poe: eight budgets that only existed in prose, and the audit that counts the rest #1992

Merged
jaylouisw merged 1 commit from catalogue/1118-1426 into main 2026-11-18 14:44:55 +00:00
Owner

Refs #130 - it does not close it. UDW-PRO still has no published port or PoE census; this fixes the class the issue's fallback exposed.

devices/ubiquiti/udw.json carried twelve poeOut ports and no poe.budgetW, so Device fell through to the model-name parser, got nothing from UDW, and show power inline printed "budget not modelled" over live ports (#1979 fixed that one file). The invariant suite only polices the opposite direction - a budget with no port to spend it on - so nothing counted the rest.

The census

node scripts/audit-poe-budget.js (also npm run audit:poe) over the whole catalogue: 401 models source PoE on at least one port; 130 silent (PoE out, no budget, no "never published" marker); 22 inferred from the model name rather than a document; 16 explicitly unknown; 0 orphaned.

It exits 0 unless --strict: gating the class today would fail the build on the 130. research/poe-budget-field-census.md records the end state (zero silent, then --strict in the test job).

The audit corrected itself twice, both written down: read the engine's hardware.poeOut rather than the spec's literal key (a port template's generated ports never carry it - the first cut reported 37 false orphans), and require the wattage before the word "budget" (Teltonika's "240 W total budget and up to 30 W on any one port" read as 30 W) with a thousands separator (1,050 W read as 50 W).

Eight budgets promoted, each re-read from the vendor today

spec W vendor line
devices/tp-link/tl-sg1016pe.json 110 https://www.tp-link.com/us/business-networking/easy-smart-switch/tl-sg1016pe/ - "total PoE power budget of 110W, up to 30W per port"
devices/tp-link/tl-sg1210mpe.json 123 https://www.tp-link.com/us/business-networking/soho-switch-poe/tl-sg1210mpe/ - spec table "PoE Power Budget: 123 W"
devices/tp-link/tl-sg1218mp.json 250 https://www.tp-link.com/us/business-networking/poe-switch/tl-sg1218mp/v1/ - "250 W for all PoE ports"
devices/teltonika/tsf000.json 120 https://wiki.teltonika-networks.com/view/TSF000 - "4x PoE+ outputs with a 120 W power budget"
devices/teltonika/tsw040.json 240 https://wiki.teltonika-networks.com/view/TSW040 - "total power budget of 240 W, or up to 30 W per port"
devices/ubiquiti/usw-pro-hd-24-poe.json 600 https://techspecs.ui.com/unifi/switching/usw-pro-hd-24-poe - "Total PoE Availability 600W"
devices/ubiquiti/usw-enterprise-48-poe.json 720 https://techspecs.ui.com/unifi/switching/usw-enterprise-48-poe - "Total PoE Availability 720W"
devices/ubiquiti/ecs-24-poe.json 1050 https://techspecs.ui.com/unifi/switching/ecs-24-poe - "1,050W with (2) 600W modules" (shared mode)

Three of those prose figures sat beside a bigger, wronger number: 141.54 W input power, 660 W including PoE output, an 870 W supply rating. That is how a lab over-subscribes a budget.

One is a stated configuration, not a number: ECS-24-PoE publishes 1,050 / 450 W with 2x600 W modules and 2,250 / 1,850 / 1,050 / 850 W with 2x1,200 W modules. The spec records 1,050 W - reachable at any input voltage only in the 600 W-module chassis - and the poe.note carries all six figures. Same shape as the E6616X AC/DC slot arithmetic (#803): a capacity that depends on what is fitted.

Gates

  • node scripts/syntax-check.js - 417 files, all OK
  • node scripts/audit-port-types.js - no port contradicts its declared type
  • node scripts/audit-devices.js - no hard defects
  • node scripts/regen-registry-doc.js - catalogue pages current (2,356 models / 258 vendors / 29,721 sockets); no doc churn
  • npm test - 343 suites, 11,908 assertions, 0 failed (653 s)
  • targeted: test/hardware-invariants.test.js 26, test/poe.test.js 18, test/poe-inference.test.js 17

No engine behaviour change: the new script is an audit, and every device edit adds a poe block the engine already understood.

Refs #130 - it does not close it. UDW-PRO still has no published port or PoE census; this fixes the *class* the issue's fallback exposed. `devices/ubiquiti/udw.json` carried twelve `poeOut` ports and no `poe.budgetW`, so `Device` fell through to the model-name parser, got nothing from `UDW`, and `show power inline` printed "budget not modelled" over live ports (#1979 fixed that one file). The invariant suite only polices the opposite direction - a budget with no port to spend it on - so nothing counted the rest. ## The census `node scripts/audit-poe-budget.js` (also `npm run audit:poe`) over the whole catalogue: **401** models source PoE on at least one port; **130 silent** (PoE out, no budget, no "never published" marker); **22 inferred from the model name** rather than a document; **16 explicitly unknown**; **0 orphaned**. It exits 0 unless `--strict`: gating the class today would fail the build on the 130. `research/poe-budget-field-census.md` records the end state (zero silent, then `--strict` in the test job). The audit corrected itself twice, both written down: read the engine's `hardware.poeOut` rather than the spec's literal key (a port template's generated ports never carry it - the first cut reported 37 false orphans), and require the wattage **before** the word "budget" (Teltonika's "240 W total budget and up to 30 W on any one port" read as 30 W) with a thousands separator (1,050 W read as 50 W). ## Eight budgets promoted, each re-read from the vendor today | spec | W | vendor line | |--|--|--| | devices/tp-link/tl-sg1016pe.json | 110 | https://www.tp-link.com/us/business-networking/easy-smart-switch/tl-sg1016pe/ - "total PoE power budget of 110W, up to 30W per port" | | devices/tp-link/tl-sg1210mpe.json | 123 | https://www.tp-link.com/us/business-networking/soho-switch-poe/tl-sg1210mpe/ - spec table "PoE Power Budget: 123 W" | | devices/tp-link/tl-sg1218mp.json | 250 | https://www.tp-link.com/us/business-networking/poe-switch/tl-sg1218mp/v1/ - "250 W for all PoE ports" | | devices/teltonika/tsf000.json | 120 | https://wiki.teltonika-networks.com/view/TSF000 - "4x PoE+ outputs with a 120 W power budget" | | devices/teltonika/tsw040.json | 240 | https://wiki.teltonika-networks.com/view/TSW040 - "total power budget of 240 W, or up to 30 W per port" | | devices/ubiquiti/usw-pro-hd-24-poe.json | 600 | https://techspecs.ui.com/unifi/switching/usw-pro-hd-24-poe - "Total PoE Availability 600W" | | devices/ubiquiti/usw-enterprise-48-poe.json | 720 | https://techspecs.ui.com/unifi/switching/usw-enterprise-48-poe - "Total PoE Availability 720W" | | devices/ubiquiti/ecs-24-poe.json | 1050 | https://techspecs.ui.com/unifi/switching/ecs-24-poe - "1,050W with (2) 600W modules" (shared mode) | Three of those prose figures sat beside a bigger, wronger number: 141.54 W input power, 660 W including PoE output, an 870 W supply rating. That is how a lab over-subscribes a budget. One is a stated configuration, not a number: `ECS-24-PoE` publishes 1,050 / 450 W with 2x600 W modules and 2,250 / 1,850 / 1,050 / 850 W with 2x1,200 W modules. The spec records 1,050 W - reachable at any input voltage only in the 600 W-module chassis - and the `poe.note` carries all six figures. Same shape as the E6616X AC/DC slot arithmetic (#803): a capacity that depends on what is fitted. ## Gates - `node scripts/syntax-check.js` - 417 files, all OK - `node scripts/audit-port-types.js` - no port contradicts its declared type - `node scripts/audit-devices.js` - no hard defects - `node scripts/regen-registry-doc.js` - catalogue pages current (2,356 models / 258 vendors / 29,721 sockets); no doc churn - `npm test` - 343 suites, 11,908 assertions, 0 failed (653 s) - targeted: `test/hardware-invariants.test.js` 26, `test/poe.test.js` 18, `test/poe-inference.test.js` 17 No engine behaviour change: the new script is an audit, and every device edit adds a `poe` block the engine already understood.
poe: eight budgets that only existed in prose, and the audit that finds the rest
All checks were successful
CI / Classify the change (pull_request) Successful in 7s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 10s
CI / Server smoke test (pull_request) Successful in 6s
CI / Docker build (pull_request) Successful in 19s
CI / Test (Node 20) (pull_request) Successful in 54s
CI / Test (Node 22) (pull_request) Successful in 58s
CI / Test (Node 18) (pull_request) Successful in 1m2s
CI / Arm auto-merge (pull_request) Successful in 19s
9f38c8aa70
#130 found the class in one file: `devices/ubiquiti/udw.json` carried twelve
`poeOut` ports and no `poe.budgetW`, so `Device` fell through to the model-name
parser, got nothing from `UDW`, and `show power inline` printed "budget not
modelled" over live ports. The device was present; the number was not. The
invariant suite only polices the opposite direction - a budget with no port to
spend it on - so nothing caught it.

This adds `scripts/audit-poe-budget.js`, which counts the four cohorts over the
whole catalogue: 401 models source PoE, 130 are silent (PoE out, no budget, no
"never published" marker), 22 take their budget from the model name rather than a
document, 16 are explicitly unknown, 0 are orphaned. It exits 0 unless --strict,
because gating the class today would fail the build on the 130; the note in
`research/poe-budget-field-census.md` records the end state.

It also promotes eight budgets that existed only as prose, each re-read from the
vendor's page today: TL-SG1016PE 110 W, TL-SG1210MPE 123 W, TL-SG1218MP 250 W,
TSF000 120 W, TSW040 240 W, USW-Pro-HD-24-PoE 600 W, USW-Enterprise-48-PoE
720 W, ECS-24-PoE 1,050 W. Three of those prose figures sat beside a bigger,
wronger number - 141.54 W input power, 660 W including PoE output, an 870 W
supply rating - which is exactly how a lab ends up over-subscribing a budget.

The audit had to correct itself twice, both recorded in the note: read the
engine's `hardware.poeOut`, not the spec's literal key (a port template's
generated ports never carry it, which reported 37 false orphans), and require
the figure BEFORE the word "budget" (Teltonika's "240 W total budget and up to
30 W on any one port" read as 30 W), with a thousands separator (1,050 W read as
50 W).

Refs #130 - it does not close it; UDW-PRO still has no published port or PoE
census.
jaylouisw force-pushed catalogue/1118-1426 from 9f38c8aa70
All checks were successful
CI / Classify the change (pull_request) Successful in 7s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 10s
CI / Server smoke test (pull_request) Successful in 6s
CI / Docker build (pull_request) Successful in 19s
CI / Test (Node 20) (pull_request) Successful in 54s
CI / Test (Node 22) (pull_request) Successful in 58s
CI / Test (Node 18) (pull_request) Successful in 1m2s
CI / Arm auto-merge (pull_request) Successful in 19s
to 94f3af3429
All checks were successful
CI / Classify the change (pull_request) Successful in 9s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 7s
CI / Server smoke test (pull_request) Successful in 7s
CI / Docker build (pull_request) Successful in 21s
CI / Test (Node 20) (pull_request) Successful in 55s
CI / Test (Node 18) (pull_request) Successful in 1m2s
CI / Test (Node 22) (pull_request) Successful in 55s
CI / Arm auto-merge (pull_request) Successful in 10s
2026-11-18 14:42:48 +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!1992
No description provided.