X-Sender: jtw@ginger.lcs.mit.edu Mime-Version: 1.0 Date: Fri, 12 Jul 1996 17:02:45 -0400 To: minutes@CNRI.Reston.Va.US Sender: minutes-request@ietf.CNRI.Reston.Va.US From: John Wroclawski Subject: Montreal IETF minutes Cc: int-serv@isi.edu Montreal IETF Proceedings Transport Service Area INTSERV Working Group Minutes of the Integrated Services Working Group (INTSERV) Wednesday, June 26 (one session) Reported by: John Wroclawski, MIT LCS The goal of the meeting was to address details remaining before submitting the guaranteed and controlled-load service standards and related documents to the IESG as potential proposed standards. Craig Partridge began the meeting by presenting a plan proposed by the Transport Area director and WG chairs. The purposes of the plan are to ensure that developers and vendors interested in INTSERV and RSVP have clear, crisp guidelines about what to build, and to ensure that the community gets the experimental experience needed to evaluate the current work and move forward. The plan has two parts: 1) Coordinated IETF last-call of RSVP and INTSERV documents by July 22. - RSVP WG products: RSVP, RSVP MD5 Integrity, RSVP Security - INTSERV specifications Guaranteed, controlled-load svc specs. General characterization params, data formats. - INTSERV informational Template document, new glue document. 2) Impose a moratorium on new standards-track services until first review of existing proposed standards, to allow implementors and the market to gain necessary experience with the current proposed standards. It was emphasized that this is not intended to discourage experimental work on new services which may be needed in the future. Craig then briefly reviewed the status of the relevant INTSERV documents. - Guaranteed service spec - approved by WG for submission as P-S at the previous (LA) IETF with minor revisions. New draft incorporating revisions available. Authors request that any further comments be sent to them immediately. - Controlled-Load service spec - approved by WG for submission as P-S at the previous (LA) IETF with minor revisions. New draft incorporating revisions available. Authors request that any further comments be sent to them immediately. - General characterization parameters spec - new draft available. Some issues arose, to be discussed later in the meeting. - Data formats spec - unchanged since approval at LA meeting. - Template document - new draft available, some relatively minor changes and clarifications. This is destined to be an informational RFC. - New glue document - intended to explain how & why pieces fit together. To be written. John Wroclawski presented the revised general parameters document. This document describes path parameters not related to a specfic service. The discussion touched on five points. - PATH MTU calculation parameter - noncontroversial. - PATH_UNPROTECTED (not covered by reservation) flag. This was changed from a count to a flag in the current draft, due to the doubtful usefulness and difficulty of calculating a count. The change was questioned. The resulting discussion reconfirmed the decision to make the change. - PATH BANDWIDTH parameter. John pointed out several issues regarding the exact meaning, usefulness, and feasability of calculating this parameter, and raised the question of whether it should be deleted. After some discussion, it was agreed to keep the parameter with a significantly refined definition in the next draft. - PATH LATENCY parameter. John reviwed the definition of this parameter, which is intended to compute the *minimum* path latency for use in conjunction with guranteed and similar services. Several questions about the exact usefulness and definition of the paramter were raised, and a long and somewhat confusing discussion followed. Agreement was reached to a) keep this as a general paramter, b) attempt to significantly clarify the definition in the next draft of the document, and c) throw much of the problem to the ISSLL WG. - PARAMETER NAMESPACE. A proposal to change the parameter namespace somewhat so that a specific service may override the value of a general characterization parameter was presented. There was little discussion, and it was agreed to make the necessary changes to various documents. This concluded the portion of the session devoted to completing current work. The meeting then turned to a presentation and discussion of Roch Guerin's Committed-Rate service specification draft. Roch presented this service as lying at a point between controlled-load and guaranteed, and targeted at applications which will never exceed their stated TSpec and do not want the ability to send bursts through the network. After the presentation, a long discussion developed, primarily focused on the level of distinction between this proposal and the controlled load service. No definite conclusions were reached. The meeting concluded with a brief discussion of the group's future direction. Possibilities raised include going dormant until time for initial review of the existing proposed-standard work, addressing remaining issues closely related to the present work (e.g., service heterogeneity) before resting for a while, and refocusing activity on aspects of integrated services intended to provide better support for traditional best-effort applications.