Minutes from NFS Version 4 working group meeting at 54th IETF in Yokohama, Japan on July 17, 2002. ------------------------------------- Network File System Version 4 (nfsv4) ------------------------------------- Chair(s): Robert Thurlow Brian Pawlowski Document Editor: Spencer Shepler AGENDA: Wednesday July 17: 13:00pm - 15:00 (2 hours) Welcome and Introduction (Pawlowski) 5 min Agenda bash Note Well/Blue Sheets Agenda Bashing Status, milestones, Bakeathon, stuff (Pawlowski) 10 min Brian went over milestones and how we're doing. NFSv4 - July 1 draft resolves most issues; a few substantive ones left. Will have a new draft out after August Bake-a-Thon (UMich), and will do WG last call. See the presentation. Open issues on spec discussion (Brittle) 15 min Dave Brittle presented specification and V4 issues discussion. - Over the wire protocol changes mostly complete - Most issues are "nice to resolve" - Must do resolutions - about 7 or 8 items? (beepy covered a few items on list to illustrate the nature of most of the open items as clarifications or error code additions). Migration/replication (Thurlow) 25 min Rob Thurlow presented migration/replication status. - Security is the primary issue in the design considerations discussion. - Compression? An option at best says Scott Bradner, perhaps best left for a separate document. - General issue on bandwidth reduction techniques. Julian Satran asked whether the general techniques are necessarily applicable to migration? Rob talked about zero block elimination, similar block factoring. Long discussion... twixt Rob and Julian. Question (Julian): what will you do about bandwidth reduction and whether there are different answers for migration and different types of replication e.g. continuous replication of e-mail different from databases. A: there are certainly differences, haven't thought about it that much yet. NFS and RDDP work (Black) 15 min David Black discussed RDDP working group (meets 13:00 - 15:00 Thursday) beepy asked David to discuss RDDP and its application to NFS Version 4. - Overlap with RDMA Consortium asked John Hufferd? Black said expects IETF to absorb. - Overlap with SNIA NFS RDMA work? asks David Brittle? Black responded that we should look at to what the right forum is moving forward. When and how to transition to IETF. Scott Bradner said RDDP is new group - if there are requirements out of NFS V4 then we need to feed into RDDP (same for iSCSI). Julian Satran stated NFS and SCSI are different and that NFS will impose restrictions on RDMA/RDDP use - but, hmmm... beepy pointed to DAFS as an example that solutions for "NFS" exist. RPC and NFS changes must be reflected in the working group specifications. More from Rob's notes: Q: RDMA Consortium? A: Original charter was to transfer its work to others, should be possible now. Q: NFS over RDMA WG at SNIA - relationship? A: AD needs to try to manage, want to avoid per-protocol shims. Should get NFS requirements fed to RDDP. Q: iSCSI and NFS seem different in nature, can they feed compatible requirements into RDDP? A: Hopefully, TBD. Any work on NFS or RPC needs to come back through NFSv4 WG, many hope to limit all impact to RPC layer. Q: Coexistance with older clients? A: Probably only via backwards compatible negotiation, with benefits restricted to new client + new server. Review of work items (Pawlowski) 15 min - Implementation Guide (info RPC) - API GSSAPI advancement - MIB draft - RPC/XDR/RPCSEC_GSS RFC plans Dave Brittle want IANA to pick up RPC numbering - Thurlow to take forward, replacing person at Sun. Scott Bradner says we need to write an RFC for IANA consideration. Backlog of RPC numbering requests. Next work: Implementation guide - bucket to catch some clarifications. API advancement draft - waiting for comments. SNMP MIB draft - want to resurrect, looking for volunteers. Inherited standards - e.g. GSS-API must advance to permit us to advance NFSv4, and need to check on the IANA owning RPC numbers: need to get this on track, Rob Thurlow has this from Brent. Need to read the RFC that says how to write IANA considerations and just do it. Open discussion (Pawlowski) 30 m Wrapup (Pawlowski) 5 m