Nicolás, all,

Received, and it is enough to write from. I am keeping your order:
definitions first, table second, so the table reads the record instead of
restating it.

What goes into the draft as agreed text: the per-check record with its four
fields and five states, cause required on the three non-verdict states and
absent on the two verdict ones; the cause vocabulary with the three
changes, including a value for void and integrity-failure kept apart from
availability-failure; the evidence object with compared and moved split,
moved recomputed rather than declared and arity recomputed from the delta;
the domain declared once at run level; carry-or-reference with the reading
table; the per-slot sort with the recompute-opaque kind that no row reads;
and the two row classes.

What goes in marked as proposed rather than agreed: the numbering of rows
13 and 14. Nobody has numbered them on the list, so the draft will show
them as proposed and say so in the change log.

What goes into known gaps, named as unfinished rather than quietly decided:
the classification of rows 4, 10, 11, 12 and 13. I am not classifying them
for you, and the text will say five rows are unclassified.

On the counts, I will carry your asymmetry exactly as you state it. The 42
objects are reproducible at commit 1e36257 through tools/emitir-42.py, and
the 0 and the 132 come from a script that is not published, so the text
will attribute those two to you rather than present them as reproducible.
Sankalp's 252 at v0.11.1 sits beside them with the same labelling. The
finding is that one field is a property of how a corpus was built.

3.1 and 3.3 publish as a single open item, who declares what is counted,
with the output side and the input side written out. 3.2 publishes as a
decision with its cost stated: the observation primitive is carried and
recomputable, and no row reads it.

On the Merkle line I will use your narrowing rather than my earlier
sentence. Binding the count is necessary and is not a general membership
proof, and domain separation added while duplicate-last padding is retained
does not remove the ambiguity, because RFC 6962 also uses a different tree
shape. The text will present neither as the fix on its own, and will
require the producer to declare the tree shape.

Dates unchanged. Full draft on this list by Wednesday 23 September,
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



On Fri, Sep 18, 2026 at 5:55 PM Nicolas Rocchia <nicolasrocchia@gmail.com>
wrote:
