fund an optimised Europe-wide IP network (RIPE can only hope to coordinate, not to optimise). Still less is there a mechanism to optimise multi-protocol networking or to design a relevant strategy. * Intercontinental coordination of research networks is carried out under the aegis of the CCIRN, but the European TCP/IP community is not formally represented in the CCIRN today. * RIPE exists as a de facto technical coordination forum, as an indirect result of a CCIRN action, and contains many of the "right" technical people (those with knowledge, skills, and commitment to IP services). Note that some IP service providers have not yet joined RIPE while awaiting clarification of RARE's attitude. 2. OPINIONS AND COMMENTS ------------------------ This chapter is a compendium of opinions and comments that have reached the authors of this report. Most of them are derived from an electronic "opinion poll" conducted during recent weeks. Where the opinion was given as an official one, the organisation concerned is indicated, but we have summarised the wording. A few comments were received too late for inclusion, but none of them were outside the range of opinions given here. * There is a widespread user perception that OSI services are arriving too late to cover all applications, and that TCP/IP services can fill the gap. Many comments addressed technical requirements to improve IP connectivity, rather than the question asked ("What should RARE do about IP coordination?"). "Need high-performance European IP services now." [Lund, Mainz] * IP is open and vendor-independent, hence should be encouraged as complementary to OSI. * "Be pragmatic." * Do not dilute OSI effort; be patient. [COSINE] * Must we adopt the same mediaeval "federative" (fragmented) approach for IP as for OSI? Why not strive for a uniform pan-European strategy? * Israel is going IP before OSI. [IARN] * GARR and UNINETT both formally support RIPE for short-term coordination. UNINETT foresees a long-term role for RARE. * RARE should participate in RIPE activities (e.g. for IXI). [FUNET] * IP coordination is vital, RARE should do it. Need IP to OSI migration strategy. In particular, should consider OSI applications running over a TCP/IP base, rather than separate stacks. [JANET] * IP can do things that OSI cannot, hence no conflict with RARE goals and policy. RARE could handle IP like it handles IXI. [DFN] * RIPE should continue as a committee of RARE. Infrastructure should be X.25 up to 2 Mbps. [OFFRIR] * RIPE is not a competitor to anybody, nor a service provider. It could fit under a RARE "umbrella". [RIPE] * RARE is wrong platform for IP coordination; let RIPE continue. RARE should tackle general strategy. 3. OPTIONS ---------- A considerable range of opinion has been exposed in the previous chapter, not all of which can be easily reconciled. The present chapter sets out rather summarily the major options we have identified as available to RARE, to set the scene for our recommendations which are inevitably a compromise. 1. "Purist". RARE stays out of IP. Advantage: Simple. Disadvantage: RARE loses all credibility. 2. "Recognition". RARE recognises RIPE activities and approach and is represented in RIPE meetings. IXI-CC perhaps participates in RIPE. RARE arranges appropriate support for cost of RIPE meetings. Advantages: Realistic, keeps "right" people involved. No need to re-invent wheel. Disadvantages: Does not solve optimisation & strategic problems. Slight increase in organisational complexity. 3. "Takeover". RARE replaces RIPE with a new WG. Advantage: Fits RARE structure. Disadvantages: Job of RIPE unsuitable for WG structure (national delegates but international services). RARE/IXI is only one IP actor among many. Probably unacceptable to some actors. 4. "Strategy". A RARE WG plans Europe-wide capacity, topology, and protocol strategy to include both OSI and IP integrated into the same infrastructure, covering the low- and high-performance markets. Should also involve strategy for coping with transition to ISO IP and the Transport Class problem. This approach is compatible with either recognition or takeover of RIPE activities. Advantage: This work is essential! Disadvantage: " " " difficult! 4. RECOMMENDATIONS ------------------ The compromise we propose is that RARE should take the lead in the strategic planning (option 4 above) involving IP and any other non-proprietary protocols as well as OSI. We leave the CoA to decide on the appropriate working group and time limit. However, we believe that RARE would be ill-advised to take the lead in technical coordination of IP, which is already in competent hands, so we prefer option 2 to either of the extremes represented by options 1 and 3. We recommend that RARE should adopt the following statements: 1. RARE recognises the TCP/IP family of protocols as an open multi-vendor suite, well suited to scientific and technical applications, which offers some facilities not available with OSI today. RARE recognises TCP/IP as complementary to OSI and preferable to proprietary protocols for immediate use. 2. A strategy is required to plan future pan-European (including Eastern Europe) networking for the A&R community, covering capacity, topology, open protocols including OSI and TCP/IP, and choice of bearer services. A possible backbone structure and gateway requirements must be planned. Cost estimates and a description of possible funding mechanisms must be developed. RARE WGx is charged to propose the first version of such a strategy by xx.xx.90. 3. Meanwhile RARE recognises RIPE as an appropriate body for current TCP/IP coordination activities, encourages RIPE to continue as planned, and wishes to be represented in RIPE meetings. In addition the IXI-CC should participate in RIPE as a service provider. The REC is requested to arrange support for RIPE travel and subsistence expenses as far as possible within the 1990 budget, and to plan for further support in later years. 4. RIPE should be represented in the CCIRN to ensure adequate intercontinental coordination. (end)