CURRENT_MEETING_REPORT_ Reported by Robert J. Reschly, Jr./BRL NJM Minutes Agenda Old Business o Unexpected routing. - Reports - Operational Impact - Action * Is there anything which should be done? * Is there anything which can be done? o Other old issues - Communities? New business o Dale Johnson on trouble tickets. Roundtable on current and expected issues o Effects of development of Internet - Scaling - Speed - ``low budget'' connections, users? - International network coordination and mgt., etc. After a brief review of the function of the NJM, there was another call for ``unexpected routing'' anecdotes. The University of Delaware to DuPont Delaware via Ithaca, NY and Reston, VA, and WestNET's 16 hops across town routes were cited as examples. Also cited was the TWB routing problem due to that router being connected directly to the campus. 1 Others mentioned examples which were found to be a result of MILNET problems, and one situation involving Argonne. All were understood and have been or are being corrected. The subject of diagnosing routing problems came up. Traceroute, especially third-party traceroute where available, still seems to be the most heavily used tool. Tony Hain of ESNET informed those present of the community name for ESNET's routers. This is strictly for use by other midlevel network operators in the performance of their tasks. Others with a requirement to access these routers should contact Tony. NSI is considering making it's community name available as well. Dale Johnson briefly outlined this week's discussions concerning NOC trouble ticket systems. He has a draft (draft-ietf-ucp-tt-00.txt), inspired by the UCP WG document (draft-ietf-ucp-connectivity-00.txt). He feels that their focus on accountability to end user problem reports and single NOC operations is not totally suitable for his purposes. Dale is more concerned with inter-NOC network oriented operations. Worth noting is that the TT discussions revealed a desire to make this more universally useful -- i.e., by central site staff as well as NOC staff. Dale will be publishing an updated document in a few weeks. When questioned about whether any systems were going to be proposed, he responded affirmatively. As a point of information, Gene Hastings stated he felt the real goal of the the UCP paper was the establishment of an inter-NOC transaction processing system for handling the passing of problem reports between NOCs. MERIT currently runs an IBM mainframe product, but is moving towards a UNIX based TT system they may develop locally. IBM Yorktown is working on xgmon; Tim Salo at MSC is funded to work on a UNIX implementation; and Sun Microsystems is working on one as well... Word on developments will be sent to the Trouble Ticket Requirements mailing list (-request for administrivia) as it becomes available. There was quite a bit of talk about the pros and cons of basing a TT system on top of a DBMS. It is very easy to expend man-years of effort in the design and integration of a DB based system -- time many organizations simply do not have. A suggestion that we encourage some company to produce and support a TT system was generally well received. It was also observed that in many cases, the integration of a TT system was going to involve some DB customization/interface work in any event. A poll was taken about current TT operations. 10 sites have some sort of online TT system (4 were ASCII --['sensibly' printable]); 5 were paper systems; and three people reported having no formal TT system in use. Someone noted there were two publicly available systems (are these 2 in NOCTOOLS?). Conversation then moved on to the desirabilty of having links to other portions of any existing DB -- examples involved things like specification of a router filling in configuration information, and mentioning a pair of routers completing link and telco contact information. Again it was noted that this was a bigger win when the ``external'' components already existed. It was observed that there must be products available which solve similar problems in areas like inventory control, but that they were not necessarily TT oriented. Unfortunately nobody could cite specific systems. There was a call to formalize an operations track within the IETF. Having this track would reduce internal schedule conflicts, and should attempt to minimize conflicts with User Services as the two have significant overlap. The group then dove into an extended discussion of the undesirability of referring all problems up towards MERIT. Members very much wanted the ability to contact relevant parties in other regionals directly, but expressed frustration at lack of contact information. Many rely on one or more of the Internet Managers Phonebook, WHOIS, or stabs into the DNS, but these often are only approximate reflections of reality. One proposal was the addition of text/hinfo records incorporating contact phone numbers. Doug Gale is working on an NSF RFP for global user services.. [something about a help server at MERIT -- call (800)66-MERIT and ask about the help server]. There was a suggestion to add DNS records for networks as well as hosts (e.g., lookup on 128.63.0.0 -- forward and inverse), along with a warning that any records should match networks.txt. Milo Medin had some comments concerning the new DDN NIC contract. The new contract does not provide for network number assignment or DNS registration among other things. [Later, Steve Wolff told us that DCA and NSF are working together to ensure the continuity of essential services.] More information will be sent to the mailing list as it becomes available. Kannan Varadhan then touched on his ongoing Telebit NetBlazer testing. He has developed a list of things he wants to discuss with Telebit, and solicits questions from others. The NJM mailing list (-request for administrivia) will host the dialog with Kannan as his testing continues (i.e., post your questions and answers to this list). 3 The basic NetBlazer is a 386 box running KA9Q, with 2 modem ports for a total cost of ~$3,000.00. Additional ports are added in 8 port increments. It offers packet driven dialup, and three authentication methods: username/password; callback; and, between boxes, a crypto handshake. NetBlazer does not do TACACS. The TACACS comment prompted a number of requests for some sort of authentication servers which may (at least optionally) be Internet-wide in scope. Dale Johnson mentioned in passing that MERIT had just deployed one for MICHNET. Milo then talked briefly about NSI's plan for having a single 800 number for his folks on travel. When called, this number would route to a hunt group of lines local to that area. He also mentioned that it was still possible to assign fixed IP addresses with this and still have routing work (under OSPF if it was a single area -- OSPF used best match.). After the discussion was wrenched back to the agenda, it was asserted that overall European routing is a disaster, even if internal (i.e., ESNET or NSI European routing appeared to be sensible). Dave O'leary noted that in many cases routing was set based on technical considerations even when they conflicted with policy considerations. SURA continues to take heat on this issue. It was felt that the FEPG/FRICC work would help. The FEPG has developed guidelines which formalize connectivity in accordance with CCIRN recommendations. At this point Milo insisted that NOCs contemplating international operation absolutely positively must have 24 x 7 NOC operations. We were told that SPRINT and Cornell (the NSF International connection managers) want to schedule a global BGP, coordination and cutover meeting. The current best guess has this meeting taking place at the July IETF in Atlanta. Someone wondered if the decisions were unilateral or bilateral. The IEPG is a technically oriented group doing sensible things, but it is not clear the IEPG is in a position to significantly affect the decision process. Their next meeting is in Paris in early May. It was also noted that many of the problems appeared to be intra-European. We then moved on to a very brief consideration of what connecting hordes of high schools would entail. A quick survey showed three regionals are planning to connect 10 or more high schools in the coming year, and in at least one case, these connections will connect whole districts. The humor quotient chose that time to take a significant nosedive so we adjourned. 4 Attendees Vikas Aggarwal vikas@JVNC.net Jordan Becker becker@nis.ans.net Tom Easterday tom@cic.net Dale Finkelson dmf@westie.unl.edu Vince Fuller vaf@Standford.EDU Shari Galitzer shari@gateway.mitre.org Joseph Golio golio@cray.com Eugene Hastings hastings@psc.edu Steven Hunter hunter@es.net Dale Johnson dsj@merit.edu Dan Jordt danj@nwnet.net Daniel Long long@nic.near.net Milo Medin medin@nsipo.nasa.gov Bill Norton wbn@merit.edu David O'Leary oleary@sura.net Robert Reschly reschly@brl.mil Ron Roberts roberts@jessica.stanford.edu Roxanne Streeter streeter@nsipo.nasa.gov Kannan Varadhan kannan@oar.net Ross Veach rrv@uiuc.edu Carol Ward cward@spot.colorado.edu C. Philip Wood cpw@lanl.gov 5