Trondheim 901029 Notes from meeting for NODRUnet ISO CLNS project ------------------------------------------------ Present: G|ran Bengtsson Juha Heinanen Gunnar Lindberg H}vard Eidnes Location/time: Chalmers, G|teborg, 24/9-90 0) Administrative matters H}vard was appointed secretary, Juha as chairman of the meeting. 1) Acceptance of the agenda The agenda was accepted with the addition of point 2.5, Minutes from the last meeting. 2) Juha reported that the final NORDUnet project contract had been sent, but not yet received back. Nonetheless he considered the project to have formally started. 2.5) Minutes from last meeting, review of actions etc. Manning of the project was considered sufficient for now. Later, when the practical activity with DECnet phase V stars, more people should be involved. NSAP allocations: SE - There had been a formal acceptance of guidelines and recommendations for NSAP allocation in Sweden. There had yet to be established a formal organization to handle the requests. NO - As H}vard Eidnes recently started within this area, he had no news to bring. FI - In Finland there is an unofficial organization called "OSI User's group" that handle the NSAP allocations for now. It was noted that there was to be an ISO meeting in Australia "this week", that were to decide on a number of important issues for the CLNS protocols: - An extension of max length NSAPs to 20 octets? - Promotion of IS-IS routing protocol for CLNS to DIS? - Inter-domain routing working paper accepted as a working paper? Juha reported on recent activities within RARE WG4. Juha has argued for the binary representation choice. Anyway, this is still a decision to be taken by the national ISO standardization body. Olav Kvittem had previously noted the importance of trying to get some coordination and/or interworking between the CONS and CLNS communication domains and routing in a mixed environment. There appears to be a DIS on CONS ES/IS, but so far nothing has been produced on the IS/IS level. It was commented that the CONS/CLNS situation is very muddy at present. In connection with this it was mentioned that the CONS/CLNS IWU issues were to be discussed at the Australia meeting of ISO. It was also reported that there had been an European/US meeting on interworking CONS/CLNS (not attended by many DECnet users, though), and it resulted in 3 reports -- short/medium and long term solutions. Among other things RFC1006 was re-worked to carry full NSAPs in packets. As a consequence of this fluid and unclear situation to the IWU issues, it was decided to keep these issues out of the NORDUnet CLNS project for the time being. The CERN field test of DECnet phase V had been still further delayed, so Ingrid hadn't gone there. At the time of this meeting there still had not been decided on a date for the CERN CLNS field test. 3) Status reports from various fronts: RARE WG4, NSFnet, Digital, cisco, etc. The article "Review of ISO IP" by Juha Heinanen was distributed. RARE had contacted Juha Heinanen to urge him to put forward a CLNS pilot project and pilot proposal for RARE. The foreseen configuration will first be: NORDUnet /\ / \ / \ / \ IXI x.25 RIPE x.25 /\ \ / \ \ / \ \ SURFnet SWITCH CERN ! ! ================ CERN ether ! NSS ! v NSFnet This configuration would allow us to get experience with the cisco routing software and its handling of networks with loops... The report on ISO IP was to be discussed by RARE COA. However, Juha didn't expect any COSINE money to be contributed (worries about the involved bureaucracy). However, the CLNS subproject of RARE is still expected to be an official activity within RARE, and Juha was going to push for the next version of IXI to provide a CLNS service in addition to x.25 and CONS. PROMs for cisco 8.2(6) software was said to be on the way, and Juha would test them as soon as they arrived. If the test is successful, he would distribute copies of the proms to the other participants. The setup of the finnish ISO network involves some 4 different routing domains: Tampere University of Technology, FUNET, Datanet and a connection to a consulting company. A question was raised as to how we should run ISO on the NORDUnet backbone. It was thought that running multiple routing domains on the same "ethernet" probably wasn't a good idea. If this turns out to be an inherent problem in the protocols, we could eg. use the swedish IDP and a swedish routing domain on all the NORDUnet routers. In this connection it was noted that the local area part of the NSAP is limited to 2 octets by the current cisco software, because the early IS-IS draft specified the use of 2 octets for this field. There should be prepared a report on how we should organize DECnet phase IV tunneling over the NORDUnet ISO backbone. In the long run we should probably try to get a separate IDP for NORDUnet to aid in the DECnet transition period (?). On the NSFnet side of things, they currently have CLNS operating on all NSSes from Michigan to CERN. NSFnet management had expressed positive interest in the NORDUnet project. There would be no NSS at INTEROP, but work was still progressing towards bringing the CLNS service to the INTEROP exhibition. The connection from the INTEROP exhibition to the outside world was going to be through a Proteon router this year, and those boxes do not yet have CLNS routing. Efforts were still ongoing to perhaps use a separate connection and a cisco router? cisco have still to give a date for an implementation of "true" IS-IS. The protocol had at least to reach DIS status first... As for Digital and DECnet phase V: Various rumours about still further delays are circulating. Now it seems that DECnet phase V will first be released on Ultrix and on DEC routers during Q1 of 91. The DEC nameing service will have to run on Ultrix during the initial start of a DECnet phase V network. There appears that there will be two "waves" of DECnet phase V for VAX/VMS, the first to come during Q1 of 91. The full VMS release is still "one year from now". Among the most significant services that will not be part of the first wave is the DEC Time Service (DTS) and the DEC Name Service (DNS) (thus the requirement for an Ultrix system on the local network). There will be a meeting at CERN in December where the field test will be discussed. Key personnel from NORDUnet should probably attend. There seems to have been a change in the transition strategy: the current plans seem to indicate that it will be possible to simultaneously support DECnet phase IV and V on the same router. There will probably be an FTAM and a VT license bundled with the release of DECnet phase V. In the mean time people should get hold of VOTS (VAX/VMS OSI Transition Services?) licenses (available today for VAX/VMS), probably two for each country, and try ot these services. One should get hold of systems to run these tests on. While discussing systems to try out, it was mentioned that both BSD 4.3 Reno (now out) and BSD 4.4 would contain the OSI network service implementation in the kernel, and would probably be suitable to test OSI networking with. Thus, any old iron (eg. VAX-11/750's) could be put to useful tasks. There also is a dual IS-IS implementation being done by U of Wisconsin for this system. The main problem seems to be able to find time to "play" with these systems... Of other likely software packages to try out, SunLink OSI was mentioned. It still costs quite a lot, but was rumoured to be included in the educational license program of Sun by now. ISODE has an interface to SunLink OSI, so is fairly straight-forward to use. The finns had tried out the Retix x.400 product (runs on top of CLNS), and it seems to work well with PP. This was actually tested to work through two cisco routers. In fact, the finnish x.400 WEP runs PP and CLNS, so it is possible to test using that system over CLNS. 4) NORDUnet CLNS project The timetable was redone once more. The phase with running static routing was removed, as cisco has ISO IGRP implemented now. The phase where we run the true ISO IS-IS was pushed to April-91 and onwards, thus including DEC routers etc. in the prototype. The deliveries and time plan were pushed further into the future, according to the delay in starting the project (April -> October). Subtasks were allocated: G|ran and Ingrid is to produce a report for the DECnet transition with recommended configuration/setup and pitfalls documented. (Pending the acknowledgement by these persons' bosses.) H}vard Eidnes is to produce an NSAP recommendation for NORDUnet. (Also pending acknowledgement by his boss.) The suggested Table of Contents for the whole project report was drafted as follows: - Intro to CLNS - Goals of NORDUnet CLNS project - NSAP format: possibilities and recommendation - DECnet phase V transition plan - Experience with pilot Different routing protocols, conclusions Interoperability, IS-IS, IS-ES Inter- and Intra-domain routing protocols - Appendices config guidelines router configuration examples configuration of end-systems etc. 5) Any other buisness No other buisness to discuss 6) Next meeting During the NETF at Chalmers, in the end of October. We had some time left, and used it to discuss the possible choices for NSAP structure: Org ID, to be decided by national ISO std. body ! ! Routing domain ! ! AFI IDC ! Reserved! Local-area Host NSEL ! ! ! ! ! ! ! ! +--+----+------------+--+----+----+------------+--+ ! 1! 2 ! ! ! ! 2 ! 6 ! 1! +--+----+------------+--+----+----+------------+--+ ! ! +-------- Area -----------+ The AFI is the Address Format Identifier. For NSAPs 38 indicates decimal representation, and 39 indicates binary representation. It was recommended that we use the binary represesentation, since that is also used by cisco, DEC and Sun. The IDC is the Initial Domain Component, usually a country code allocated by ISO. The DEC transition strategy allocates the two last octets of the Area ID for use as a "Local area" identifier. The host number is fixed at 6 octets by IS-IS routing (?), and the NSEL (Network Service selector?) is fixed at 1 octet. Thus, the fixed portion of the address is 12 octet. If one is to have a non-null routing domain (perferably 2 octets) and 1 octet reserved for "future purposes" (different routing protocol etc. ?), a 17-octet NSAP will probably not suffice (17 was the previous recommented maximum size of NSAPs?). If we instead use full length (20 octets) NSAPs and the allocations discussed so far, this leaves 5 octets for organization ID, which should be sufficient for most countries. The Organization ID is handed out by the national NSAP addressing authority (delegated by the national ISO standardization body), and could further be sub-devided into a format identifier to distinguish between different formats of the rest of the NSAPs. The distintion between routing domain and organization and how these should be used is still open to discussion. Should one allocate eg. one org ID to SUNET as a whole, and hand out different routing domains for each university, or should eg. each FUNET member have its own organization ID? Host numbering was also discussed, and this is strictly a local matter. One should decide on some allocation strategy, and stick to it. One possibility was to encode a machine's IP address using 3 digits in the hexa-decimal representation for each octet of the IP address. Another possibility is to use the hardware media address for the host part (this has the undesireable consequence that once one changes the ethernet board, the host has suddenly been renumbered, ie. not such a good idea). Possible pitfalls for DECnet phase IV and V transition and consequences for the NSAP allocation plan have to be sorted out.