Summary
- draft-ietf-6man-eh-limits-19 printed a tight family of IPv6 extension-header limits — 8 non-padding options, 64-byte option headers, 60-byte single options, a 104-byte header chain — and its Appendix A.2 derives them from a 128-byte minimum parsing buffer, which the printed numbers satisfy exactly.
- The one reviewed surface that names the years of the studies behind those figures, and the ossification risk of publishing them, is Éric Vyncke's Discuss ballot of 11 April 2025; the draft then went back to 6MAN at the author's request, lost its IETF stream and expired as an individual draft.
One buffer, five numbers
The draft's operative sentence about routers is blunt and, read alone, entirely arithmetic: "If a router needs to parse the upper layer protocol (as discussed in Section 7 of [RFC9098]) then it MUST be able to correctly forward packets that contain an IPv6 header chain of 104 or fewer bytes, or equivalently, a router MUST be able to process a packet with an aggregate length of extension headers less than or equal to 64 bytes" (the -19 draft text).
Around that MUST sit the recommendations the draft is known for. Sending hosts are told not to put more than 8 non-padding options in a Destination Options header, not to make that header larger than 64 bytes, and not to make any single option larger than 60 bytes, with the same three limits repeated for Hop-by-Hop Options; receiving hosts and intermediate nodes are given the mirror-image minimums and a recommended default of 8. Every one of those sentences carries its own escape clause — the explicit-knowledge exception — inside the same sentence, which is why quoting the number without the clause misstates the document.
The derivation is in Appendix A.2, which takes an expected minimum parsing buffer of 128 bytes and splits it: 16 bytes of Layer 2 header, 40 bytes of IPv6 header, 64 bytes of extension headers, 8 bytes of transport-layer header. It adds that typical parsing buffers come in 64, 128, 256 or 384 bytes. My own arithmetic, and nothing more: 16 + 40 + 64 + 8 = 128 closes exactly on the buffer the appendix names, and 40 + 64 = 104 restates the chain limit as the header plus the extension-header allowance the draft's own "equivalently" clause already asserts.
Neither sum tells us that any deployed router actually has a 128-byte buffer; the appendix offers a buffer size, not a measurement of hardware.
The sentence that carries the years
A number of that kind is only as good as the population it describes. The draft prints the population for each limit; what it does not print, anywhere in the reviewed text, is when the underlying behaviour was observed or why publishing the number might change it.
Éric Vyncke's ballot does. It is printed under the heading "Éric Vyncke, INT AD, comments for draft-ietf-6man-eh-limits-19", dated 11 April 2025, and archived in the ipv6 list (the archived ballot). He opens by crediting the authors' intent and then saying he is "deeply concerned about having a reverse effect and actually leading to an IPv6 ossification." Under a subheading, Race to the bottom, he spells out the mechanism: if the I-D is published, "it will most probably end up in call for tender or other BCP, i.e., vendors will limit themselves to the different limits written in this document and not try to go beyond, leading to an ossification of the Internet and preventing creative use of extension headers."
The provenance problem follows in the next paragraph, and it is the sentence this article is built around: "The limits are based on studies done in 2022 and 2023 over the public Internet." From that he draws two objections — that the document should perhaps restrict its applicability to the public Internet, and that anchoring the future of the protocol to studies of 2022 and 2023, "going over routers dating possibly back to 2017" and without knowing whether a limit is operator policy or a hardware ceiling, is too restrictive. Then, in five words: "64 bytes for the extension header chain is really limiting."
There is a second, quieter act in the same document. Section 5.3 of RFC 8504, a BCP, uses "MAY" about the same limits; he asks how an Informational I-D can convert a BCP "MAY" into a "SHOULD", proposes "a normal 'should' w/o BCP14 weight", and asks that "defined in this document" become "suggested" because the intended status is informational. He also records that he commissioned two directorate reviews himself, from Jean-Michel Combes and Jouni Korhonen, and thanks the working-group shepherd for a write-up he labels incomplete.
Whatever else the ballot is, it is not a drive-by: it is an editor at work on someone else's document, with the reviewers' names attached.
What the ballot page prints, and what it does not
The Datatracker ballot page for the same revision dates his entry 2025-04-11 and carries the document-level summary line "Needs a YES. Has 5 DISCUSSes." (ballot page). That count is the page's, not mine: the comment text of the other four Discuss positions was not read for this piece, so five is a printed summary, not a survey of objections.
What the page cannot tell a stranger is what happened next, because nothing there reconciles a five-Discuss ballot with the outcome. The status page does: the document is an "Expired Internet-Draft (individual) Expired & archived", the latest revision is -19 dated 2025-02-27, the expiry date shown is 31 August 2025, the IESG state is "Expired", the stream state "No stream defined", and no RFC number appears (status page).
The history table supplies the shape of the exit and, in its own wording, the cause: a 2025-11-02 entry reading "Document has expired", a return to 6MAN from the IESG because "the author requested it and the working group showed apparent lack of interest", release by the 6MAN chairs, conversion to an individual draft, the stream changing to None from IETF, consensus to No from Yes, and the IESG state to I-D Exists from IESG Evaluation::Revised I-D Needed (history). No revision after -19 appears.
That is a withdrawal, not a verdict. No reviewed document says the draft was rejected on its technical merits, and this piece does not say so either. By my own subtraction of two printed dates, the gap between 11 April 2025 and 2 November 2025 is just under seven months; the record does not say the ballot caused the expiry, and neither do I.
Two renderings, one word apart
The archive copy and the Datatracker rendering of the same 11 April 2025 ballot disagree about how many blocking points it contains: the mailing-list original reads "one blocking DISCUSS points (easy to address)", while the Datatracker rendering reads "several blocking DISCUSS points". Both are printed renderings of one ballot. The substantive DISCUSS text is the same in both. A one-word difference between two surfaces that describe the same act is worth naming precisely because it is small: it is the kind of drift that turns a single objection into a chorus when someone quotes the summary instead of the message.
Where the record stops
Three gaps bound everything above. The COMMENT and nit sections of the ballot were not read, only the DISCUSS section, so the non-blocking points — the Routing Header grammar, the overbroad IPsec AH sentence, the repeated RFC 8200 ordering advice — are known here only from their headings. The 2022 and 2023 studies the ballot names were not obtained, so the question of whether a measured limit reflects operator policy or hardware cannot be answered from the documents reviewed. And the 6MAN discussion itself was not read, so whether the working group ever answered his race-to-the-bottom argument is unknown to this piece.
The finding does not depend on those gaps. It depends on this: the draft's numbers can be audited from outside with a hundred and twenty-eight bytes of arithmetic, and the only reviewed sentence that says what population those numbers came from, and what publishing them might do to the protocol, is one Area Director's dated objection. The document he objected to is now an archive entry. The numbers are not.
Member Briefing
Read the full trend briefing
Members read the full briefing with the source notes behind every finding.
Strategic Circle
A daily briefing in your language, built from changes we have verified.
Join for USD 20/monthCancel anytime.
Members also get the full research library, a daily briefing on the networks they follow and the private forum.






