research: the hedex content API, its two ids, and the retired host that used to serve it (#807) #1894

Merged
jaylouisw merged 1 commit from fix/1117-2006 into main 2026-11-17 20:46:24 +00:00
Owner

Refs #807 - records evidence, does not close the issue.

What is new. The hedex viewer is a shell that loads one bundle, /hedex/lst/js/index-legacy-DaAKRolI.js (200, 361,996 B). That bundle names the content endpoints and their template: /hedex/api/pages/${docId}/${topicLibId}/${topicLibVer}/${topicUrl} (its own RegExp is /hedex/api/pages/[^/]+/([^/]+)/([^/]+)/([^#?]+)), plus /hedex/api/navi/ and /hedex/api/search/. Earlier passes had the document id only; the 302 from any /hedex/pages/... URL hands back the missing pair in its Location header (tocLibId=BEJ09047&tocLibVer=03 for the OSN 9800 set).

What the API answers (crawler UA, 2026-11-17):

  • /hedex/api/pages/EDOC1100126589/BEJ09047/03/resources/help/wdm/9800/hd/line/n210/n210.html -> 404, 18 B, No doc data found. (same without .html)
  • /hedex/api/navi/EDOC1100126589/first-level-tree -> 404, {"status":404,"message":"No doc data found.","data":null}
  • https://support.huawei.cn/hedex/api/pages/EDOC1100625542/BEJ09047/03/resources/help -> 401, NOT LOGIN!
  • https://support.huawei.cn/enterprise/en/doc/EDOC1100163901 -> 302 -> view/notFound, 404, 452,479 B

So the wall is per-document (404 = not served, 401 = needs a session), the .cn mirror repeats it byte-for-byte, and no first-level-tree/topic call will disclose a lib/ver pair to a caller without a session.

The one host that used to leak it is retired. info.support.huawei.com still appears in search as an openly readable hedex content host (.../hedex/api/pages/EDOC1000163559/YEI0812D/19/resources/en-us_topic_0000001238552187.html, and the pt/br/fr E9605 product pages), but every path now answers 302 -> https://support.huawei.com/supportindex/ (200, 91,294 B). Wayback has no capture of .../hedex/api/pages/EDOC1100625542 or .../EDOC1100163901 (CDX prefix query: 200, 0 bytes), while the host itself is archived from 2021-02-18 on.

Not an access gap in the ordinary sense. e-file.huawei.com/.../OptiXtrans%20E9600%20TNG2SRAPXF.pdf fetched today: 200, application/pdf, 1,104,102 B, pdftotext -layout prints section 1.9 Valid Slots for E9605/E9612/E9624/E9624X. Same family, same server, anonymous - so a published client-board chapter would be reachable; the TNG1N* ones are simply not published.

Gates: node scripts/syntax-check.js (393 files, all OK), node scripts/audit-port-types.js (no port contradicts its declared type), node scripts/check-canonical-urls.js (3071 files OK). No spec touched, so no importer/catalogue change.

Spec path: research/optixtrans-hedex-content-api.md.

Refs #807 - records evidence, does not close the issue. **What is new.** The hedex viewer is a shell that loads one bundle, `/hedex/lst/js/index-legacy-DaAKRolI.js` (200, 361,996 B). That bundle names the content endpoints and their template: `/hedex/api/pages/${docId}/${topicLibId}/${topicLibVer}/${topicUrl}` (its own RegExp is `/hedex/api/pages/[^/]+/([^/]+)/([^/]+)/([^#?]+)`), plus `/hedex/api/navi/` and `/hedex/api/search/`. Earlier passes had the document id only; the `302` from any `/hedex/pages/...` URL hands back the missing pair in its `Location` header (`tocLibId=BEJ09047&tocLibVer=03` for the OSN 9800 set). **What the API answers** (crawler UA, 2026-11-17): - `/hedex/api/pages/EDOC1100126589/BEJ09047/03/resources/help/wdm/9800/hd/line/n210/n210.html` -> 404, 18 B, `No doc data found.` (same without `.html`) - `/hedex/api/navi/EDOC1100126589/first-level-tree` -> 404, `{"status":404,"message":"No doc data found.","data":null}` - `https://support.huawei.cn/hedex/api/pages/EDOC1100625542/BEJ09047/03/resources/help` -> 401, `NOT LOGIN!` - `https://support.huawei.cn/enterprise/en/doc/EDOC1100163901` -> 302 -> `view/notFound`, 404, 452,479 B So the wall is per-document (404 = not served, 401 = needs a session), the `.cn` mirror repeats it byte-for-byte, and no `first-level-tree`/`topic` call will disclose a lib/ver pair to a caller without a session. **The one host that used to leak it is retired.** `info.support.huawei.com` still appears in search as an openly readable hedex content host (`.../hedex/api/pages/EDOC1000163559/YEI0812D/19/resources/en-us_topic_0000001238552187.html`, and the `pt`/`br`/`fr` E9605 product pages), but every path now answers 302 -> `https://support.huawei.com/supportindex/` (200, 91,294 B). Wayback has no capture of `.../hedex/api/pages/EDOC1100625542` or `.../EDOC1100163901` (CDX prefix query: 200, 0 bytes), while the host itself is archived from 2021-02-18 on. **Not an access gap in the ordinary sense.** `e-file.huawei.com/.../OptiXtrans%20E9600%20TNG2SRAPXF.pdf` fetched today: 200, `application/pdf`, 1,104,102 B, `pdftotext -layout` prints section 1.9 Valid Slots for E9605/E9612/E9624/E9624X. Same family, same server, anonymous - so a published client-board chapter would be reachable; the `TNG1N*` ones are simply not published. Gates: `node scripts/syntax-check.js` (393 files, all OK), `node scripts/audit-port-types.js` (no port contradicts its declared type), `node scripts/check-canonical-urls.js` (3071 files OK). No spec touched, so no importer/catalogue change. Spec path: `research/optixtrans-hedex-content-api.md`.
ping: a box answers an echo to an address it holds, off the wire
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 24s
CI / Server smoke test (pull_request) Successful in 11s
CI / Test (Node 18) (pull_request) Successful in 49s
CI / Test (Node 20) (pull_request) Successful in 47s
CI / Docker build (pull_request) Successful in 20s
CI / Test (Node 22) (pull_request) Successful in 48s
CI / Arm auto-merge (pull_request) Successful in 8s
7c1b637694
`ping 10.0.0.1` on the router that configured 10.0.0.1 lost all five echoes:

    % Echoes lost: 5 drops (no ARP reply from 10.0.0.1 x5)

The engine routed the request out of the port that owns the address and waited for an
ARP reply to its own address, which no NIC sends, so no device in the lab could reach
itself - a PC could not ping its own address either, and no box could ping the address
it answers BGP with.

A packet whose destination exactly matches one of the box's own addresses is delivered
locally, not transmitted (RFC 1812 §5.2.3), and the echo server must answer what is sent
to the router (RFC 1812 §4.3.3.6, RFC 1122 §3.2.2.6). `_sendIcmp()` now asks
`Device.holdsIp()` before it looks for a route and answers from the stack; `_sendIcmp6()`
does the same through `Device.interfaceWithIp6()`, covering a configured IPv6 address and
the EUI-64 link-local each port forms. The round trip comes from the run's own event list,
which holds no frame for this exchange, so it prints 0.000 ms - the echo crossed no link.

Unchanged: a neighbour's address still goes the wire and still loses with `no ARP reply`,
and a host still reaches the router across the cable. `holdsIp()` names only an enabled
port, so an address on a disabled one still fails the routing way.

research/ping-own-address.md has the citations; test/ping-own-address.test.js holds the
local echo (IPv4 and IPv6, console and engine API), the wire control and the one-hop trace.
jaylouisw force-pushed fix/1117-2006 from 7c1b637694
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 24s
CI / Server smoke test (pull_request) Successful in 11s
CI / Test (Node 18) (pull_request) Successful in 49s
CI / Test (Node 20) (pull_request) Successful in 47s
CI / Docker build (pull_request) Successful in 20s
CI / Test (Node 22) (pull_request) Successful in 48s
CI / Arm auto-merge (pull_request) Successful in 8s
to c65063efb6
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 8s
CI / Docker build (pull_request) Successful in 24s
CI / Test (Node 18) (pull_request) Successful in 58s
CI / Test (Node 20) (pull_request) Successful in 54s
CI / Test (Node 22) (pull_request) Successful in 54s
CI / Arm auto-merge (pull_request) Successful in 18s
2026-11-17 20:41:28 +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!1894
No description provided.