RSWG

Summary

The RSWG met to discuss several key items concerning the RFC series, including guidelines for authorship assignment, formatting mathematical content in RFCs, and the future of the "updates" metadata tag.

A draft proposing ten principles for RFC authorship sparked a lively debate on the role of generative AI, the boundaries of author responsibility vs. tool disclosure, and how stream-specific policies interact with RPC-level enforcement. The working group agreed to send the Math in RFCs draft (draft-editorial-rswg-mathinrfcs) to the RSAB following a minor text amendment. The discussion on redefining the "updates" tag was put on hold (in abeyance) pending a broader, future discussion on "Evolving RFCs." Due to time constraints, the main session on Evolving RFCs was postponed, and the chairs planned to schedule an interim meeting.

Key Discussion Points

1. Chair Introduction & Evolving RFCs

Discussion: Rich Salz opened the session with the Chair slides. He noted that while Evolving RFCs are a key area of upcoming work, the IETF's internal standardization process is out of scope for the RSWG. He suggested hosting an interim meeting with the IESG to coordinate on how the IETF stream would map to any newly evolved formats.
Process Coordination: Jay Daly and Joel Halpern (via chat) pointed out that under RFC Editor Model Version 3, coordination with the streams is chartered to occur through the RSAB and stream representatives. Russ Housley clarified that early communication with stream leadership is critical to avoid past integration issues (e.g., the previous "new track" effort). The chairs agreed to engage the RSAB directly.

2. Guidelines for Assignment of RFC Authorship

Presentation: Brian Carpenter presented the Guidelines for Assignmentof RFC Authorship. He introduced ten draft principles designed to establish stream-consistent expectations for authors, editors, and contributors, including provisions for acknowledging generative AI-generated content.
Debate on Principles and Conventions:
- Eric Rescorla strongly contested several of the principles, arguing they represent subjective conventions rather than rigid ethical mandates. He noted that common publishing practices (like ghostwriting, academic advisor credits, or reusing boilerplate payload structures without explicit citation) contradict the draft's strict rules on creative contribution.
- Colin Perkins defended the principles, arguing that authorship should strictly reflect a creative contribution and that hidden ghostwriting or uncredited authorship is indeed a issue the community should avoid.
Generative AI and Tool Attribution:
- Christopher Janz raised concerns about where to draw the threshold for AI attribution, noting that AI is increasingly embedded as a standard tool for translation, grammar check, and draft structuring.
- Jay Daly observed that a growing challenge involves participants using AI to generate complex technical text that they themselves do not intellectually understand, which poses a threat to technical standards integrity.
- Martin Thomson and Mark Nottingham argued against mandatory AI disclosure. They explained that mandatory disclosure is a "cat and mouse game" that is hard to define and police. Instead, they advocated for framing the issue around author responsibility—meaning authors are fully accountable for the entirety of the text they sign their names to, regardless of how it was generated.
- Roman Danyliw noted in the chat that the IRTF already has established guidelines regarding generative AI disclosure under RFC 9775, showcasing that streams already differ.
Enforcement Feasibility:
- Jean Mahoney (RPC) stated that the RPC cannot police or verify who made creative contributions, or check for undisclosed AI text. She noted that these verifications are impossible to perform at the final publication queue stage.
- Eliot Lear and Dhruv Dhody agreed that the enforcement of authorship and ethical guidelines must be managed within the individual streams (e.g., by WG chairs, ADs, or stream managers) long before a document is sent to the RPC.
Way Forward: The chairs suggested that Brian Carpenter reframe the draft to focus primarily on "author responsibility" (as articulated by Martin Thomson) rather than trying to regulate specific tools or attribution types.

3. Math in RFCs

Discussion: The chairs reviewed the proposed XML and text styling guidelines from the Math in RFCs slide deck (draft-editorial-rswg-mathinrfcs).
Proposed Amendment: Eliot Lear proposed a minor amendment to the draft text to avoid setting strict procedural constraints on the RPC. He suggested changing:

"...the RPC is expected to adjust their requirements on this..."

to:

"...the RPC may adjust requirements on this as they gain experience."
Consensus: Jean Mahoney (RPC) and Martin Thomson supported this change. Rich Salz and Jay Daly agreed that using "may" preserves necessary RPC discretion when addressing formatting limitations, without creating a hard rule that the author's format preferences must always override layout limitations. The group reached consensus on this change.

4. What's about the "updates" tag?

Presentation: Mirja Kühlewind presented What’s about the “updates” tag?. The draft proposes splitting the poorly defined "updates" metadata relationship into three precise, machine-readable tags: amends, extends, and see also.
Skepticism: Colin Perkins and Eric Rescorla questioned the utility of the split, suggesting that the working group would spend significant effort trying to define and police the difference between "amends" and "extends" without resolving the core disputes. Colin noted that, in thirty years of participation, he has not observed the current "updates" tag causing significant real-world harm.
Implementer Value: Suresh Krishnan defended the effort, stating that the reverse relationships (e.g., "updated by") are incredibly useful for implementers looking at immutable, published RFCs who need to know if a critical change has occurred.
Abeyance: Eliot Lear suggested holding this draft in abeyance because the entire concept of metadata relationships and document updates is heavily tied to the "Evolving RFCs" effort, which explores document mutability and versioning. Mirja Kühlewind and the chairs agreed to place the document on hold until the direction of the Evolving RFCs work becomes clearer.

Decisions and Action Items

Decision: The Math in RFCs draft (draft-editorial-rswg-mathinrfcs) is approved as amended by Eliot Lear's text change ("may" instead of "is expected to"). The chairs will submit this completed document to the RSAB.
Decision: Mirja Kühlewind’s "updates" tag draft is officially put on hold (in abeyance) pending the outcomes of the Evolving RFCs discussion.
Action Item: Jean Mahoney (RPC) to draft an outline/list of existing RPC policies and conventions regarding author/editor handling (such as procedures for deceased, missing, or non-responsive authors) to help level-set the authorship discussion.
Action Item: Brian Carpenter to revise the authorship guidelines draft, reframing its core principles around "author responsibility" instead of tool/AI disclosure.
Action Item: The chairs to issue a mailing list poll to schedule an interim RSWG meeting to discuss the Evolving RFCs initiative.

Next Steps

Poll for Interim: A poll will be distributed to the RSWG list to schedule an interim session specifically dedicated to Evolving RFCs.
RSAB Submission: Submit draft-editorial-rswg-mathinrfcs to the RSAB.

Automatic IETF Minutes