Integration direction#
Keep the historical draft and its receipt immutable. The canonical guide belongs
under Products → V2 competitive fighting → Netcode, after Orientation, with
v2-product-promise and v2-contracts as prerequisites. This introduces a
learning branch for local netcode mechanisms without claiming the entire
rollback/transport curriculum is present. Delivery and release remain last.
Visual thesis: retain the verified draft's cream-and-ink evidence comparisons, precise rule tables and selectable source annotation at actual embedded sizes. Content plan: preserve its twelve-slide arc and all local-versus-consumer qualifications; replace only the closing manuscript-publication note with a source-revision/environment reminder suitable for the permanent guide. Interaction thesis: use the existing hierarchy disclosure, selected-guide transition and slide/source controls, with reduced motion; verify expanded and collapsed navigation plus actual compact playback after audio generation.
The frontend skill directs this restrained integration. No renderer redesign, new imagery or decorative motion is needed. Integration verification, narration and final reader/listening acceptance are recorded separately in the integration receipt; a ready audio duration does not mean a reader study or complete release acceptance.
Historical draft direction#
Audience: a product or engineering reader who understands the V2 product promise
and evidence conventions, and wants to distinguish a network recommendation from
an enacted match policy. Pin: 1ceae52c7896a59ac74a4429d410f3218f0b91c6. This
is an unpublished manuscript, not an integrated, narrated guide or reviewed
teaching assignment. The architecture unit is 020598d821f509c36fd7427f, lines
262–309; its entire prose, threshold table, diagram and related-boundary claims
need to be addressed, including corrections against actual C++.
Visual thesis: use the cream-and-ink Oshun reading plane with restrained terracotta emphasis on the distinction between a calculated record and a live effect; exact comparisons and a short selectable source artifact carry evidence.
Content plan: establish the boundary, explain measurements and the pre-match grade, work an aggregation example, separate recommendations from the session, then examine pressure, branch precedence and acceptance. Compare config with policy validation, work the reliability score, inspect diagnostic flags, and close with a concrete investigation exercise.
Interaction thesis: preserve slide entrance, source-dialog disclosure and navigation transitions with reduced motion. All relationships must survive static PDF; no additional decorative motion or invented gameplay/dashboard screenshot is justified. The frontend skill directs this content-led choice of native evidence rows, exact tables and source annotation, not a marketing hero or repeated card grid.
Reading plan#
| Slide suffix | Job and relationship | Transition |
|---|---|---|
boundary |
Independent evidence questions, resolving on enacted behavior. | Ask what the analyzer receives. |
measurements |
Named measurements with units and the local aggregation rule. | Compare the resulting grade. |
handshake |
Ordered Poor/Borderline/Good threshold table with exact inequalities. | Calculate one concrete example. |
handshake-case |
Two samples and their aggregate; no performance ranking. | Ask what changes when a recommendation returns. |
session-boundary |
Independent analyzer, session and consumer responsibilities. | Examine the mid-match pressure input. |
pressure |
Trailing-entry comparisons, count versus elapsed time. | Follow branch precedence. |
fallback-order |
First-matching branch table, not a diagram implying live transitions. | Separate offered from accepted. |
acceptance |
Ranked offer/acceptance and stable-branch counterexample. | Locate the separate validators. |
policy |
Config validity versus default-policy alignment. | Keep scoring separate from mode and admission. |
score |
Explicit per-term penalty weights and a 75-point worked case. | Inspect the record that carries the score. |
diagnostics |
Exact source lines assign flags and return a value. | Ask what would establish persistence. |
investigation |
Ordered reader investigation from input through predicate to actual consumer. | Continue into session, transport and online-operation guides. |
Every slide has one primary claim, explanatory narration paragraphs, three reader questions and pinned sources. The native experiments are explicitly adapted-host observations, not an unchanged-pin build or a runtime network pass. The guide must not copy the source diagram's automatic handshake-to-session arrows, one-time-offer claim or offered-to-settled no-contest implication.
Verification and acceptance ledger#
Run the complete isolated draft Playwright flow at desktop, compact and mobile, including every source panel, table column, accessibility audit and PDF page. Extend exact-content checks and include deliberately corrupted conditions, source characters and hidden evidence. Inspect screenshots and actual PDF renders; test narrow/enlarged text without reducing the established type floors. Incomplete narration stays unavailable; no synthesized track or reader result is implied by a draft build. Recheck active playback after real narration exists.
The eight quality standards remain explicit: this is not an accepted benchmark; typography needs final font/zoom and reader evidence; quantitative examples are local calculations with stated units, not measured population statistics; record/consumer vocabulary must remain consistent; no image is justified here; actual-size browser/PDF review is separate from geometry; willing-reader comprehension and normal-speed listening remain pending; remove any repeated qualification that adds no new distinction while preserving safety boundaries.
Neutral reader tasks, before assessment or coaching:
- Two samples have RTT 80/120 ms, jitter 3/25 ms and loss 2/6 percent. What does the handshake return, and what is still unknown about a live match?
- A ranked fallback record has both offered and no-contest flags. What can you infer about the player's response and durable match outcome?
- A configuration passes its predicate. What else must be checked before using the default fallback feature policy?
- Extreme RTT alone yields a score of 75. Why is that compatible with a Poor handshake, and what proves an account or dashboard was actually updated?
Bind willing-reader answers to the final stimuli and record accommodations; never substitute automated checks or model review for participant evidence.
Assess only after retaining the reader's unaided answer. Expected boundaries: the worked example returns Borderline/Hybrid and ranked eligibility, with no proven live mode change; jointly offered/no-contest flags do not establish a decline or durable settlement; config validity is separate from aligned feature policy and returned-decision validity; and a saturated RTT term contributes only 25 penalty points, independently of the Poor handshake threshold. Account or dashboard effects require the actual authorized consumer and persisted/delivered result. Equivalent accurate explanations are acceptable. These are assessment criteria, not observed reader results or text to coach a first response.
Mobile review correction#
The first rendered numeric example split “Sample” and “Aggregate” inside words at 390px despite passing cell-bound checks. The final headers are “A”, “B” and “Used”; the caption names samples A/B and the value used for grading. Exact numbers, units, aggregation rules and narrative meaning are unchanged. A browser word-range check now covers numeric headers at 320px and with 20px body text, with a deliberately split header required to fail. Source-symbol text is enlarged to 24px in that stress case, not reduced below its existing 22px floor.
Eight-standard draft rubric#
Exact stimuli, captures, defect history and automated results are bound in the dated receipt. The reviewer is the authoring agent, not a recruited reader. These statuses do not promote any treatment to an accepted benchmark.
| Standard | Evidence in this draft | Still required |
|---|---|---|
| Benchmark | Existing Oshun content-led vocabulary applied to a complete twelve-slide argument. | Accepted pilot references and reader evidence; no benchmark acceptance claimed. |
| Typography | Complete desktop/compact/mobile/PDF inspection; numeric-heading repair; narrow and 20px body stress, with 24px source symbol. | Broader font substitution, zoom and reader-backed typography acceptance. |
| Quantities | Typed two-sample illustrative calculation and configured penalty weights, exact units/provenance; native, browser and PDF checks. | Narrated interpretation and reader transfer task; no production measurement claim. |
| Continuity | Supplied input → returned record → actual consumer vocabulary; the two-sample example follows threshold rules; branch alternatives stay explicit. | Integrated route and actual narration review. |
| Images | Not applicable to this local code/record explanation; selectable source and exact tables supply the evidence. | No fabricated gameplay or dashboard screenshot may stand in for a real consumer. |
| Actual sizes | All twelve standalone desktop/compact/mobile slides and PDF page-fit views; all eight table-end views; 320px/enlargement automation. | This draft's embedded and active-audio states after integration; reader accommodations. |
| Comprehension | Four neutral tasks above, written before reader assessment. | Willing participants, exact stimuli/setup, observations and defect/retest records; no simulated acceptance. |
| Subtraction | Short A/B/Used headings with a full caption; unnumbered peers; ordered rows only for branch precedence; detailed qualifiers in notes. | Normal-speed listening and final integrated editorial acceptance. |
Historical integration review and findings#
The canonical twelve-slide guide uses the same visual fields as the frozen draft. The only narrative edit replaces the closing publication-status sentence with a reminder to retain the source revision and environment. The guide adds 1,190.383 seconds of recorded narration. Neither source-unit assignments nor a new native execution result are inferred from publication.
The final focused matrix adds real playback at 1280×720 and all twelve slides in five actual embedded contexts: 1440×900 and 1280×720 with navigation expanded and collapsed, plus 390×844 mobile. The three smaller contexts also play each track and scroll to the seek control, verifying that all playback controls are visible and usable. These 36 additional checks test reachability, not whole-slide first-screen fit. A hidden embedded seek control must fail. The full matrix has 168 checks, ten rejected defects and 108 captures.
The measured embedded frame sizes below are CSS pixels, after the actual center header and navigation have taken their space. They are not fixed-canvas scale factors or assumed dimensions. Capture pixels equal the outer viewport pixels in this run; each final capture retains its own frame/viewport record.
| Workspace | Outer viewport | Actual guide frame |
|---|---|---|
| Desktop, sidebar expanded | 1440×900 | 1148×797 |
| Desktop, sidebar collapsed | 1440×900 | 1440×797 |
| Compact, sidebar expanded | 1280×720 | 988×617 |
| Compact, sidebar collapsed | 1280×720 | 1280×617 |
| Mobile, drawer closed | 390×844 | 390×741 |
The smaller-frame playing captures record 0–565 CSS pixels of vertical scroll, depending on the slide. This is why a standalone 1280×720 fit cannot be inherited as an embedded 988×617 fit. Browser zoom, fallback fonts and other accommodation states need their own explicit coverage; this matrix does not establish them.
Model inspection covers those 108 captures, all 44 standalone initial/table-end images, all twelve delivered PDF pages at 960px and 1440px width, and a complete guide contact sheet. Sixty final focused captures inherit inspection only by byte equality with the earlier reviewed run; the other 48 are inspected anew. Tall standalone mobile images are reading captures, not initial phone viewports. The contact sheet establishes sequence rhythm, not small-text legibility.
The eight draft rubric entries above remain historical. The integration-stage update below predates the shared table repair recorded at the end of this file:
| Standard | Current evidence and disposition |
|---|---|
| Benchmark | Existing evidence rows, exact tables and source annotation serve their named jobs. No accepted reference or reader benchmark is claimed. |
| Typography | Numeric headings remain intact under narrow/enlarged checks. Open defects: score row labels separate 200/ms and 40/ms on mobile; ordinary prose-table headers are noticeably small at PDF page fit. Resolve number/unit cohesion and effective header sizing with font/zoom and visual regression evidence. |
| Quantities | Exact configured/illustrative values, units, precedence and selectable source characters survive browser and PDF checks. Actual playback is verified; narrated interpretation and reader transfer still need listening and participant evidence. |
| Continuity | The complete route carries input → local record → actual consumer consistently. Similar adjacent tables compare exact rules; the responsibility slide, source artifact and final investigation change pace where the content changes. |
| Images | No bitmap is warranted for these local predicates and records. A hypothetical match screen or dashboard would falsely imply consumer evidence. |
| Actual sizes | Complete new-guide browser/PDF inspection and expanded/collapsed playing controls are recorded. Vertical/horizontal reading remains explicit; font substitution, full browser zoom and reader-supported effective-size thresholds remain open. |
| Comprehension | Four neutral tasks and source-backed assessment criteria are ready. No willing-reader observations, accommodation results or retests have been collected. |
| Subtraction | The publication-only sentence is removed from permanent narration; short numeric headings retain their full caption. Independent peers have no ordinals and actual branch/investigation order remains. Final listening and editorial acceptance are pending. |
Neither this integration review nor the shared repair establishes final craft acceptance. Required reader/listening evidence remains open; repeat affected checks and reviews after any artifact revision.
The original header finding had a concrete shared-CSS cause: ordinary thead th
used 11px, the compact active-audio rule repeated 11px, and print supplied no
larger ordinary-header override. Quantitative headers separately used 16px. With
the 13.333-inch print width, a 960px page-fit raster gives the ordinary header
an approximately 8.25px effective size (derived from CSS/page scale, not a
measured font-ink height or reader-approved floor). The mobile quantity row used
a narrow clamp(88px, 32cqw, 140px) sticky column with unrestricted text
wrapping. These findings required shared typography/quantity repairs with
representative regressions, preserving exact text and quantities.
Shared table repair and remaining acceptance#
The
shared repair receipt
records intact 200 ms, 40 ms and > 8 frames groups and authored-case 16px
table headers, including compact playback. All 40 exported PDFs retain complete
tables with 12px effective headers at 960px page fit, measured through Poppler's
integer font metadata and paired with bounded raster inspection. A separate
print-layout guard rejects clipping that survives text extraction.
The whole table inventory now has narrow, enlarged, missing/preferred-font and actual 200%/400% browser-zoom checks. The full 421-slide bank passes all twelve checks with observed exit zero, completing automated verification for the two concrete table findings. The font probe observes Liberation Sans for both the preferred and missing-font stacks on this host, so it does not prove a distinct fallback face. At 400% embedded zoom, navigation occupies substantial height and truncates the guide title; that accommodation issue remains open.
The shared repair leaves every configured/illustrative value, exact source excerpt and recording unchanged. Source-unit assignments, normal-speed listening, willing-reader tasks/accommodations/retests, accepted benchmarks, complete optical refinement and reader-backed effective-size thresholds remain required. The historical integration matrix above is not current-artifact evidence; use the repair receipt's exact hashes and declared scope.
Reviewed architecture-unit assignment#
The
teaching receipt
records review of the complete architecture unit at lines 262–309 of
V2/architecture/rollback-netcode-and-tag-team.md, under the pinned source
revision. The assignment binds all
twelve assembled slide fingerprints and their explanations to that unit.
The review includes every grade row and threshold, the eight-node fallback diagram, consecutive pressure, wireless wording, score expression, diagnostics fields and both follow-up links. It qualifies those source claims using the actual analyzer and separate predicates: returned targets are not demonstrated session transitions; sample count needs a sampling contract to imply seconds; false acceptance does not establish a decline; and flags or an endpoint name do not perform persistence or delivery. The guide's worked cases teach those distinctions with their qualified native-evidence limits.
This closes one documentation teaching unit. It does not assign the referenced code, generated duplicates or other sections of the architecture page. The guide's visual/audio artifacts are unchanged from the shared-table repair; normal-speed listening, participant comprehension and final craft acceptance remain pending.