Engine: a CFP2 cage, so optical transport stops apologising for its own specs #1213

Merged
jaylouisw merged 1 commit from feat/media-cfp2 into main 2026-11-09 14:06:25 +00:00
Owner

The OptiXtrans E6616 and E6608 have 200G line ports. There was no CFP cage in the engine, so those ports were modelled as QSFP56 and the note said: "the physical module is CFP2, so a lab planning optics should order CFP2, not QSFP56."

A spec telling the reader not to believe the spec is worse than a missing entry - a lab that wires the modelled port with the modelled module builds a link the hardware cannot build, and the simulator would have accepted it.

cfp2 is now a cage: 100/200G, optical, media 200gbase-fr4/dr4/100gbase-lr4/40gbase-lr4, no DAC (it is a line-side module, not a patch lead). 40G because a CFP2 port takes the older optic; 100G because a muxponder board has a 100G line mode.

The interesting part is the refusal

CAGE_MATE.cfp2 contains exactly one entry: cfp2. This is the first cage in the catalogue where two cages carry the same rate and must still refuse each other - a 200G CFP2 module is roughly four times the footprint of a 200G QSFP56 and latches differently. "The rates match, so it fits" is wrong here, and the validator says so by name:

E_CONNECTOR_MISMATCH: Cannot mate CFP2 to QSFP56 (incompatible connectors).

Only CFP2

CFP, CFP4 and CFP8 are real standards, one line of table each. None is declared: no device uses them yet, and the cable matrix reports a declared-but-unreferenced cage as unused - speculative vocabulary makes that signal useless. When a box needs CFP8 it is a one-line addition plus its matrix row.

Changed with it

  • Both OptiXtrans specs: 40 line ports moved from qsfp56 to cfp2, apology paragraphs replaced with the refusal they should always have had.
  • test/cable-matrix.test.js: cfp2 x cfp2 accepts, cfp2 x * refuses.
  • scripts/audit-port-types.js: cfp2 joins FIBRE_CAGES.
  • docs/CONNECTORS.md: table row plus §4.5 "CFP2 is not a large QSFP". While there, fixed a stale count in docs/PHYSICAL_LAYER.md - it claimed "18 of them" when the engine has had 27 cages for a while. Numbers in prose rot; that sentence no longer carries one.

30 checks in test/media-cfp2.test.js, including the E6616 line ports cabling to each other at 200G and refusing a QSFP56 port with both cage names in the message. Suite: 117 suites, 5,435 assertions, 0 failed.

The OptiXtrans E6616 and E6608 have 200G line ports. There was no CFP cage in the engine, so those ports were modelled as **QSFP56** and the note said: *"the physical module is CFP2, so a lab planning optics should order CFP2, not QSFP56."* A spec telling the reader not to believe the spec is worse than a missing entry - a lab that wires the modelled port with the modelled module builds a link the hardware cannot build, and the simulator would have accepted it. **`cfp2` is now a cage:** 100/200G, optical, media `200gbase-fr4`/`dr4`/`100gbase-lr4`/`40gbase-lr4`, no DAC (it is a line-side module, not a patch lead). 40G because a CFP2 port takes the older optic; 100G because a muxponder board has a 100G line mode. ### The interesting part is the refusal `CAGE_MATE.cfp2` contains exactly one entry: `cfp2`. This is the first cage in the catalogue where two cages carry the **same rate** and must still refuse each other - a 200G CFP2 module is roughly four times the footprint of a 200G QSFP56 and latches differently. "The rates match, so it fits" is wrong here, and the validator says so by name: ``` E_CONNECTOR_MISMATCH: Cannot mate CFP2 to QSFP56 (incompatible connectors). ``` ### Only CFP2 CFP, CFP4 and CFP8 are real standards, one line of table each. None is declared: no device uses them yet, and the cable matrix reports a declared-but-unreferenced cage as *unused* - speculative vocabulary makes that signal useless. When a box needs CFP8 it is a one-line addition plus its matrix row. ### Changed with it - Both OptiXtrans specs: 40 line ports moved from `qsfp56` to `cfp2`, apology paragraphs replaced with the refusal they should always have had. - `test/cable-matrix.test.js`: `cfp2 x cfp2` accepts, `cfp2 x *` refuses. - `scripts/audit-port-types.js`: `cfp2` joins `FIBRE_CAGES`. - `docs/CONNECTORS.md`: table row plus **§4.5 "CFP2 is not a large QSFP"**. While there, fixed a stale count in `docs/PHYSICAL_LAYER.md` - it claimed "18 of them" when the engine has had 27 cages for a while. Numbers in prose rot; that sentence no longer carries one. 30 checks in `test/media-cfp2.test.js`, including the E6616 line ports cabling to each other at 200G and **refusing** a QSFP56 port with both cage names in the message. Suite: 117 suites, 5,435 assertions, 0 failed.
Engine: a CFP2 cage, so optical transport stops apologising for its own specs
All checks were successful
CI / Classify the change (pull_request) Successful in 10s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 12s
CI / Test (Node 20) (pull_request) Successful in 1m52s
CI / Test (Node 18) (pull_request) Successful in 1m56s
CI / Server smoke test (pull_request) Successful in 36s
CI / Docker build (pull_request) Successful in 37s
CI / Test (Node 22) (pull_request) Successful in 1m44s
471c4b67bd
The OptiXtrans E6616 and E6608 have 200G line ports. There was no CFP cage in the engine, so those
ports were modelled as **QSFP56** and the note said: *"the physical module is CFP2, so a lab planning
optics should order CFP2, not QSFP56."*

A spec telling the reader not to believe the spec is worse than a missing entry. A lab that wires the
modelled port with the modelled module builds a link the hardware cannot build - and the simulator
would have accepted it.

**`cfp2` is now a cage: 100/200G, optical, media `200gbase-fr4`/`dr4`/`100gbase-lr4`/`40gbase-lr4`,
no DAC** (it is a line-side module, not a patch lead). 40G is in the list because a CFP2 port takes
the older optic; 100G because a muxponder board has a 100G line mode.

### The interesting part is the refusal

`CAGE_MATE.cfp2` contains exactly one entry: `cfp2`. This is the first cage in the catalogue where
two cages carry the **same rate** and must still refuse each other - a 200G CFP2 module is roughly
four times the footprint of a 200G QSFP56 and latches differently. "The rates match, so it fits" is
wrong here, and the validator says so by name:

```
E_CONNECTOR_MISMATCH: Cannot mate CFP2 to QSFP56 (incompatible connectors).
```

### Only CFP2

CFP, CFP4 and CFP8 are real standards and each would be one line of table. None is declared, because
no device in the catalogue uses them yet and the cable matrix reports a declared-but-unreferenced
cage as *unused* - speculative vocabulary makes that signal useless. When a box needs CFP8, it is a
one-line addition plus its matrix row.

### Changed with it

- Both OptiXtrans specs: 40 line ports moved from `qsfp56` to `cfp2`, and the apology paragraphs
  replaced with the refusal they should always have had.
- `test/cable-matrix.test.js`: `cfp2 x cfp2` accepts, `cfp2 x *` refuses, so the matrix has a row to
  hold the new cage to.
- `scripts/audit-port-types.js`: `cfp2` joins `FIBRE_CAGES`, so the copper-in-a-fibre-cage rule
  applies to it.
- `docs/CONNECTORS.md`: table row plus **§4.5 "CFP2 is not a large QSFP"**. While there, fixed a
  stale count in `docs/PHYSICAL_LAYER.md` - it claimed "18 of them" when the engine has had 27 cages
  for a while. Numbers in prose rot; the sentence no longer carries one.

30 checks in `test/media-cfp2.test.js`, including the E6616 line ports cabling to each other at 200G
and **refusing** a QSFP56 port with both cage names in the message. Suite: 117 suites, 5,435
assertions, 0 failed.
jaylouisw scheduled this pull request to auto merge when all checks succeed 2026-11-09 14:06:22 +00:00
jaylouisw force-pushed feat/media-cfp2 from 471c4b67bd
All checks were successful
CI / Classify the change (pull_request) Successful in 10s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 12s
CI / Test (Node 20) (pull_request) Successful in 1m52s
CI / Test (Node 18) (pull_request) Successful in 1m56s
CI / Server smoke test (pull_request) Successful in 36s
CI / Docker build (pull_request) Successful in 37s
CI / Test (Node 22) (pull_request) Successful in 1m44s
to 1d2bbf6b12
All checks were successful
CI / Classify the change (pull_request) Successful in 10s
PR rebase / Rebase onto main and arm auto-merge (pull_request) Successful in 12s
CI / Test (Node 18) (pull_request) Successful in 1m55s
CI / Test (Node 20) (pull_request) Successful in 1m39s
CI / Server smoke test (pull_request) Successful in 27s
CI / Docker build (pull_request) Successful in 41s
CI / Test (Node 22) (pull_request) Successful in 1m38s
2026-11-09 14:06:23 +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!1213
No description provided.