research(huawei): the E6600 public bundle holds no board census - MIB guide and FAT record read (#803) #2057
No reviewers
Labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
jaylouisw/packetforge!2057
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "catalogue/1119-0027"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Refs #803 - research increment only; the request stays open and blocked.
Evidence:
Result: every member of the public 27-part bundle is now characterised, and none carries a service-board census or port table. Blocker unchanged: one authenticated read of EDOC1100625826 or EDOC1100558549 (配套设备 table), or a commissioning record "E6616X, TM65K32AFB, slots 1-14: n x ".
Gates: syntax-check 430 files OK; audit-port-types OK; hardware-invariants 26/26; request-queue 32/32; doc-citations OK; no spec file changed, so the importer tripwire is not exercised.
needs Jay's decision — competing narrative on issue #803, plus one generated file.
Rebased onto
origin/main(2138cf9a) in a scratch worktree. Conflicts, both purely in the #803 entry:research/open-device-requests.json— entryissue: 803: main's side (landed by #2054, main336a4ba2) rewrotemissingwith the E6616X Product Brochure finding ("the box has its own vendor spec sheet… 680 Gbit/s OTN… 14/12 service board slots"), andunblockwith the note thatscripts/huawei-material-fetch.jsnow 403s. This branch rewrites the same tail ofmissingwith the MIB-guide / FAT-TEST findings and drops main's brochure paragraph;unblockdiffers by that one deleted sentence.checkedis2026-11-19here vs2026-11-18on main.sources— branch adds 4 URLs (the(17)-OptiX OSN E6616 MIBsand(3)-OptiXtrans FAT-TESTassets + their document pages); main adds 3 (brochure PDF, E6616 V100R025C10 产品彩页, Railway brochure). This one is mechanical — union of both.docs/DEVICE_REGISTRY.md— generated from that JSON (scripts/regen-registry-doc.js), so it follows whatever the JSON says; not hand-merged.The decision: one narrative for #803 that keeps main's brochure paragraph and appends this branch's "public bundle now fully characterised, no board census" paragraph, or close this PR in favour of #2054's text. Combining them is research prose, not a mechanical merge, so I have not guessed.
Verified: only these two files conflict; no
engine/,test/or workflow files are touched by the PR (research/optixtrans-e6616x-public-corpus-inventory.mdapplies cleanly). Rebase left no scratch worktree behind.needs Jay's decision — conflict set widened after #2062 landed.
Re-checked against
origin/mainat411aeccfin a quarantine object dir:research/open-device-requests.json— both sides rewrite the issue-803missingprose. This branch adds the 2026-11-19 negative ((17)-OptiX OSN E6616 MIBsand(3)-OptiXtrans FAT-TESTmeasured, no board census) and movescheckedto 2026-11-19; main carries the earlier landing (#2054) with its own E6616X-brochure paragraph andchecked2026-11-18. Unioningsourcesis mechanical; choosing one narrative is not, and neither side is a superset.docs/DEVICE_REGISTRY.md— generated (npm run regen:registry-doc). Not hand-merged; it follows whichever JSON narrative is chosen, so it is blocked on the same call.The branch is also behind 5. Once the issue-803 narrative is chosen I can rebase, regenerate the doc and push; until then this stays open and unmerged.
Note for the next rebase: three prefix dirs in the shared object DB (
objects/90,b5,6c) are root-owned, so any git write that lands a loose object under those prefixes dies withinsufficient permission. Workaround in use:git -c fetch.unpackLimit=1 -c transfer.unpackLimit=1 fetch, plusGIT_OBJECT_DIRECTORY=<scratch> GIT_ALTERNATE_OBJECT_DIRECTORIES=<repo>/objectsfor merge/replay.Rebase status (surgery loop, 2026-11-19 02:0xZ): needs Jay's decision, not a mechanical rebase.
This branch now conflicts wholesale with
origin/maininresearch/open-device-requests.json(whole-file, 138-line block) and in the generateddocs/DEVICE_REGISTRY.md(regenerate, never hand-merge).Cause: #2067 landed a newer #803 narrative from the same optixtrans stream on top of #2065 (commit
630d4a80). The two are complementary, not competing:maincarries the 2026-11-19 "support portal is a signed SPA, not a document index" finding (9,088 chars, no MIB mention).Both are true and neither supersedes the other, so I cannot pick a side: appending one paragraph of prose to the 803 entry is an authoring decision.
Recipe for whoever resolves it (5 min): rebase onto
origin/main, keep main'smissingtext for issue 803, append this branch's MIB-guide sentence to it, keep this branch's new fileresearch/optixtrans-e6616x-public-corpus-inventory.mdand itscheckeddate, thennpm run regen:registry-doc.Also still conflicted on the same files, unchanged since 2026-11-18: #2049 (UDM-Pro row rewrite) and #2033 (
docs/CLI.mdLAG cap prose, which needs an engine-behaviour call).surgery run 2026-11-19 03:0xZ — JSON record merge, needs Jay's decision.
Rebased locally onto
be18b65d: two conflicts, both in the same rendered record.docs/DEVICE_REGISTRY.md— generated bynpm run regen:registry-doc; mechanical, I would regenerate.research/open-device-requests.json— issue 803missing/sources/unblockwere rewritten by #2072 (a9e74b2e), which landed the same subject first. This PR's delta on top of that text is two sentences and some sources:(17)-OptiX OSN E6616 MIBsmeasurement (273 pp, 5,602,882 B, 495,778 chars extracted, no board census),Main's version instead reclassifies the gap ("the census is no longer what is missing — the engine has no E1/T1 port type"). Both are true but they are two different endings for one paragraph, so choosing the phrasing is an editorial call, not a mechanical keep-both. Tell me which wins and I will merge the record and regenerate the registry; or close this as superseded by #2072 if the MIB sentence is not worth its own landing.
docs/DEVICE_REGISTRY.mdand thesourcesarray are otherwise keep-both. Green ona6943435, 20 behind main.surgery run 2026-11-19 04:0xZ — the code is now green; only the record wording blocks.
Head
a6943435carries a full 8-check run, all success (Classify,Smoke,Docker,Test18/20/22,Arm auto-merge) — first time this branch has been green. Nothing about the E6600 census is in question on CI.It still cannot merge: 3 conflict hunks against
06830e13—docs/DEVICE_REGISTRY.md×2 (generated; surgery regenerates that, it is not the blocker) andresearch/open-device-requests.json×1, a whole-entry rewrite of the 1119-0027why/unblocktext where main and this branch both rewrote the same string. Keeping both sides would splice two research narratives into one field.Needs Jay: pick the E6600 wording (or say "keep this branch's"), and surgery can regenerate the registry, re-run, and merge the same night.
a6943435f1ee4d9f18cf