Editor's Note: Minutes received 12/9/92 CURRENT_MEETING_REPORT_ Reported by Marshall Rose/DBC Minutes of the SNMP Version 2 Working Group (SNMPV2) The Agenda was reviewed and approved. In the discussion which follows, the decisions reached by the Working Group are summarized. In the majority of cases, there was significant, protracted discussion. In the interests of brevity, that discussion is not reproduced here. Outstanding Procedural Issues were Discussed o Deadline to Finish: Although a meeting slot has been identified for December, the Chair wanted to try to conclude business this week as several slots were scheduled for the Working Group. There was strong consensus that an additional meeting should be avoided if at all possible. o No New Proposals: There was consensus that only ``bug fixes'' and ``show stoppers'' would be addressed after the conclusion of this meeting. The one exception is the row-creation and associated proposals, see III.10 below. o Deadlock Shelf: There was consensus that deadlock shelf would remain in place for proposals for which consensus could not be reached. From time to time, these items will be taken off the shelf to see if there is a new consensus. o More Implementation Experience: There was consensus that no additional implementation requirements would be placed on the documents prior to the Working Group completing its work. Deadlock Shelf: 1. Changing descriptors/enumerations w/o changing object's OID There was consensus that this would not be allowed, because descriptors could be IMPORTed to other modules. As such, a change in either would result in a change in OID. 2. Adding enumerations w/o changing object's OID An action was taken by the editor and Dave Perkins to make a proposal in this area. After discussion, the proposal was rejected and instead there was consensus that adding enumerations w/o changing an object's OID was permissible. 3. LABEL clause 1 There was consensus that this clause was to be dropped. However, a new effort outside of the SNMPv2 Working Group should be formed to investigate related functionality outside of the OBJECT-TYPE macro. 4. Accessibilities of auxiliary objects There was considerable deadlock and intransigence on this issue. Finally, it was observed that the two camps had polarized into agent implementors (not-accessible) and management station implementors (read-only), so the chair decided the issue in favor of the agent implementors: not-accessible was selected. 5. readOnly error-status Since some implementations return this value, the PROTO and COEX documents were updated to reflect this and indicate the appropriate actions to take. 6. RowStatus: a SHOULD or MUST No consensus was reached as Marshall Rose argued the absent Karl's Auerbach's position. As such, the current text, ``should'' remains. III. New proposals: In order to facilitate the discussion, each presenter was required to first demonstrate a problem, before presenting a solution. 1. Tracy Cox of Bellcore demonstrated that delayed operations (e.g., due to slow proxy) was a problem. Discussion of solutions was tabled until after the SNMP Security WG meeting later that evening. At that meeting, two proposals were suggested. As such, this issue has been moved to the SNMP Security WG. 2. Dave Arneson of Cabletron suggested that efficient retrieval of tabular objects was a problem. There was consensus that, in bandwidth-limited environments, retrieval should be more efficient. However, there was no consensus that this problem was specific to tables. 3. Anil Rijsinghani of DEC was absent, but a colleague demonstrated that auto-discovery of SNMP agents was a problem. There was consensus that the proposal was on the right track, but that this work could proceed independently from the SNMPv2 effort. 4. Dave Perkins of SynOptics suggested that retrieval of non-rectangular tables was a problem and there was sufficient interest to look at the solution. However, there was consensus that there wasn't enough of a problem to warrant the solution. 5. Dave Perkins of SynOptics presented his 41 SMI issues. A few of these were postponed to an off-line editing meeting (see Section IV below). Although all of the issues were discussed, in the interests of brevity, only those issues which led to a change in the document are 2 presented: - Module labels, e.g., ``FOO'' in ``FOO DEFINITIONS ::= BEGIN'' must not change across revisions of an information module. - The module revision procedures didn't indicate how to revise invocations of the OBJECT-GROUP macro. - The introductory text for each document will be normalized. - An action was taken by Jeff Case to provide an (approximately two page) introduction to the components in the network management system and their relationship. - Parts of the SMI were re-ordered for ease of reading. - The MODULE-COMPLIANCE and AGENT-CAPABILITIES macros were moved to a new document, ``CONF''. - An unsigned 32-bit integer-valued tagged type was defined. - Full ASN.1 sub-typing, appropriate to the ASN.1 type being refined, is allowed. (This is a clarification.) - The OID-VAL macro for registration assignments was created, but the editor changed the name to OBJECT-IDENTITY. - Text noting that the tagged types for IpAddress and NsapAddress were historical was added. - Clarifications of the AUGMENTS clause were made. - Missing SMI-level coexistence issues were codified. 6. Sam Roberts of Farallon presented his SMI issues. - Various ASN.1 grammar typos in the macro were corrected. - Clarifying text indicating that Counters can not, but that Gauges can be sub-typed was added. - Hyphens are not allowed in descriptor labels, enumerated labels, or the names of textual conventions. (This used to be a requirement ONLY for ``standard'' MIBs, now all MIBs must obey this rule.) The COEX document indicates that these changes may be made without deprecating objects. - Problems with the IMPLIED clause were identified and a solution provided. 7. Anil Rijsinghani of DEC was absent, but a Jon Saperia discussed a need for a unsigned 64-bit type. However, the Group could not achieve consensus on any adequate choke rule. As such, following there was consensus that despite some usefulness, that such a type would not be 3 added. 8. Marshall Rose of DBC described a problem in the definition of the TEXTUAL-CONVENTION macro along with a solution. Textual conventions are now written as ::= TEXTUAL-CONVENTION SYNTAX This is necessary due to macro definition restrictions in ASN.1 9. Jeff Case of UTK suggested that the limitations on enumerated values in INTEGERs was causing problems when translating MIBs written by other groups. There was consensus that the limitations should be removed with a recommendation that newly defined objects follow the old rules. 10. Bill Norton of Merit presented the row-creation portion of the multi-part proposal by Guenther Schreiner, et. al. Discussion lasted for over two hours. Group consensus was that Create/Delete operators were not the solution to row creation, but there is a problem with complexity and multiple ways to use RowStatus. Jeff Case took an action to reconsider this problem. The chair set a deadline of 12/4 for final resolution on this issue and consideration of the other proposals that came with this one. In comparing the row-creation proposal to the RowStatus mechanism, it was agreed that the row-creation proposal did not solve the general problem of row creation, as: 1. Sometimes multiple PDU exchanges were necessary in order to create a row, e.g., either because of a resource negotiation process between the agent and manager, or because there might be too much data to fit in a single creation request. 2. The response from the creation PDU added varbinds in order to indicate what mandatory columns are missing. However, this could make the request too big to send back. 3. The creation request is not idempotent due to potential packet duplication and loss from the underlying transport service (i.e., the request gets duplicated, the first succeeds, but the response is lost, the second fails, and its response is returned.) It was also observed that with the RowStatus mechanism, creation could be done in a single exchange, if the DEFVAL clause was active and the manager did a set to active. However, it was agreed that this text should be made more clear. An action was taken by Steve Waldbusser. After much discussion, there was consensus that the real problem was that the community had three requirements: 1. A single, consistent way to do row-creation. 2. Some row-creations take more than 1 exchange. 3. Some agent writers wish to implement a simple table in such a way so that row creation must be done in a single exchange. An action was taken by Jeff Case on behalf of the four SMP authors to see if some solution could be found which had these properties: 1. Avoided the tooBig problem. 2. Dealt with the discovery problem of missing columns and defvals. 3. Avoided stateful behavior. Jeff Case was careful to stress that this issue had been look at, in great detail 4 by the SMP authors prior to the publication of the SMP specification, and he was doubtful that a solution could be found. IV. Off-line editing: With the approval of the Working Group, Dave Perkins met with the editor to deal with numerous minor issues: - Because groups deal with both conformance and naming, the OBJECT-GROUP macro was moved to the new CONF document to be used for conformance purposes, and the SMI (and MIB and M2M) documents use the SNMPv1 mechanism for naming object groups. - A usage example was clarified. - The intention of textual conventions was clarified. - It was redundantly noted that Counter objects do not have DEFVAL clauses. - An example of ``epoch'' was given for TimeTicks. - The text concerning Opaque type was stream-lined. - The use of the experimental branch was aligned with reality. - When the STATUS clause of an object changes, its DESCRIPTION clause should be updated accordingly. V. Actions Outstanding: - Jeff Case: introductory text - Steve Waldbusser: look at clarifying RowStatus/DEFVAL active text. - Four authors: look at row-creation issues - WG: discuss and resolve the Schreiner, et. al. proposals on Set2Default, short termination of get-bulk, etc. VI. Timetable: There was strong consensus that the row-creation issue and other associated, unresolved proposals. would be given until Friday, December 4 to achieve resolution. There was *complete* consensus that the final deadline for comments on the 9 SNMPv2 documents would be Friday, December 11 Unless the SNMP Security effort raised new issues, then the documents would be sent forward to the IESG with a recommendation for advancement to the standards-track from the Working Group. 5 Finally, it was observed that the SNMPv2 documents could not go forward without the revisions the 3 SNMP Security documents. As such, it was suggested that the membership of the SNMPv2 Working Group now focus its energies on the issues before the SNMP Security Working Group. VII. Documents: Revised versions of the SNMPv2 documents were submitted to the Internet-Drafts area. In addition, ``unofficial'' copies are available via anonymous ftp: host ftp.ics.uci.edu area mrose/snmpv2/ files *.txt The documents can also be retrieved via e-mail: mailbox archive-server@ftp.ics.uci.edu body MIMESEND mrose/mh-mime/snmpv2 These documents will be removed once the actual Internet-Drafts are announced. Attendees Elizabeth Adams adamse@attmail.com Steve Alexander stevea@i88.isc.com David Arneson arneson@ctron.com Toshiya Asaba asaba@wide.sfc.keio.ac.jp Fred Baker fbaker@acc.com Jim Barnes barnes@xylogics.com Brian Bataille bataillebc@afotec.af.mil Andy Bierman abierman@synoptics.com Fred Bohle fab@interlink.com Jack Brown jbrown@huachuca-emh8.army.mil Theodore Brunner tob@thumper.bellcore.com Stephen Bush sfb@ncoast.org Jeff Case case@cs.utk.edu John Chang changj@ralvm6.vnet.ibm.com Szusin Chen szusin.chen@eng.sun.com Robert Ching rching@nat.com Chris Chiotasso chris@andr.ub.com Bobby Clay clay@eagle.msfc.nasa.gov John Cook cook@chipcom.com Tracy Cox tacox@sabre.bellcore.com Juan Cruz juan@dss.com Dave Cullerot cullerot@ctron.com Cathy Cunningham cmc@microcom.com James Davin davin@bellcore.com Michael Davis mad@spirit.clearpoint.com Michael Davison davison@fibercom.com Cynthia Della Torre cindy@gateway.mitre.org Manuel Diaz diaz@davidsys.com Jon Dreyer Jon.Dreyer@east.sun.com Jacques Dugast dugast@issy.cnet.fr Donald Eastlake dee@ranger.enet.dec.com 6 David Engel david@ods.com Michael Erlinger mike@jarthur.claremont.edu Roger Fajman raf@cu.nih.gov Daniel Fauvarque dfauvarq@france.sun.com Karen Frisa karen.frisa@andrew.cmu.edu Shari Galitzer shari@mitre.org Shawn Gallagher gallagher@quiver.enet.dec.com Richard Graveman rfg@ctt.bellcore.com Maria Greene mngreene@eng.xyplex.com Michel Guittet guittet1@applelink.apple.com Robert Gutierrez gutierre@nsipo.nasa.gov William Haggerty haggerty@ctron.com Patrick Hanel hanel@yoyodyne.dco.ntc.nokia.com Ed Heiner eah@pau.synnet.com Gerd Holzhauer holzhauer1@applelink.apple.com John Hopprich hopprich@davidsys.com Jeff Hughes jeff@col.hp.com David Husak dave@synnet.com Robin Iddon robini@cix.compulink.co.uk Kevin Jackson kmj@concord.com Ole Jacobsen ole@interop.com Ronald Jacoby rj@sgi.com Frank Kastenholz kasten@ftp.com Mark Kepke mak@cnd.hp.com Zbigniew Kielczewski zbig@eicon.qc.ca Jong Yeol Kim kimjy@ring.kotel.co.kr Andrew Knutsen andrewk@sco.com Michael Kornegay mlk@bir.com Deirdre Kostick dck2@sabre.bellcore.com Michael Laufer mlaufer@bbn.com Mark Lewis mlewis@telebit.com David Lin lind@janus-ccm.zenith.com David Lindemulder dcl@mtung.att.com Benjamin Lisowski Ben.Lisowski@sprint.sprint.com David Liu dliu@bnr.ca John Lunny jlunny@twg.com Carl Madison carl@startek.com Keith McCloghrie kzm@hls.com Evan McGinnis bem@3com.com William McKenzie mckenzie@ralvma.vnet.ibm.com Donna McMaster mcmaster@synoptics.com John Medicke medicke@ralvm11.vnet.ibm.com Douglas Miller dmm@telebit.com David Minnich dwm@fibercom.com Mohammad Mirhakkak mmirhakk@mitre.org Rohit Mital rm@protools.com George Mouradian gvm@arch3.att.com Patrick Mullaney mullaney@ctron.com Daniel Myers dan@nsd.3com.com Rina Nathaniel rina!rnd!rndi@uunet.uu.net Hien Nguyen h.nguyen@sprintintl@sprint.com Mo Nikain mo@bss.com Tom Nisbet nisbet@tt.com Bill Norton wbn@merit.edu 7 Steven Onishi sonishi@wellfleet.com David Perkins dperkins@synoptics.com Carl Powell cpowell@bbn.com Ilan Raab iraab@synoptics.com Richard Ramos ramos@mtunm.att.com Venkat Rangan venkat@geoduck.matrix.com Louise Reingold l.reingold@sprint.sprint.com Sam Roberts sroberts@farallon.com Kary Robertson kr@concord.com Dan Romascanu dan@lannet.com Marshall Rose mrose@dbc.mtview.ca.us Shawn Routhier sar@epilogue.com Chris Rozman chrisr@usr.com Assaf Rubissa asaf@fibhaifa.com Jon Saperia saperia@tcpjon.ogo.dec.com Michael Sapich sapich@conware.de Michael Scanlon scanlon@interlan.com Sam Schaen schaen@mitre.org John Seligson johns@ultra.com Paul Serice serice@cos.com Chris Shaw cshaw@banyan.com Timon Sloane timon@rahul.net Robert Snyder snyder@cisco.com Joo Young Song jysong@ring.kotel.co.kr Roy Spitzer roy.spitzer@sprint.com Einar Stefferud stef@nma.com John Stephens john@cayman.com Bob Stewart rlstewart@eng.xyplex.com Kaj Tesink kaj@cc.bellcore.com Dean Throop throop@dg-rtp.dg.com Ahmet Tuncay atuncay@synoptics.com Warren Vik wmv@i88.isc.com Ioannis Viniotis candice@ececho.ncsu.edu Steven Waldbusser waldbusser@andrew.cmu.edu Alice Wang alice.wang@eng.sun.com James Watt james@newbridge.com Luanne Waul luanne@wwtc.timeplex.com Gerry White gerry@lancity.com Peter Wilson peter_wilson@3com.com Steven Wong wong@took.enet.dec.com Randall Worzella worzella@ralvm29.unet.ibm.com Daniel Woycke woycke@smiley.mitre.org Honda Wu honda@nat.com Jeff Yarnell jeffya@protools.com Kiho Yum kxy@nsd.3com.com 8