跳转到内容

20260922-0747-mate-package-document-structure Put the Mate package on the four-register model used by the Switch line

  • status: completed
  • createdAt: 2026-09-22 07:47
  • approvedAt: 2026-09-22 08:05
  • relatedTask: 20260922-0746-mate-package-document-structure

The Mate package is five pages, 983 lines: P03-mate.md (spec, 434), two design appendices (-design-hardware 282, -design-data-model 205), a brief (45) and the index. Mate is the product the Switch line's Air Hub, Studio, Eye, Tag, View and Control all cite, so its sections are referenced from 65 lines outside the package.

Register Where Problem
说明展示 P03-mate-brief.md stub: §2–§7 read "(待编写)"
功能描述 spec §3.4 (FR-01…22, hardware) and §4.4 (SFR-01…30, software) two lists, two chapter sequences, no acceptance column
规格书 spec §3 (16 sub-chapters) and §4 (11 sub-chapters) two Level 1 documents glued together; sub-chapter numbers restart, so "P03-mate.md §9" means the hardware OTA chapter to one reader and the software verification chapter to another; three change logs (§3.16, §4.11, §5)
开发需求 -design-hardware and -design-data-model hardware design restates the spec §3.7 chip table figure by figure; no guide (Mate is in-house, product plan §P3)
  1. Governance drift. product-plan.md §P3 decided on 2026-09-13 that Mate is one product, "不再作为'硬件 Hub 加运行其上的软件'两个半边对外表述". Spec §1 still opens with "硬件半边 … 软件半边" and a two-row table of separate tiers; the decision is not in the spec's change log (only 09-12 and 09-14 are).
  2. Ambiguous inbound references. Outside the package: P03-mate.md §10 (5 lines), §7 (2), §9, §6, §3, §5, plus [Hub §7], [Mate §6], [Mate] §5. Each was written against the pre-merge files P03-mate-hub.md / P03-mate-software.md; after the 09-14 merge the same "§n" exists twice. The design appendices' own relations lists still read [P03-mate-hub.md](P03-mate.md) and [P03-mate-software.md](P03-mate.md).
  3. CAN FD after the Classical CAN decision. Spec §3.2 SKU table and §3.14 EVT row say "2× CAN FD"; hardware design §2 block diagram, §4 pin budget ("CAN FD"), §5 ("YBO-CAN 默认速率 按 X01(500 kbit/s / 2 Mbit/s)") and §9 EOL ("500 kbit/s / 2 Mbit/s") still describe the FD bus. Spec §3.5 bus table already says Classical CAN 2.0B, 500 kbit/s, ≤8 bytes.
  4. Chip table in the spec. Spec §3.7 is a 14-row datasheet table (i.MX93, SOM, Murata 2EL, TJA1463, ISO1042, THVD1450, DP83867, LTE, GNSS, LM5143, LM74700, LTC3350, RTC). Hardware design §3 repeats the same figures rail by rail. By the Switch rule the table is Level 2.
  5. Level 1 / Level 2 disagreement on the third CAN. Spec §3.3 block diagram muxes J1939 onto FlexCAN1 via ISO1042 (N2K and J1939 mutually exclusive), while FR-14 requires republishing engine data on N2K concurrently. Hardware design §2/§5/§11 chose MCP2518FD as a third controller and notes "Level-1 框图经 PMA 更新" at M3. The spec does not list this as an open item.
  6. Stale names. "Mate 云 (P8)" (spec §3.3), "Core/Sense/Key(P01/P02/P06)" (spec §4.3), "Studio (P7)" style IDs; "OPEN-QUESTIONS P3-1…P3-8 / P3S-1…P3S-11" cite item IDs that open-questions.md §P3 no longer carries (it uses numbered lists under 硬件/软件).
  7. Layout. Titles use "—" and ASCII "(Level 2)"; spec relations list carries the non-standard label 合并记录; all three change logs are three-column (日期 | 变更 | 参考) instead of the standard four; the design appendices' updated (2026-08-28 / 08-31) predates their 09-14 relation-list rewrite.
  8. Sleep. Unlike Key and Air Hub, the Hub keeps a sleep state: X01 §9 (2026-09-11) removed sleep only for KEY and AIR-HUB. The spec's "休眠 ≤5 mA" row and the design's sleep budget are consistent with X01 and stay.
  9. Leftover from the Switch task. X01 §9 line 271 still cites P05-key.md FR-10; it should be KEY-05.
章节标题:“A. Recommended: one spec with one chapter sequence, chips to the design, brief written”

A1 — Spec skeleton (P03-mate.md, R0.3). Eleven chapters, hardware and software side by side inside each chapter, chapter numbers used once:

# Chapter Takes from today
1 定位与边界 §1, §3.1, §4.1, §4.2 — rewritten as one product with a hardware face and a software face (product plan 2026-09-13); tiers, revenue lines and compliance paths stay per face
2 型号、SKU 与目标规格 §3.2 SKU table; §3.5 electrical/I-O/bus tables; §3.8 mechanical; software tiers and schedule from §4.1; a "平台方向" row (i.MX93 SOM, EVT SOM / DVT custom carrier) replaces §3.7
3 功能与状态行为 3.1 硬件 FR-01…22 as a table (ID · 要求 · 验收 row of §9); 3.2 软件 SFR-01…30 grouped as today; 3.3 失效安全与控制面行为 (§3.10)
4 系统位置 §2, §3.3 tables and block diagram, §4.3
5 接口 §3.6 connectors, §4.5 数据、配置与接口
6 固件、配置与 OTA §3.9
7 安全、隐私与合规 §3.11, §4.6
8 成本与商业挂钩 §3.12, §4.7
9 参照基准与验证计划 §3.13, §4.8; §3.14, §4.9
10 待决事项与依赖 §3.15, §4.10, plus the third-CAN item (A4)
11 变更记录 §3.16, §4.11, §5 merged into one four-column table

Wording of every FR and SFR is unchanged; only the container changes. Verification is kept in the spec because there is no supplier guide.

A2 — Hardware design (P03-mate-design-hardware.md, R0.2). New §2 平台与选型 receives the spec §3.7 chip table with its M3/M8 freeze rule; existing §2–§11 become §3–§12. CAN FD wording in the diagram, pin budget, circuit list and EOL row is replaced by Classical CAN 500 kbit/s per X01 (the MCP2518FD stays: it is the J1939 controller and the J1939 port may run FD-capable silicon at Classical rates). Relations list, title dashes, four-column change log, updated.

A3 — Software design (P03-mate-design-data-model.md, R0.2). Relations list, title, change log format, updated; content unchanged. The name stays (it is a data model and API appendix); no rename, no redirect.

A4 — Contradictions. Governed ones are resolved, ungoverned ones marked 待决:

Item Governing record Treatment
"two halves" framing product plan §P3, 2026-09-13 spec §1 reworded; change-log row added
CAN FD / 2 Mbit/s in spec §2, §9 and hardware design X01 §1, 2026-09-11 Classical CAN 500 kbit/s; FD sentences deleted
Third CAN controller: L1 mux vs L2 MCP2518FD none (design §11 says M3) new open item in spec §10 stating both readings; L1 diagram keeps the mux with a note pointing at the item
Stale IDs (P8, P06, P7; OPEN-QUESTIONS P3-n / P3S-n) renames of 2026-09-11/12 P8 → Mate 云, P06 → Key, P7 → Studio; open-question citations rewritten as "OPEN-QUESTIONS §P3 硬件 n / 软件 n"

A5 — Brief. All eight sections written from the spec: one sentence, the problem, what it does (hardware face and software face), what it does not do (not a safety dependency, not a manned security service, not an MFD or NVR, no white label), position, SKUs and tiers, installation notes, status.

A6 — Inbound references. About 65 lines in X01, X02, Studio design, View design, Eye design, Control, Switch, research. Section citations are rewritten by a mapping table (old hardware §n / software §n → new chapter; design §n → §n+1), ID citations (FR-nn, SFR-nn) are unchanged. P05-key.md FR-10 in X01 §9 → KEY-05.

B. Alternative: split the spec into two Level 1 files again

章节标题:“B. Alternative: split the spec into two Level 1 files again”

P03-mate-hardware.md and P03-mate-software.md under one package. Removes the ambiguity too, but the lint allows one spec per package and the product plan says one product; it would need a decision record and would reverse the 09-14 merge.

Keep the two-half structure; fix CAN FD, stale names, layout, brief. Leaves the double chapter numbering and the ambiguous references.

  • Reference mapping. The "§n" citations outside the package are the largest manual step; a wrong mapping resolves to a real but wrong chapter and the build cannot catch it. Each is checked against the chapter it names.
  • Requirement tables. Turning the FR bullet list into a table with a verification column adds a mapping (FR → §9 test row) that does not exist today; where no test row covers an FR the cell is left "—" rather than invented.
  • Parallel session (task 20260921-1659) touches site tooling only; docs/sitemap.md is regenerated by both builds.
  • Rewritten: P03-mate.md, P03-mate-design-hardware.md, P03-mate-brief.md.
  • Updated: P03-mate-design-data-model.md (head and tail only), mate/index.md.
  • Inbound references: shared/X01, shared/X02, studio/P07-studio-design.md, view/P10-view-design.md, eye/P11-eye-design.md, control/P09-control.md, control/P09-control-brief.md, switch/P05-switch.md, switch/air-hub/P05-air-hub.md, research/engineering/vehicle-pdm-split-module-architecture.md.
  • docs/changelog.md; task and plan status. No governance file changes (the product plan already states the single-product decision). No renames, no _redirects, no decision record (no rule is deviated from).
  1. ID set before/after over mate/: every FR-nn, SFR-nn and open-item present once.
  2. Unit-bearing figures in the two design files checked against the spec; datasheet values and derived budgets are the only remainder.
  3. rg "CAN FD|2 Mbit" bizplan/product/mate returns only the J1939 controller capability and the TJA1463 datasheet note; rg "半边|P8\)|P06\)|P3S-|P3-[0-9]" bizplan/product/mate returns nothing.
  4. Relations lists and four-column change logs on all four controlled pages.
  5. Every rewritten inbound citation opened and matched to its chapter.
  6. bun run docs:lint && bun run docs:build && git diff --check.
  • Whether Mate should get a -guide slot now (supplier development requirements) although the product plan keeps the carrier board in-house; A assumes not.
  • 2026-09-22 08:05 owner: "开始处理". Option A approved; no -guide slot (Mate stays in-house per product plan §P3).
  • 2026-09-22 08:30: completed. Verification 1–6 passed: 52 FR/SFR IDs before and after; remaining Level 2 figures are datasheet values, rails, budgets and EOL limits; no "CAN FD", "半边", P8/P06/P3S citation left in mate/ except negations; four-column change logs and fixed-order relations lists on all three controlled pages; 32 inbound section citations repointed by the mapping (spec hardware §n / software §n → new chapter; design §n → §n+1); lint 222 pages 0 findings, build and git diff --check pass. Also done outside the package: product-plan.md §P3 "2× CAN FD" → Classical CAN, Studio/Tag "P8"/"P6" IDs, Eye "P03 §15" → §10, X01 P05-key.md FR-10 → KEY-05, and four hardware-FR citations in the software design corrected to SFR-. Not done: Sense, View, Core and Tag still carry "CAN FD" in their own rows and diagrams.