A frame that refuses before it believes
Reading the Draft · No. 28 · Sources cited in this piece · the text read sentence by sentence: draft-li-rttp-iqa-addressing-02 (full text) · series page and revision history: datatracker · the family as it entered the mainline: the pulse-header family page (sealed record) · the merge record: lab PR #222 · Sister pillars: RTTP (rttp://, CRI 27) · IQA (iqa://, CRI 28)
A protocol frame has exactly one chance to be introduced, and most introductions waste it on praise. Ours arrives with a stranger habit: the pulse header of the pulse-header family introduces itself by listing what it refuses. The family — our case set for the fixed 128-octet frame of section 7.1, the extension-block rules of 7.1.1, and the time fields of 7.1.2 — entered the conformance lab's mainline as PR #222 this week, merged at 04:51:07Z on 2026-11-17, and this piece walks the discipline it carries: the frame's first act toward any input is rejection, and only bytes that survive the refusal list earn the right to be reasoned about.
The frame is fixed, so the reading is fixed
The frame's design starts from a refusal to parse. Section 7.1, verbatim: “The pulse header is the fixed 128-octet structure that carries an rttp pulse. Its fields are fixed-offset, so a reader validates a header with fixed comparisons rather than by parsing a variable layout.” Nine fields at nine fixed offsets — magic, version, sequence, timestamp, TTL, priority, the derived address, the subject's AID, and the extension block. Fixed comparisons are cheap to implement and expensive to get subtly wrong, which is exactly the trade a conformance suite wants: the family's frame cases are concrete 128-octet headers, and the checker re-derives each decision from the bytes with the same fixed comparisons the text demands, not from any library.
The extension block is the design's quietest virtue. The 26 octets at 0x66–0x7F were reserved and all-zero before the extension existed; the block occupies them “without moving or redefining any earlier field, and VERSION_ID is not incremented”, because “implementations that read only offsets 0x00-0x65 continue to validate unchanged, because the bytes they have not read were previously zero and remain ignorable.” Two of the nine negative cases pin this boundary from both sides: SPEC_REV 0 with a non-zero block is rejected, and SPEC_REV 1 with non-zero RESERVED is rejected. The extension earns its presence only inside a frame that stays provably identical to the pre-extension shape everywhere else.
The refusal list is the rule, not a preface
The fail-closed paragraph of section 7.1.1 is the family's center of gravity, and it is written as a list precisely so no case of judgement can creep in early: “A receiver MUST reject a frame if any of the following holds: the ACTION field contains characters outside the set of Section 4.1, exceeds 16 significant bytes, or carries non-zero padding; SPEC_REV is outside the known set; SPEC_REV is 0 and the extension block is not all zero; or SPEC_REV is 1 and RESERVED is not all zero or unknown FLAGS bits are set.” Nothing in that list asks the receiver what the frame means. The rejection happens on structure — character sets, lengths, paddings, reserved bytes — before any interpretation of the verb inside is entitled to happen at all.
Then the design does something stricter: it refuses even the inference that a frame which survived the list is safe. When an action is present, section 4.3 warns “a client MUST NOT infer safety from this section”, and the unknown-verb rule of 7.1.1 turns that warning into wire behavior: “On encountering an unrecognised action verb, an implementation MUST NOT infer safety from the default-operation statement of Section 4.3: the verb MUST be treated as having unknown safety and either require explicit authorisation under the local security policy or be rejected (Section 9).” The rule exists because the action space is open — “this document deliberately defines no enumeration of verbs, because enumerating an open set would close it” — and an open set guarantees that every receiver will eventually meet a verb it has never seen. Four behavior cases (PH-B03 to PH-B05 and PH-B12) pin the rule in both directions: silent admission is scored as nonconformant, because neither accepting by default nor the default operation is a verdict the receiver was entitled to give.
Time fields that claim almost nothing
The frame carries three time-and-order fields, and section 7.1.2 gives them deliberately narrow jobs: “a pulse is a claim of intent, and the only freshness that can authorize anything rides the answer side (Section 6); the pulse's time fields make the claim's moment checkable, not the subject's validity.” TIMESTAMP makes the pulse's age derivable offline, from the bytes alone — “a pulse states how old it is, whatever channel carried it” — and the very next sentence forbids the reader from promoting that age into trust: “A receiver MUST NOT treat TIMESTAMP, or the age it yields, as a statement about the subject's standing.” The TTL rules are equally unromantic: a receiver MUST NOT forward a pulse received with TTL 0 and MUST decrement before forwarding, while verification deployments that consume pulses where they receive them place no trust in the field at all. Four behavior cases pin the TTL boundaries and two more pin the timestamp boundary, and the mutation battery includes a mutant that drops the timestamp boundary entirely — it dies, because a case depends on the rule.
One line: a frame earns interpretation only by surviving its refusal list; an unknown verb earns nothing, least of all the default operation; and a pulse's timestamp says how old a claim is, never whether it stands — because fail-closed is not a feature of this design, it is the load-bearing wall: rttp resolves where & when; iqa states whether — and where the two agree, action begins.