[SCITT] Re: Action requested: Focused Working Group Last Call on Section 4 of the CCF Profile

[Cliff Singletary <[e-mail]>]

Konrad, all,
I independently recomputed the fifteen Section 4 vectors (b7n0de/proofbundle@8994f52) from the -05 text alone: Python standard library plus cryptography, my own CBOR decoding, and none of the vectors' checker or generator code. The per-vector results table is available on request.
Figure 9 as written reproduces all 15 recorded outcomes. On the four rejects (S4-11, S4-13, S4-14, S4-15) it also fails at the same first statement. A strict reading built only from -05's MUSTs and CDDL agrees with the recorded reading B on all 15 outcomes, with and without tree sizes.
For Section 4:

     *
Figure 9 accepts S4-03, S4-05, S4-06 and S4-12, which 4.1, 5 and Figure 11 exclude. I support Figure 9 asserting that every proof in a receipt, consistency or inclusion, recomputes the same newer root. Figure 9 should also say whether verifiers reject or ignore verifiable-proofs keys other than -1 and -2.
     *
Two anchor conditions are checkable without m: the first sibling is a right sibling, and at least one sibling is a right sibling. Both hold for every canonical proof with 0 < m < n ≤ 128, and they reject S4-08 and S4-09. 4.2 could state them.
     *
S4-10 passes every size-free check and fails only once (m, n) = (6, 7) is known. That makes "neither tree size is needed for verification" look too strong. If the full anchor rule needs sizes, how is a verifier meant to obtain m and n bound to the roots?

Thank you,
Cliff Singletary