All,

Three answers and one request, so the draft has no surprises in it.

Nicholas, your line moves into v0.1. I said it would go in a proposed
section until someone on the list supported it; Kenne has, and he is right
that it belongs next to the roll-up rather than in v0.2. So section 4 will
carry it as agreed text: a roll-up states whether the checks it aggregates
were capable of a negative verdict, either because the run already shows a
non-pass, or by naming what the producer has instead - a control built to
fail that did fail, a prior run of the same declared checks on a corpus
where they did fail, or nothing. Nothing is a permitted answer. Silence in
that field is not. The change log will show it as a late addition with your
withdrawal of 44 signed results as the reason, and with the limit I raised:
a prior discriminating run only counts where check identity survives across
runs.

On the Merkle point, I am taking it into section 3 rather than leaving it
to a reader. The reading table for carry-or-reference gets one more line: a
digest match establishes a set only when the count of leaves is bound too,
and a report says which tree shape it uses. Your forgery passed every
per-leaf digest check, every inclusion proof and the root comparison, and
only the leaf count caught it, which is the whole argument for binding the
count. RFC 6962 domain separation is the other way to get there and I will
name both rather than pick one for producers who already have a tree.

Sankalp, yes, please co-edit the conformance appendix. Two things from your
message go into the known-gaps section as they stand: void has no slot in
your emitter today, and your not-exercised column carries no cause, so row
1 rejects your own output. That is exactly the kind of thing v0.1 should
say out loud about a reference implementation rather than quietly fix. I
will reference agent-evidence-vectors and the 461 vectors the eight
manifests pin, and leave your split of projection into compared and moved
to the row that reads moved, as you propose.

Nicolás, when you have a moment: the consolidated table with each row's
provenance, the record and slot definitions as they stand, and the 42
evidence objects with the three things the shape could not express. The
three become the known-gaps section beside the two above.

Dates are unchanged. Full draft on this list by Wednesday 23 September,
inline so the archive holds it whole, comments until Wednesday 30
September, then v0.1 with a change log that cites each message by what it
said.

Evgenii Arsentev, PhD

Chief Executive Officer, digital health

ORCID 0000-0002-9120-7298 https://arsentev.ai
