Just-In-Time RFC Publication

Status of This Memo

This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶

Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶

Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."¶

This Internet-Draft will expire on 14 August 2024.¶

Copyright Notice

Copyright (c) 2024 IETF Trust and the persons identified as the document authors. All rights reserved.¶

This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document.¶

Table of Contents

1. Introduction

The current RFC publication process is unwieldy and slow. This is a poor match for an environment where protocol specifications routinely see wide deployment well in advance of IESG approval, let alone RFC publication. Despite the long publication time, RFCs also routinely contain errors (as an example, TLS 1.3 [RFC8446] currently has >40 errata). However, fixing these errors is prohibitively difficult as it currently requires publishing a new RFC, which incurs new delay, at which point new errors will have accumulated and the cycle begins again.¶

This document proposes a new approach to RFC publication, termed "just-in-time" (JIT) publication, which is designed to address this issue. JIT publication is centered around two basic ideas:¶

• A series of documents which are intended to be semantically identical, even though the text may be different.¶

• The ability to rapidly publish new documents within a series as soon as the requisite level of approval has been obtained.¶

When put together, these allow us to cheaply address issues as they are discovered. This document focuses on the IETF Stream as that is the dominant source of RFCs. However, the contents potentially apply to other streams as well.¶

2. Conventions and Definitions

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶

3. Requirements

The method in this document is intended to balance a number of requirements.¶

• Easy updating: Once a document is published it MUST be easy to make non-semantic changes (editorial fixes, clarifications, etc.) without re-running the whole process.¶

• Referential Integrity: It MUST be possible to reference a specific version of a document (i.e., the exact bits).¶

• Semantic References: It MUST be possible to reference a document as a semantic entity, including whatever editorial updates have been made.¶

4. Generic Structure

This section describes a generic publication structure that does not take into account the RFC Series. In Section 5 we describe how it might be mapped onto the existing RFC Series. It makes use of a two-level structure of Document and Version.¶

• Documents refer to the semantics of a given specification and thus may appear in multiple Versions. Each document has a document identifier D which is persistent through the life of the document and refers to the latest Version.¶

• Versions refer to specific instances of a Document. All versions of a given document have the same semantics. Each version has a unique version identifier under the document, as in D.V, indicating version V of document D.¶

In this design, documents would proceed through the process as usual until they were first published as an RFC. At the point where they were published by the RPC they would receive a document identifier and the initial version would be published, e.g., at https://rfc-editor.org/documents/D.0.¶

Once the initial version was published, new versions could be immediately published based on approval from the Area Director (or perhaps the WG Chair), who would be responsible for confirming that the changes did not change the document semantics.¶

This process allows for errata, etc. to be immediately applied to the document in place. Note that this is consistent with AUTH48 process because the AD can sign off on changes after IESG approval (see Section 6 for more on this). Moreover, the stakes here are quite low because we can always publish a new Version that reverts any change which has been inappropriately applied.¶

5. Mapping onto the RFC Series

There are at least two ways to map this onto the existing RFC Series, which only has one level with each RFC being unchanging.¶

• Add a new level of Document identifiers with RFCs as versions (e.g., D1234.1 -> RFC 123456)¶

• Add a new level of Version below RFC (e.g., RFC8446.0, RFC8446.1, etc.), with the main RFC# pointing to the most current version.¶

Either of these would probably work, and which people prefer to some extent depends on their priors. However, it's worth noting that the former approach would create a new reference that people would generally be pointed to that wasn't RFCs, whereas the second would not. In addition, it would start to burn through the RFC numbers very quickly, especially if each erratum creates a new version.¶

6. IETF Consensus

All IETF Stream documents require IETF rough consensus [RFC8789]. As with the existing publication process, the combination of IETF Last Call and IESG review is intended to ensure that the initial published document as reviewed by the IESG has IETF Consensus. The JIT publication process in principle allows for subsequent revisions to incorporate material that does not have consensus. This would be a process violation because ADs are supposed to only approve non-semantic changes, but of course ADs might make mistakes.¶

However, it is already possible to introduce non-consensus changes in RFCs during AUTH48. Authors routinely make changes to RFCs during the AUTH48 process; the RPC does not prevent these changes but just requires that substantive changes be approved by the appropriate AD. This proposal extends the period when changes can be made past initial RFC publication but the risk of non-consensus changes being mistakenly made is mitigated by several factors:¶

1. Changes are explicitly restricted to those without any semantic content, whereas AUTH48 changes can be semantically meaningful at the ADs discretion. The RPC could detect at least some of these cases, as they do now.¶

2. Any such changes are clearly visible as a diff from the previous version, so this situation is readily detectable. One could imagine having a short "objection period" between approval and publication to allow for incorrect changes to be detected.¶

3. It is trivial to publish a new version reverting any non-consensus changes.¶

Together, these changes minimize the risk of semantic changes being introduced to published RFCs.¶

As a side effect, JIT publication may also make AUTH48 faster, as authors would not need to worry so much about having every single sentence be perfect.¶

7. Version Publication Logistics

Once the RPC has published the initial version of an RFC, it should be trivial to make a change and re-spin the document without significant work by the RPC. Ideally, there would be a simple process in which the changes were proposed, approved by the ADs, vetted by the RPC and then mechanically published as a new RFC version. This should only require changing the specific text, rather than metadata, etc., which should all change automatically.¶

For readers familiar with GitHub, one could imagine this a workflow powered by GitHub actions:¶

1. Once the RFC is published, the XML is published on GitHub (e.g., in a repo containing the XML for every RFC).¶

2. The proposed change is submitted as a pull request to the XML.¶

3. A GitHub action automatically generates the resulting publication formats (HTML, TXT, PDF) for review.¶

4. The AD approves the pull request.¶

5. Upon AD approval and verification that no inappropriate changes have been made, the RPC merges the pull request.¶

6. A GitHub action publishes the RFC Version.¶

Obviously, these processes need not be based on GitHub, but this example illustrates the desired level of automation.¶

8. Published Version Adjustments

9. Security Considerations

This document in theory might make it possible to make semantically relevant changes to RFCs post-publication, which could have security implications. In the event that this happens, it is straightforward to revert these changes.¶

10. IANA Considerations

This document has no IANA actions.¶

Acknowledgments

TODO acknowledge.¶

Author's Address