devices(nvr): three recorders still drew from a 0 W budget, and a fourth cited a reseller #2297

Merged
jaylouisw merged 2 commits from catalogue/1120-1822 into main 2026-11-20 20:10:05 +00:00
Owner

Three Dahua/Hikvision recorders named their camera ports PoE n, quoted a wattage in description/notes, and carried no poe block at all - so engine/device.js fell through to parsePoeBudget(model), found no number in DS-7616NXI-K2/16P, and gave each box 0 W to supply. A fourth had the number but sourced it from a reseller's reprint. scripts/audit-poe-budget.js: main 83e2e7c4 49 silent / 411 PSEs, this branch 46.

Reconciled with #2295. This PR opened with seven rows. #2295 merged 18:38 and landed three of them (nvr5216-16p-ei2, ds-7608ni-k2-8p, ds-7616ni-k2-16p); those three files take main's landed wording and are not contested here. What remains is the part main still lacks.

spec budget vendor source
devices/dahua/nvr2108hs-8p-4ks3.json 72 W (25.5 W/port) https://www.dahuasecurity.com/products/network-products/network-recorders/lite-series/nvr2-4ks3-series/nvr2108hs-8p-4ks3 (200, 514,739 B)
devices/dahua/nvr4208-8p-4ks3.json 130 W (25.5 W/port) https://www.dahuasecurity.com/products/network-products/network-recorders/Lite-Series/NVR4-4KS3-Series/2HDD/NVR4208-8P-4KS3 (200, 524,439 B)
devices/dahua/nvr5216-16p-4ks2e.json 130 W rated / 80% control https://www.dahuasecurity.com/asset/upload/product/20180201/DHI-NVR5216_5232-16P-4KS2E_Datasheet_20180201.pdf (200, pdf, 452,417 B, 3 pp)
devices/hikvision/ds-7616nxi-k2-16p.json 200 W https://www.hikvision.com/content/dam/hikvision/products/S000000001/S000000132/S000000007/S000000026/OFR031212/M000141849/Data_Sheet/Datasheet-of-DS-7616NXI-K2_16PD_20240523.pdf (200, pdf, 987,036 B, 5 pp)

All four vendor documents were re-read this pass (Dahua HTML spec tables; both PDFs through their text layers), not inherited from the branch point.

Two of them were worse than a missing number. nvr4208-8p-4ks3 and ds-7616nxi-k2-16p had no poeOut at the spec level on any of their 8 / 16 camera ports; the engine's name rule inferred them, which is why the audit still counted 8 and 16 PSE - but an inferred port is not a declaration. Both now declare it, and all four gain poe-pse / 802.3af / 802.3at, which they were missing while their prose promised 802.3af/at.

The NVR5216-16P-4KS2E row also stops citing a reseller's re-hosted copy of the Dahua datasheet (rxtx.com.au/...) in both sourceUrl and poe.sourceUrl; it now points at the Dahua asset the sibling NVR5232 row already used, and the 130 W was re-read there rather than inherited.

Full evidence, including the port-ceiling-vs-budget arithmetic and how the Hikvision asset's JS cookie challenge was passed to reach the PDF, is in research/nvr-integrated-poe-census.md.

Gates: syntax-check.js 495 files OK; audit-port-types.js 0 contradictions; audit-poe-budget.js 46 silent / 0 inferred-from-name / 0 budget-without-PSE; regen-registry-doc.js --check no drift (2,355 models); scripts/run-tests.js 415 suites, 13,817 assertions, 0 failed.

Closes nothing: this is catalogue hygiene under the device stream. #803 (E6616X) remains blocked upstream and is not touched here.

Three Dahua/Hikvision recorders named their camera ports `PoE n`, quoted a wattage in `description`/`notes`, and carried no `poe` block at all - so `engine/device.js` fell through to `parsePoeBudget(model)`, found no number in `DS-7616NXI-K2/16P`, and gave each box **0 W** to supply. A fourth had the number but sourced it from a reseller's reprint. `scripts/audit-poe-budget.js`: main `83e2e7c4` **49 silent / 411 PSEs**, this branch **46**. **Reconciled with #2295.** This PR opened with seven rows. #2295 merged 18:38 and landed three of them (`nvr5216-16p-ei2`, `ds-7608ni-k2-8p`, `ds-7616ni-k2-16p`); those three files take main's landed wording and are not contested here. What remains is the part main still lacks. | spec | budget | vendor source | |--|--|--| | `devices/dahua/nvr2108hs-8p-4ks3.json` | 72 W (25.5 W/port) | <https://www.dahuasecurity.com/products/network-products/network-recorders/lite-series/nvr2-4ks3-series/nvr2108hs-8p-4ks3> (200, 514,739 B) | | `devices/dahua/nvr4208-8p-4ks3.json` | 130 W (25.5 W/port) | <https://www.dahuasecurity.com/products/network-products/network-recorders/Lite-Series/NVR4-4KS3-Series/2HDD/NVR4208-8P-4KS3> (200, 524,439 B) | | `devices/dahua/nvr5216-16p-4ks2e.json` | 130 W rated / 80% control | <https://www.dahuasecurity.com/asset/upload/product/20180201/DHI-NVR5216_5232-16P-4KS2E_Datasheet_20180201.pdf> (200, pdf, 452,417 B, 3 pp) | | `devices/hikvision/ds-7616nxi-k2-16p.json` | 200 W | <https://www.hikvision.com/content/dam/hikvision/products/S000000001/S000000132/S000000007/S000000026/OFR031212/M000141849/Data_Sheet/Datasheet-of-DS-7616NXI-K2_16PD_20240523.pdf> (200, pdf, 987,036 B, 5 pp) | All four vendor documents were re-read this pass (Dahua HTML spec tables; both PDFs through their text layers), not inherited from the branch point. Two of them were worse than a missing number. `nvr4208-8p-4ks3` and `ds-7616nxi-k2-16p` had **no `poeOut` at the spec level** on any of their 8 / 16 camera ports; the engine's name rule inferred them, which is why the audit still counted 8 and 16 PSE - but an inferred port is not a declaration. Both now declare it, and all four gain `poe-pse` / `802.3af` / `802.3at`, which they were missing while their prose promised 802.3af/at. The NVR5216-16P-4KS2E row also stops citing a **reseller's re-hosted copy** of the Dahua datasheet (`rxtx.com.au/...`) in both `sourceUrl` and `poe.sourceUrl`; it now points at the Dahua asset the sibling NVR5232 row already used, and the 130 W was re-read there rather than inherited. Full evidence, including the port-ceiling-vs-budget arithmetic and how the Hikvision asset's JS cookie challenge was passed to reach the PDF, is in `research/nvr-integrated-poe-census.md`. Gates: `syntax-check.js` 495 files OK; `audit-port-types.js` 0 contradictions; `audit-poe-budget.js` 46 silent / 0 inferred-from-name / 0 budget-without-PSE; `regen-registry-doc.js --check` no drift (2,355 models); `scripts/run-tests.js` 415 suites, 13,817 assertions, 0 failed. Closes nothing: this is catalogue hygiene under the device stream. #803 (E6616X) remains blocked upstream and is not touched here.
devices(nvr): six recorders sourced eight to sixteen PoE ports from a 0 W budget
All checks were successful
CI / Classify the change (pull_request) Successful in 26s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 22s
CI / Test (Node 22) (pull_request) Successful in 1m5s
CI / Test (Node 20) (pull_request) Successful in 1m8s
CI / Test (Node 18) (pull_request) Successful in 1m11s
CI / Server smoke and Docker build (pull_request) Successful in 23s
CI / Arm auto-merge (pull_request) Successful in 4s
2cb73c6080
Dahua NVR2108HS-8P-4KS3 / NVR4208-8P-4KS3 / NVR5216-16P-EI2 and Hikvision
DS-7608NI-K2/8P / DS-7616NI-K2/16P / DS-7616NXI-K2/16P name their camera ports
'PoE n' and quoted a wattage in prose, but carried no poe block, so
engine/device.js fell through to parsePoeBudget(model) and gave each box 0 W to
supply. audit-poe-budget: 61 silent PSEs / 6 prose -> 55 / 3.

Budgets and per-port claims from the vendors' own documents: 72 W and 130 W
(Dahua product spec tables, per port 25.5 W), 130 W (DHI-NVR5216/32-16P-4KS2E
datasheet, which also replaces the reseller copy that row cited), 80 W and
200 W (Hikvision datasheets). NVR4208 and DS-7616NXI also gain the poeOut they
had only by inference, and all six gain poe-pse/802.3af/802.3at.
forgejo-actions scheduled this pull request to auto merge when all checks succeed 2026-11-20 18:34:30 +00:00
Author
Owner

needs Jay's decision — this PR now collides with main on three files, and the collision is data, not formatting.

What landed first: #2295 (2ff67857, "devices(poe): six budgets that were only in prose…"). It was opened at 18:12, this PR at 18:22, and it carries test/poe-budget-prose-six.test.js, which asserts poe.budgetW for all six models. Both PRs were green 7/7 and individually mergeable into main; they are not mergeable with each other.

Verified locally (git merge --no-commit origin/catalogue/1120-1822 onto 2ff67857):

  • conflicts, all CONFLICT (content) on the same object key:
    • devices/dahua/nvr5216-16p-ei2.json
    • devices/hikvision/ds-7608ni-k2-8p.json
    • devices/hikvision/ds-7616ni-k2-16p.json
  • everything else in this PR auto-merges cleanly: the capabilities additions (poe-pse, 802.3af, 802.3at), the notes rewrites, research/nvr-integrated-poe-census.md, and the other five device files.

Why I did not resolve it: the wattages agree exactly (130 / 80 / 200 W), but both sides supply a hand-written poe.note for the same key, and the two notes make different factual claims:

model #2295 note says this PR's note says
DS-7608NI-K2/8P "Eight ports at 25.5 W would want 204 W" per-port ceiling is "the 802.3at 30 W the standard names, not a vendor figure, so the row does not claim one"
DS-7616NI-K2/16P "Sixteen ports at 25.5 W would want 408 W" "16 × 30 W of 802.3at ceiling = 480 W against a 200 W supply"
NVR5216-16P-EI2 same 25.5 W per-port reading same budget, different quote and read-metadata

Picking one note over the other is a sourcing judgement, not a merge — and sourceUrl also differs for the 7616 (vendor product page in main vs. the 2022-10-26 datasheet PDF here).

Decision needed: which per-port ceiling the catalogue quotes for these integrated-PoE NVRs — 2295's 25.5 W vendor figure or this PR's 30 W 802.3at ceiling — and which sourceUrl/note survives for the three overlapping models. Once that is settled the rebase is trivial: keep one poe block per file, keep this PR's capabilities and notes changes. Say the word and I will do exactly that mechanically.

**needs Jay's decision** — this PR now collides with main on three files, and the collision is data, not formatting. **What landed first:** #2295 (`2ff67857`, "devices(poe): six budgets that were only in prose…"). It was opened at 18:12, this PR at 18:22, and it carries `test/poe-budget-prose-six.test.js`, which asserts `poe.budgetW` for all six models. Both PRs were green 7/7 and individually mergeable into main; they are not mergeable *with each other*. **Verified locally** (`git merge --no-commit origin/catalogue/1120-1822` onto `2ff67857`): - conflicts, all `CONFLICT (content)` on the same object key: - `devices/dahua/nvr5216-16p-ei2.json` - `devices/hikvision/ds-7608ni-k2-8p.json` - `devices/hikvision/ds-7616ni-k2-16p.json` - everything else in this PR auto-merges cleanly: the `capabilities` additions (`poe-pse`, `802.3af`, `802.3at`), the `notes` rewrites, `research/nvr-integrated-poe-census.md`, and the other five device files. **Why I did not resolve it:** the wattages agree exactly (130 / 80 / 200 W), but both sides supply a hand-written `poe.note` for the same key, and the two notes make different factual claims: | model | #2295 note says | this PR's note says | |---|---|---| | DS-7608NI-K2/8P | "Eight ports at **25.5 W** would want 204 W" | per-port ceiling is "the 802.3at **30 W** the standard names, not a vendor figure, so the row does not claim one" | | DS-7616NI-K2/16P | "Sixteen ports at **25.5 W** would want 408 W" | "16 × **30 W** of 802.3at ceiling = 480 W against a 200 W supply" | | NVR5216-16P-EI2 | same 25.5 W per-port reading | same budget, different quote and read-metadata | Picking one note over the other is a sourcing judgement, not a merge — and `sourceUrl` also differs for the 7616 (vendor product page in main vs. the 2022-10-26 datasheet PDF here). **Decision needed:** which per-port ceiling the catalogue quotes for these integrated-PoE NVRs — 2295's 25.5 W vendor figure or this PR's 30 W 802.3at ceiling — and which `sourceUrl`/`note` survives for the three overlapping models. Once that is settled the rebase is trivial: keep one `poe` block per file, keep this PR's `capabilities` and `notes` changes. Say the word and I will do exactly that mechanically.
Author
Owner

Still needs Jay's decision — re-checked against main 83e2e7c4 (10 commits newer than the original escalation, 2ff67857). Nothing has changed mechanically; the blocker is still which figure is authoritative.

Verified again just now (git merge --no-commit origin/catalogue/1120-1822 in a scratch worktree at 83e2e7c4):

  • 3 CONFLICT (content), unchanged: devices/dahua/nvr5216-16p-ei2.json, devices/hikvision/ds-7608ni-k2-8p.json, devices/hikvision/ds-7616ni-k2-16p.json.
  • One conflict got wider, not narrower: ds-7616ni-k2-16p.json now also disagrees on poe.sourceUrl — main points at the Hikvision product page, this PR at the 2022-10-26 datasheet PDF. Same for the per-port ceiling: main states 802.3at's 30 W "not a vendor figure", this PR states the vendor's 25.5 W.
  • The other five files merge cleanly and are real net-new content, not duplicates: nvr2108hs-8p-4ks3, nvr4208-8p-4ks3, ds-7616nxi-k2-16p have no poe object on main at all, and the capabilities additions (poe-pse, 802.3af, 802.3at) apply cleanly everywhere.

So this is not superseded and not mechanical. Picking a side means choosing a sourcing standard, which is Jay's call. Two viable shapes:

  1. Keep main's figures for the three collided files and let this PR contribute only the net-new devices + capabilities — I can execute that in one rebase on your word, but it means deliberately dropping this PR's note/sourceUrl text on those three.
  2. Take the vendor numbers (25.5 W ceiling, datasheet URLs) and re-source main's three files in a follow-up, which also touches test/poe-budget-prose-six.test.js from #2295.

The branch is 10 commits behind and will keep rotting until one is chosen.

**Still needs Jay's decision** — re-checked against main `83e2e7c4` (10 commits newer than the original escalation, `2ff67857`). Nothing has changed mechanically; the blocker is still which figure is authoritative. **Verified again just now** (`git merge --no-commit origin/catalogue/1120-1822` in a scratch worktree at `83e2e7c4`): - 3 `CONFLICT (content)`, unchanged: `devices/dahua/nvr5216-16p-ei2.json`, `devices/hikvision/ds-7608ni-k2-8p.json`, `devices/hikvision/ds-7616ni-k2-16p.json`. - One conflict got wider, not narrower: `ds-7616ni-k2-16p.json` now also disagrees on `poe.sourceUrl` — main points at the Hikvision product page, this PR at the 2022-10-26 datasheet PDF. Same for the per-port ceiling: main states 802.3at's 30 W "not a vendor figure", this PR states the vendor's 25.5 W. - The other five files merge cleanly and are real net-new content, not duplicates: `nvr2108hs-8p-4ks3`, `nvr4208-8p-4ks3`, `ds-7616nxi-k2-16p` have no `poe` object on main at all, and the `capabilities` additions (`poe-pse`, `802.3af`, `802.3at`) apply cleanly everywhere. So this is not superseded and not mechanical. Picking a side means choosing a sourcing standard, which is Jay's call. Two viable shapes: 1. **Keep main's figures** for the three collided files and let this PR contribute only the net-new devices + capabilities — I can execute that in one rebase on your word, but it means deliberately dropping this PR's `note`/`sourceUrl` text on those three. 2. **Take the vendor numbers** (25.5 W ceiling, datasheet URLs) and re-source main's three files in a follow-up, which also touches `test/poe-budget-prose-six.test.js` from #2295. The branch is 10 commits behind and will keep rotting until one is chosen.
Merge main: #2295 landed three of these seven rows, keep the four still missing
All checks were successful
CI / Classify the change (pull_request) Successful in 18s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 14s
CI / Test (Node 22) (pull_request) Successful in 59s
CI / Test (Node 20) (pull_request) Successful in 1m1s
CI / Server smoke and Docker build (pull_request) Successful in 25s
CI / Test (Node 18) (pull_request) Successful in 1m5s
CI / Arm auto-merge (pull_request) Successful in 4s
0940d922e9
Main's #2295 fixed nvr5216-16p-ei2, ds-7608ni-k2-8p and ds-7616ni-k2-16p
(the three files that conflicted) while this branch was open. Those three take
main's landed wording; this branch now carries only what main still lacks:
three recorders with no poe block at all and a fourth whose 130 W cited a
reseller reprint. The research note is re-based on main (49 silent PSEs ->
46) instead of the branch-point baseline.
jaylouisw changed title from devices(nvr): six recorders sourced 8-16 PoE ports from a 0 W budget to devices(nvr): three recorders still drew from a 0 W budget, and a fourth cited a reseller 2026-11-20 20:05:39 +00:00
jaylouisw deleted branch catalogue/1120-1822 2026-11-20 20:10:07 +00:00
Author
Owner

Resolved — merged as 59b1e7eb without needing the sourcing call.

The branch was reconciled to shape 1 before this merge: it now carries 5 files, and the three contested rows (nvr5216-16p-ei2, ds-7608ni-k2-8p, ds-7616ni-k2-16p) are no longer in the diff at all. Main's wording on those three is untouched on 59b1e7eb — verified on the merged tree: ds-7616ni-k2-16p still reads 200 W, 25.5 W in prose, product-page sourceUrl.

What landed is only the uncontested net-new: nvr2108hs-8p-4ks3, nvr4208-8p-4ks3, nvr5216-16p-4ks2e (incl. the reseller-URL repair), ds-7616nxi-k2-16p, research/nvr-integrated-poe-census.md, and the poe-pse/802.3af/802.3at capabilities.

Verified before merging: merge-tree clean against ab233500, and full suite 498/498 pass, 0 fail on the merged tree.

So the question I escalated — 802.3at's 30 W vs the vendor's 25.5 W — is not a blocker for this PR, but it is still open as an editorial standard: main claims 30 W "not a vendor figure" on Hikvision rows while Dahua rows in the same family quote 25.5 W. Worth one decision eventually; nothing is inconsistent within any single row today.

**Resolved — merged as `59b1e7eb` without needing the sourcing call.** The branch was reconciled to shape 1 before this merge: it now carries 5 files, and the three contested rows (`nvr5216-16p-ei2`, `ds-7608ni-k2-8p`, `ds-7616ni-k2-16p`) are no longer in the diff at all. Main's wording on those three is untouched on 59b1e7eb — verified on the merged tree: `ds-7616ni-k2-16p` still reads 200 W, 25.5 W in prose, product-page `sourceUrl`. What landed is only the uncontested net-new: `nvr2108hs-8p-4ks3`, `nvr4208-8p-4ks3`, `nvr5216-16p-4ks2e` (incl. the reseller-URL repair), `ds-7616nxi-k2-16p`, `research/nvr-integrated-poe-census.md`, and the `poe-pse`/`802.3af`/`802.3at` capabilities. Verified before merging: merge-tree clean against `ab233500`, and full suite **498/498 pass, 0 fail** on the merged tree. So the question I escalated — 802.3at's 30 W vs the vendor's 25.5 W — is not a blocker for this PR, but it is still open as an editorial standard: main claims 30 W "not a vendor figure" on Hikvision rows while Dahua rows in the same family quote 25.5 W. Worth one decision eventually; nothing is inconsistent within any single row today.
Sign in to join this conversation.
No reviewers
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
jaylouisw/packetforge!2297
No description provided.