NFSV4

Summary

The NFSv4 Working Group met to discuss the status of active drafts, outline a more structured process for document adoption, and review technical progress on uncacheable directories and Flex Files v2. Key highlights include the progression of the internationalization draft through IETF Last Call, proposals to restructure and split up large working group documents (such as the Security draft and Flex Files v2) to make them reviewable, and plans to address outstanding items on the mailing list rather than holding interim meetings.

Key Discussion Points

Working Group Management and Document Adoption

Brian Pawlowski and Chuck Lever presented the Chair slides and discussed the working group's current document queue, referencing RFC 7221 regarding participation and progress.

Document Backlog & Carrying Capacity: Brian Pawlowski expressed concern over the high volume of active drafts and the limited bandwidth of working group participants to actively review and implement them. David Black and Sorin Faibish agreed that the working group and the IESG have a limited "carrying capacity" and cannot process too many complex documents at once.
Adoption Rules: Brian Pawlowski emphasized that silence does not equal consensus. Future adoption calls will require active feedback, statements of intent to review, and implementation commitments.
Security Draft Restructuring: A significant discussion took place regarding the pending individual security draft associated with the draft-ietf-nfsv4-rfc8881bis respecification effort:
- Gorry Fairhurst (AD) cautioned that a 150-page security document containing a mix of problem statements, considerations, and normative requirements would face severe hurdles during IESG and security area reviews. He suggested separating standards-track normative requirements from informational security considerations.
- Christoph Hellwig argued that separating normative security aspects from the main protocol document (draft-ietf-nfsv4-rfc8881bis) is structurally problematic and that normative changes should remain integrated or tightly coupled. He also noted that the current draft is difficult to review without clean diffs.
- Tom Haynes found the current security document unreviewable and argued that using interim meetings to make progress is unsustainable.
- Chuck Lever noted that most of the document's proposals represent current practice rather than new protocol features. He suggested that proposed AuthSys changes are RPC-wide and belong in a companion document to RFC 9289 (RPC-with-TLS) rather than an NFS-specific document.
- David Black counter-argued that AuthSys discussions are too fundamental to NFSv4 security to be completely omitted.

Uncacheable Directories

Tom Haynes presented Uncacheable Directories: WGLC status to address issues raised during the Working Group Last Call (WGLC) for draft-ietf-nfsv4-uncacheable-directories.

Caching Strategy Options:
- Option A: Disable reader attribute caching across the board.
- Option B: Keep caching as a per-deployment opt-out.
- Option C: Require deployments to use directory delegations.
Tom Haynes advocated for Option B due to scaling and performance concerns with directory delegations.
David Black and Christoph Hellwig agreed that Option A is not viable due to 25 years of existing client implementations, focusing the technical choice between Option B and Option C. Christoph Hellwig warned of potential "dragons" when mixing READDIR and READDIR-PLUS attributes and suggested adding clarifying language to the draft.

Flex Files v2 and Proxy Server

Tom Haynes presented Flex Files v2 + Proxy Server, summarizing progress on a user-space prototype (RefFS), a new trust model using "trust state IDs," and erasure coding implementations.

Erasure Coding & Patents: Tom Haynes proposed Reed-Solomon Vandermonde as the Mandatory-to-Implement (MTI) encoding. Brian Pawlowski made a patent disclosure on behalf of Hammerspace, noting that Hammerspace acquired patents related to Mojette encoding via its acquisition of Roseau FS. He indicated that Hammerspace would provide a formal disclosure but did not expect it to present a concern for the IETF.
Standardization of Encodings: Christoph Hellwig urged the group to stick to a small family of algorithms already established in open-source storage projects (e.g., Linux RAID6 or SnapRAID) rather than proprietary or divergent variants.
Interoperability and Proxy Servers: David Black raised concerns about whether a single MTI encoding is sufficient if clients and servers use divergent encodings. Tom Haynes suggested using a proxy server to translate encodings for legacy clients. David Black expressed skepticism regarding the performance impact of proxy translation in production environments. Christoph Hellwig pointed out that systematic erasure coding allows direct, non-degraded reads without decoding, mitigating some performance concerns.
Draft Split Proposal: Due to the large size of the draft (216 pages), Tom Haynes proposed splitting it into 8 distinct documents, separating specific encoding algorithms into individual standards-track documents. David Black supported this, noting that the security area frequently uses this pattern. Christoph Hellwig agreed, suggesting that basic striping (zero encoding) should remain in the core document while complex encodings are split out.

NFSv4.1 Respecification Effort

Due to technical issues with Meetecho, David Noveck was unable to join the session to present Needed Steps for NFSv4.1 Respecification Effort.

Decisions and Action Items

draft-ietf-nfsv4-uncacheable-directories: Chuck Lever will open a second Working Group Last Call (WGLC) within 1 to 2 weeks to close out outstanding technical feedback.
Security Draft: The chairs will initiate a structured discussion and adoption call on the mailing list to determine if the security draft should be adopted, edited down, split, or folded back into the main draft-ietf-nfsv4-rfc8881bis document.
Flex Files v2: The call for adoption and the proposed document-split strategy will be taken to the mailing list.
NFSv4.1 Respecification (draft-ietf-nfsv4-rfc8881bis): Rather than scheduling an interim meeting, the chairs decided to prioritize moving the respecification discussion to the mailing list to encourage wider participation.

Next Steps

Brian Pawlowski will post a summary/write-up of the recent testing status from the last bakeathon to the mailing list.
Brian Pawlowski (on behalf of Hammerspace) will follow up with more detailed information regarding the Mojette encoding patent situation.
Tom Haynes will incorporate feedback regarding client interoperability requirements and standard open-source encoding families into the next revision of the Flex Files v2 documents.

Automatic IETF Minutes