Minutes for NFSV4 at IETF-92
minutes-92-nfsv4-1

IETF 92 - nfsv4 Working Group - Dallas, TX
March 26, 2015

=== Agenda Bashed (beepy)

=== NFS V4.2 (Haynes)

Working implementations out there. Finding issues. Server side copy.
Added state id, removed NFS URL. Slide 1.

Slide 2. Tom suggests we include both clone and copy - even though
it is 

=== RFC 7530 and 7531 are ready to be published

=== FlexFiles

Good: Three implementations in progress.

Bad: WG last call found confusion on uid/gid for fencing...

Ugly: Fencing interactions with kerberized access...

The protocol requires, customer decides... Not on object or block requires
Kerberos... - Trond

Spencer: How do plan on closing on requirements. 

=== Layout Types

Past WG Last Call, changed from info to std.

Really great review of the draft from Dave Noveck, address and go through
another working group last call.

LNFS set for April 9 for telechat.

=== FedFS (Lever)

Two documents, see slide. All authors ready to move forward.

Couple docs from Chuck, LDAP related, need review. Expertise gap.
AI: Chuck to send out pointers to mailing list... and reminder to review
docs.

=== Updating the NFS RDMA Standards (Lever)

Clarifications needed and address inadequacies...

Bulk payload... bindings for V2 and V3, and NFS V4.1.

Really, see the slides... Chuck did a great job of making them
readable...

NOMSG - gray areas in specification... we are suffering under some
incompatibilities/inconsistencies? Are there. RFC 5666 - isufficiently
specified - David Black says errata insufficient, rev the spec.
Non-interoperable - rev the spec...

Tom Talpey - spec good for V2 and V3, not for V4. Need an update.
(Richard relaying comments). 

Page 10 of slides talk about RFC 5667 fixes for V4.

(Richard for Tom) - since 5667 written, compound operation evolved.
Tom says replace entire document. Tom and Chuck to work on bis ddocument.

Slide 13: Matt Benjamin - any particular implementation (???) - unanswered
question.

No upper layer binding for NFS V4.2.

Decision: Slide 14: Preference for approach 1. Move to mail list...

Slide 16 - persistent memory technologies and iSER etc. Radar screen?

AI: Tom/Chuck to send Outline needed soonish. Take this to alias
to prep for next meeting.  Keep ball moving on this.

AI: Tom/Chuck Communicate that we will bis both docs to unsnarl incompatibility.

AI: Tom/Chuck Create experimental docs as needed for proposed new features.

AI: Include V4.2 Upper Layer binding in 5667bis (verify this) - Chuck/Tom.

=== RPC/RDMA Bi-direction (Lever)

Prototype built for two approaches. 

Sidecar transports work already, would require trunking.

Second approach, is a single connection - complex changes may be required.
Prototyping being done in Linux and Solaris implementations.
(Trond weighed in here on need to do this approach).

See slides... Slide 14 defines challenges found in the prototype.
May require protocol change... 5666 RPC over RDMA changes.
New add on doc or bis open question once the changes scoped...

David Black - downrev behaviour. RFC .... Tom says does not
require a doc change?

AI: Chuck/Tom Talpey to propose protocol changes that are experimental
and keep out of bis doc.

AI: Tabled topic on whether we need upper level bindings for new minor
versions. Revisit. Tom raised issued.

=== NFS Version 4 Management (Noveck)

Pretty complete slides.

Large spec management - took four years for 100 pages...

Review the version management doc.

Send out tickler to list - Noveck - asking for feedback...

=== Adopting WG document (beepy/Shepler/Haynes)

Christoph document to be adopted.

Continue to vet XATTR doc on alias... fish or cut bait. Requirement,
now discuss technical merit. Spencer or I will contact authors.
Personal draft.

=== State of the migration document

AD is working on the document, the document is in queue