Catalogue: a stated build belongs in the label, not the model (stated-configurations steps 1-2) #1879

Merged
jaylouisw merged 1 commit from catalogue/1117-1755 into main 2026-11-17 18:14:10 +00:00
Owner

Steps 1-2 of docs/plans/stated-configurations.md, which had to land before any OptiXtrans spec does.

What changed

The rule. docs/DEVICE_SCHEMA.md gains § "A port list may be a stated build": a modular shelf has no ports of its own, so a spec states a card complement; the complement goes in label, never in model; aliases carries the bare names that keep search working; notes must state the slot count, the card part number and the table it came from; and when the cited document does not carry the census, notes says so and names the document that does. Precedent cited: devices/arista/7504r.json, devices/huawei/netengine-8000-x4.json.

The three specs. optixtrans-e6616, optixtrans-e6608 and optixtrans-e6608t lose the complement from model (OptiXtrans E6616 (14x B1LDCD) -> OptiXtrans E6616). Their label already carried it verbatim, so nothing moved into thin air; the port lists are untouched. Each alias that became identical to its own model is removed, because test/device-aliases.test.js refuses an alias that is its own model name — the old complement was what kept "OptiXtrans E6616" a distinct alias. E6616, E6608, E6608T, B1LDCD and TMB1LDCD still resolve, which is the check the plan asked for rather than a change to make.

The citation gap, stated where the claim is. For the two specs whose gap is measured (research/optixtrans-e6600-xt-census.md §5), notes now records that the cited E6608/E6616 document carries the frame and cross-connect figures but not the board census: the E6616 datasheet's text layer has no board part numbers, and TMB1LDCD's front panel is stated by Huawei's boards-overview chapter EDOC1100156575BEJ09044, which answers 302 to every user-agent (re-measured today: 302 on support.huawei.com and support.huawei.cn, 404 under e.huawei.com/hedex/). A spec that quotes a source it did not read is the error this catalogue has corrected most often, so the gap now sits next to the claim instead of only in research.

Also in this branch, from today's pass on the queue: nothing else — the five open [Device]: requests are unchanged and unexplained by this PR (see below).

What this does not close

  • #803 E6616X / #804 E6608T: the E6600 board list is still hedex-only. Today's pass did find a genuinely new vendor route — https://e.huawei.com/en/documents/products/enterprise-optical-network/<hash> is a server-rendered document tree that hands over the asset path, and through it the OptiXtrans E6616X V100R026C00 Product Brochure (.../Material/91ebb2.../M3T1A590N1296415560500404267/OptiXtrans E6616X V100R026C00 Product Brochure.docx, 1,628,729 B) and the R025 PDF brochure (216,145 B, 4 pp). Both were read end to end: the only board part number in either is the cross-connect TM65UXCL, so the brochure route does not carry the client-board census either. That is recorded, not guessed at.
  • #807 / #808 / #810: still blocked on an OTN port type (step 6, needs Jay's approval). Their slot arithmetic is untouched by this PR.

Gates

node scripts/syntax-check.js — 392 files, all OK · node scripts/audit-port-types.js — "No port contradicts its declared type" · check-control-bytes.js — clean · regen-registry-doc.js — "The catalogue pages are current: 2,348 models across 258 vendors, 29,235 sockets".

Suites run individually (each exit 0): catalogue-hygiene 25/25, device-aliases 13/13, palette-search-keys 19/19, inspector-port-count 10/10, spec-vs-description 3/3, duplicate-models 21/21, capability-vocabulary 13/13, hardware-invariants 26/26, import-devices-capabilities 34/34 (the importer tripwire), importer-roundtrip 26/26, inspector-labels 47/47, palette-selfsearch 5/5.

Co-authored-by: Hermes Agent jay@jayw.uk

Steps 1-2 of `docs/plans/stated-configurations.md`, which had to land **before** any OptiXtrans spec does. ## What changed **The rule.** `docs/DEVICE_SCHEMA.md` gains § "A port list may be a stated build": a modular shelf has no ports of its own, so a spec states a card complement; the complement goes in `label`, never in `model`; `aliases` carries the bare names that keep search working; `notes` must state the slot count, the card part number and **the table it came from**; and when the cited document does not carry the census, `notes` says so and names the document that does. Precedent cited: `devices/arista/7504r.json`, `devices/huawei/netengine-8000-x4.json`. **The three specs.** `optixtrans-e6616`, `optixtrans-e6608` and `optixtrans-e6608t` lose the complement from `model` (`OptiXtrans E6616 (14x B1LDCD)` -> `OptiXtrans E6616`). Their `label` already carried it verbatim, so nothing moved into thin air; the port lists are untouched. Each alias that became identical to its own model is removed, because `test/device-aliases.test.js` refuses an alias that is its own model name — the old complement was what kept `"OptiXtrans E6616"` a distinct alias. `E6616`, `E6608`, `E6608T`, `B1LDCD` and `TMB1LDCD` still resolve, which is the check the plan asked for rather than a change to make. **The citation gap, stated where the claim is.** For the two specs whose gap is measured (`research/optixtrans-e6600-xt-census.md` §5), `notes` now records that the cited E6608/E6616 document carries the frame and cross-connect figures but **not** the board census: the E6616 datasheet's text layer has no board part numbers, and TMB1LDCD's front panel is stated by Huawei's boards-overview chapter `EDOC1100156575BEJ09044`, which answers **302** to every user-agent (re-measured today: 302 on `support.huawei.com` and `support.huawei.cn`, 404 under `e.huawei.com/hedex/`). A spec that quotes a source it did not read is the error this catalogue has corrected most often, so the gap now sits next to the claim instead of only in research. **Also in this branch, from today's pass on the queue:** nothing else — the five open `[Device]:` requests are unchanged and unexplained by this PR (see below). ## What this does not close - **#803 E6616X / #804 E6608T**: the E6600 board list is still hedex-only. Today's pass did find a genuinely new vendor route — `https://e.huawei.com/en/documents/products/enterprise-optical-network/<hash>` is a server-rendered document tree that hands over the asset path, and through it the **OptiXtrans E6616X V100R026C00 Product Brochure** (`.../Material/91ebb2.../M3T1A590N1296415560500404267/OptiXtrans E6616X V100R026C00 Product Brochure.docx`, 1,628,729 B) and the R025 PDF brochure (216,145 B, 4 pp). Both were read end to end: the *only* board part number in either is the cross-connect `TM65UXCL`, so the brochure route does not carry the client-board census either. That is recorded, not guessed at. - **#807 / #808 / #810**: still blocked on an OTN port type (step 6, needs Jay's approval). Their slot arithmetic is untouched by this PR. ## Gates `node scripts/syntax-check.js` — 392 files, all OK · `node scripts/audit-port-types.js` — "No port contradicts its declared type" · `check-control-bytes.js` — clean · `regen-registry-doc.js` — "The catalogue pages are current: 2,348 models across 258 vendors, 29,235 sockets". Suites run individually (each exit 0): `catalogue-hygiene` 25/25, `device-aliases` 13/13, `palette-search-keys` 19/19, `inspector-port-count` 10/10, `spec-vs-description` 3/3, `duplicate-models` 21/21, `capability-vocabulary` 13/13, `hardware-invariants` 26/26, `import-devices-capabilities` 34/34 (the importer tripwire), `importer-roundtrip` 26/26, `inspector-labels` 47/47, `palette-selfsearch` 5/5. Co-authored-by: Hermes Agent <jay@jayw.uk>
catalogue: a stated build belongs in the label, not the model
All checks were successful
CI / Classify the change (pull_request) Successful in 11s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 15s
CI / Server smoke test (pull_request) Successful in 7s
CI / Test (Node 20) (pull_request) Successful in 51s
CI / Test (Node 18) (pull_request) Successful in 1m2s
CI / Docker build (pull_request) Successful in 34s
CI / Test (Node 22) (pull_request) Successful in 1m9s
CI / Arm auto-merge (pull_request) Successful in 6s
7fc7caf65f
The E6608, E6608T and E6616 carried their card complement inside `model`
("OptiXtrans E6616 (14x B1LDCD)"). `model` is the variant and the filename,
so a complement there makes the same shelf unfindable under the name a
reader is holding, and it puts a build where the schema says the SKU goes.
The complement moves to `label`, which already carried it verbatim, and the
bare names stay in `aliases` - minus each alias that has just become
identical to its own model, which the alias invariant refuses by name.

`docs/DEVICE_SCHEMA.md` gains the rule that made this a decision instead of a
style preference: the complement lives in `label`, `notes` must state the slot
count, the card part number and the table it came from, and when the cited
document does not carry the card census, `notes` says so and names the
document that does. Precedent cited: arista/7504r and huawei/netengine-8000-x4.

Applied to the two specs where the citation gap is measured
(research/optixtrans-e6600-xt-census.md section 5): the cited E6616 datasheet
has no board part numbers in its text layer, and the TMB1LDCD front panel lives
in the hedex-only boards-overview chapter EDOC1100156575BEJ090044, which answers
302 to every user-agent. A spec that quotes a source it did not read is the
error this catalogue has corrected most often, so the gap is now stated where
the claim is made.

Does not close any request: #803/#804 still need the E6600 board list, and
#807/#808/#810 still need an OTN port type (docs/plans/stated-configurations.md
step 6). This is steps 1-2 of that plan, which had to land before any OptiXtrans
spec does.

Gates: syntax-check 392 files OK; audit-port-types "no port contradicts its
declared type"; check-control-bytes clean; catalogue-hygiene 25/25;
device-aliases 13/13; palette-search-keys 19/19; inspector-port-count 10/10;
spec-vs-description 3/3; duplicate-models 21/21; capability-vocabulary 13/13;
hardware-invariants 26/26; import-devices-capabilities 34/34;
importer-roundtrip 26/26; inspector-labels 47/47; regen-registry-doc reports
the pages current.
jaylouisw force-pushed catalogue/1117-1755 from 7fc7caf65f
All checks were successful
CI / Classify the change (pull_request) Successful in 11s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 15s
CI / Server smoke test (pull_request) Successful in 7s
CI / Test (Node 20) (pull_request) Successful in 51s
CI / Test (Node 18) (pull_request) Successful in 1m2s
CI / Docker build (pull_request) Successful in 34s
CI / Test (Node 22) (pull_request) Successful in 1m9s
CI / Arm auto-merge (pull_request) Successful in 6s
to 989d2a5c7b
All checks were successful
CI / Classify the change (pull_request) Successful in 8s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 10s
CI / Server smoke test (pull_request) Successful in 10s
CI / Docker build (pull_request) Successful in 23s
CI / Test (Node 20) (pull_request) Successful in 50s
CI / Test (Node 22) (pull_request) Successful in 53s
CI / Test (Node 18) (pull_request) Successful in 58s
CI / Arm auto-merge (pull_request) Successful in 4s
2026-11-17 18:10:31 +00:00
Compare
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!1879
No description provided.