# Submission review and change log

## Preservation

The eight original deliverable files were copied byte-for-byte to `output/original-submission/` **before revision**. `original-submission/snapshot-manifest.json` records their SHA-256 hashes. The review script verified all eight archived files against those hashes. Raw supplied inputs remain unchanged. The revised runnable site is still `output/site/`.

## Findings and changes

1. **Corrected a no-JavaScript fallback usability issue.** The original submitted form had an enabled button, field names and no explicit action/method. Source review identified that ordinary native submission could navigate using GET, put entered data in a URL and not save the signup. This was a source-inspection finding, not a reproduced JavaScript-disabled browser observation. In revised `site/index.html`, the submit button starts disabled; revised `site/app.js` enables it only after calculator initialization and submit-handler registration. The existing no-JavaScript notice remains. No fallback success is implied. The obsolete JavaScript comment promising native fallback validation was corrected.
2. **No pricing or endpoint correction needed.** The delivered calculation already uses the remaining-price factor (90% for 3+ seats), then adds 12% tax. Independent Decimal arithmetic confirmed all eight allowed quantities; new browser checks confirmed the four brief examples again. `/api/leads`, JSON fields and server behavior agree with the supplied contract. Sources: `inputs/brief.md`, **Workshop site acceptance brief**; `site/server.py`, **Handler.do_POST**. The broken `/api/lead` and discount multiplier in the supplied `site/app.js` are implementation defects, not competing requirements. No unresolved source conflict was found.
3. **Expanded evidence, without recasting old checks as new ones.** Added 11 revised-site browser checks, 13 in-process handler cases, a reproducible standard-library review script, independent arithmetic, archive-integrity checks, and local-asset/content checks. Preserved the initial evidence unchanged. No shell-launched server, screenshot, network-failure, or accessibility test is claimed.
4. **Clean delivery.** Archived the two new browser-created records to `evidence/review-browser-submissions.jsonl`, then removed live test storage. Temporary handler-test storage was also removed. No dependencies, framework, tracking or external service were added.

## Revised-site browser evidence

Actual calls used `site_path: output/site`, fresh browsers and the tool's trusted original server behavior. All **11** returned `overflow: false` and `errors: []`. These are text/control observations, not screenshots. IDs identify cases; batched calls can complete in a different order.

| ID | Width | Actions | Observed result |
|---|---:|---|---|
| R01 | 390 | Review Mobile / review-mobile@example.test / default 1 seat; submit | CAD 268.80; “Thanks! Your workshop request is saved.” |
| R02 | 1280 | Review Desktop / review-desktop@example.test / 2 seats; submit | CAD 537.60; same saved confirmation |
| R03 | 390 | Set seats=3 | CAD 725.76 |
| R04 | 1280 | Set seats=8 | CAD 1935.36 |
| R05 | 390 | Valid test name/email; blank seats; submit | Seat error; CAD — |
| R06 | 1280 | Valid test name/email; seats=0; submit | Seat error; CAD — |
| R07 | 390 | Valid test name/email; seats=-1; submit | Seat error; CAD — |
| R08 | 1280 | Valid test name/email; seats=1.5; submit | Seat error; CAD — |
| R09 | 390 | Valid test name/email; seats=9; submit | Seat error; CAD — |
| R10 | 390 | Three-space name / review-invalid@example.test / 1 seat; submit | “Enter your name.”; CAD 268.80 |
| R11 | 1280 | Review Invalid / review-invalid@example / 1 seat; submit | “Enter a valid email address.”; CAD 268.80 |

R05–R09 used `Review Invalid` and `review-invalid@example.test`. Seat error text was “Enter a whole number of seats from 1 to 8.”, present by the field and in the status. Seat constraint validity was false for each.

After R01–R04, a direct read of live storage found exactly the R01/R02 records. After R05–R11, the review script found exactly those same two records, confirming **no records added by the seven invalid submissions**. Successful clicks in R01/R02 also establish that JavaScript re-enabled the initially disabled button. The disabled-JavaScript case itself was not browser-tested: this tool has no JavaScript-disable action.

## Independent checks actually executed

`evidence/review_checks.py` ran successfully with Python's standard library. Output: `evidence/review-checks.json`.

- Archive hashes: all eight original deliverables matched.
- Server: byte-identical to supplied `site/server.py`; Python compile check passed.
- HTML: every original nonempty text node retained; IDs unique; both local asset references exist; static submit button has `disabled`.
- Decimal calculations: seats 1–8 give CAD **268.80, 537.60, 725.76, 967.68, 1209.60, 1451.52, 1693.44, 1935.36**. The four supplied examples match exactly (`inputs/brief.md`, **Workshop site acceptance brief**). This is an independent arithmetic check, not execution of the JavaScript calculator; the browser checks separately execute the calculator.
- **13 in-process tests of the actual `Handler.do_POST`:** valid seats 1 and 8 returned response code 201 and each added one record. Empty/whitespace name, malformed email, blank/zero/negative/fractional/9/boolean/string seats and malformed JSON returned code 400 and left storage unchanged. These tests used in-memory request/response streams, temporary storage and captured response-method calls. They did **not** launch a server, open sockets, or send HTTP traffic.

## Requirement and evidence audit

- Runnable files and instructions: server, HTML, CSS and JS all present; local links resolve by file inspection. Server startup was not independently launched.
- Calculator and invalid quantities: satisfied by initial and revised browser results plus arithmetic checks.
- Valid storage and invalid rejection: browser records and in-process handler checks support both layers.
- Mobile/desktop: no horizontal overflow observed at both required widths, including error and success states. Labels/controls remain present; readable sizing and minimum control heights are configured in CSS. Pixel appearance and real-device touch use were not observed.
- Identity/content: input title/copy retained by automated text-node comparison; original color values retained by stylesheet inspection. This is not a screenshot comparison.
- Initial claims: the recorded 14 browser cases, three records, reload caveat and file assertions agree with available initial evidence. The initial handoff did not claim server startup, screenshots or forced network failures, and the revised handoff retains those limits. The new 11 cases are separately identified here.

## Remaining limitations and estimated follow-up

No JavaScript-disabled browser test, standalone server-launch test, HTTP-level malformed-request test, forced network/server failure, rapid double-click, keyboard-only, screen-reader or screenshot check was performed. Browser-engine, zoom and other-width coverage remain untested. The original local server is not production-hardened. No known acceptance failure remains in the performed checks.

**Estimated human follow-up: 10–15 minutes**, not a measured result: launch the delivered server, inspect both widths visually, check keyboard focus and the disabled-JavaScript notice/button, and exercise a disconnected-server response.
