decisions: 100, row 76 round 2 also carries Filbert's round 1 notes (#1570)

Co-Authored-By: Claude Opus 5.5 <[email protected]>
2026-10-11 20:48:10 -05:00
parent 76631edd7d
commit fff031c4a8
1 changed files with 23 additions and 0 deletions
@@ -2164,3 +2164,26 @@ which stay with him. Each item names who decided it and what happened.
    - The tradeoff: round 2 takes longer to review. In exchange the proxy
      contract line is in the README before any send box is served, and
      no follow-up row is needed for five small notes.
+
+100. **Row 76's round 2 also carries Filbert's round 1 notes (2026-10-12).**
+    Filbert approved round 1 (#1570, comment 27637) with notes N1 to N4.
+    Decision 99 already covers Darkwing's notes, and two of Filbert's
+    overlap them: N2 is Darkwing's N1 (the Reply text test) and N4 is
+    Darkwing's N5 (the receipt repeats itself).
+    - N1 goes in: a 408 counts as unknown, not "not sent", in `bus.js`
+      and in the app.js Reply check, with a test for each and one README
+      line. Envoy's `stream_idle_timeout` can answer 408 after it has
+      forwarded the request, so a stored message could read as not sent.
+      Calling it unknown is the safe side: the page offers the trail
+      instead of inviting a resend.
+    - N3 goes in as tests only: field focus and caret after a refresh that
+      leaves focus in place (this also covers Darkwing's N2), a draft
+      whose role is no longer held, Inbox current on a message trail,
+      `in_reply_to: ""` reaching `send`, a coded plain error giving 503
+      unknown, and a 65-character `to` in the log assertion. None of these
+      changes behaviour.
+    - The decision 99 limit still applies. A note that turns out to need
+      more than a small change leaves round 2 and becomes a follow-up row.
+    - The tradeoff: round 2 grows by one behaviour change and about seven
+      tests. In exchange the 408 case is closed before any send box is
+      served, and the new page behaviour has tests before it lands.