Digest · Topology and routing · read after Chapter 5
導讀 · 拓樸與路由 · 第 5 章後閱讀
Stable Hierarchical Routing for Operational LEO Networks (SHORT)面向營運中 LEO 網路的穩定階層式路由（SHORT）
This digest explains how Earth-fixed cells stabilize routing as satellites move. It then shows how SHORT’s hierarchy handles the remaining link, attachment and membership changes.讀完後，你能說明為什麼在鏈路、地面連接甚至成員都持續變動的星座中，以衛星身分路由可行性待驗證穩定，把路由命名空間錨定在地球固定的地理格子如何使星間座標距離隨時間維持固定，以及三層階層如何處理剩餘的變動。
Main result
主要結果
SHORT addresses geographic cells through a hierarchy of ground network, shells and orbits. The namespace absorbs predictable motion, leaving local rerouting to handle random disruption.
SHORT 透過地面網路、殼層與軌道的階層來定址地理格子，命名空間吸收可預測運動，隨機擾動則交由局部重繞處理。
| Item項目 | Values數值 | Conditions and source條件與來源 |
|---|---|---|
| Evidence and scope | Qualitative scope | SHORT starts from an observation about operations rather than about physics: inter-satellite links have limited large-scale use because in-space routing is unstable [69, slides 3–5] |
| Evidence and scope | Qualitative scope | Routing state faces three simultaneous changes: ground attachments, intra-shell and inter-shell ISLs, and random operational events. Earth and shell motion drives attachment changes. |
| Evidence and scope | Qualitative scope | and real disruptions from drag, manoeuvres, failures and partial deployment [69, slides 7–12]. |
| Dataset and workload | Qualitative scope | OLSR floods proactive updates. AODV and DSR flood reactive requests. Random maneuvers and failures limit orbit-prediction approaches [69, slides 15–17]. |
| Evidence and scope | Qualitative scope | SHORT addresses Earth-fixed geographic cells through a hierarchy of terrestrial networks, orbital shells and orbits [69, slides 21–22]. This replaces satellite identities in destination routing. |
| Orbital configuration | Qualitative scope | Orbit-aligned coordinates (α, γ) make satellite sub-points move linearly. Same-shell coordinate differences therefore stay constant, preserving routing entries during motion. Local rerouting absorbs ISL changes. Inter-shell routing serves cases where one shell covers only one endpoint [69, slides 29–35]. |
| Evidence and scope | 81 · 1,489 | The slides claim 81–1,489 times fewer routing updates. Their chart compares routing and switch-table updates for OLSR, TS-SDN, AODV, LBP and SHORT. Constellations are Starlink, Iridium and Telesat [69, slide 39]. |
| Latency comparison | Qualitative scope | The full paper awaited access for this primer. Thus, the method is read from the authors' slides and only that relative figure is used. In comparison, the absolute update rates, convergence times, latency values and constellation parameters are left to the paper. |
| 證據與範圍 | 定性範圍 | SHORT 從營運觀察出發：星間鏈路大規模使用仍受限制，是因為太空中的路由變動[69, slides 3–5]，而變動是因為三種動態同時衝擊路由狀態，一是地球與殼層異步運動造成的地面到衛星連接變動，二是幾何變化造成的殼層內與殼層間 ISL 變動，三是來自阻力、機動、故障與部分部署的真實干擾[69, slides 7–12]。 |
| 證據與範圍 | 定性範圍 | OLSR 這類平面式主動協定會洪泛路由更新，AODV 與 DSR 這類被動協定會洪泛路由請求，而軌道預測方案會失敗，因為機動與故障是隨機的[69, slides 15–17]。 |
| 證據與範圍 | 定性範圍 | SHORT 的解答是以錨定在地球上的地理格子命名，並置於地面網路、軌道殼層與軌道三層的階層中[69, slides 21–22]。 |
| 證據與範圍 | 定性範圍 | 由於格子是在與軌道對齊的座標 (α, γ) 中劃分，衛星的星下點在其中線性移動，同一殼層內兩顆衛星之間的座標距離因此隨時間維持固定，所以路由表在衛星移動時維持原有項目，殼層內的局部重繞吸收 ISL 變動，只有當一個殼層僅能覆蓋其中一個端點時才啟用殼層間路由[69, slides 29–35]。 |
| 證據與範圍 | 81 · 1,489 | 評估投影片以「路由更新減少 81 到 1,489 倍」為標題，旁邊是 OLSR、TS-SDN、AODV、LBP 與 SHORT 在 Starlink、Iridium 與 Telesat 上每秒路由與交換表項目更新數的圖[69, slide 39]。 |
| 證據與範圍 | 定性範圍 | 本教材取得仍待完成完整論文，所以方法是從作者投影片讀出，也只使用這個相對數字，絕對更新率、收斂時間、延遲數值與星座參數留給論文本身。 |
1. Where this paper sits1. 這篇論文的位置
Earlier work quantified path churn and operational maneuvers. SHORT uses Earth-fixed geography to reduce the control-plane work of following satellite identities.
早期研究量化了路徑變動與營運機動，SHORT 則以地球固定的地理位置，降低追蹤衛星身分所需的控制平面工作。
| Item項目 | Values數值 | Conditions and source條件與來源 |
|---|---|---|
| Evidence and scope | 5 | Chapter 5 tells the routing story in three steps. |
| Evidence and scope | 2018 · 10 ms · 200 s · 2024 · 2025 | Handley's 2018 simulator recomputes satellite-identity Dijkstra routes every 10 ms, with recomputation cost outside its model. Hypatia finds four median path changes per 200 s on Starlink and Kuiper. Work in 2024–2025 shifts routing identity toward geographic patches. |
| Orbital configuration | Qualitative scope | SHORT is the control-plane statement of that last step, and TinyLEO's geographic segment anycast is the same group's data-plane statement a year later. |
| Orbital configuration | 2023 | The overlapping-author MobiCom 2023 study finds operations rewriting neighbors tens of times daily. This supports SHORT's need for control beyond orbital prediction. |
| Evidence and scope | Qualitative scope | The design separates locator and identifier roles. Here, the geographic-cell locator stays fixed while the satellite identifier moves. |
| Evidence and scope | 54 · 29 · 2026 | Crossref counted 54 citing works on 29 October 2026, treated as a lower bound. |
| 證據與範圍 | 5 | 第 5 章用三個步驟講路由的故事。 |
| 證據與範圍 | 2018 · 10 ms | Handley 2018 年的模擬器每 10 ms 在衛星身分上重算一次 Dijkstra，並把結果視為免費。 |
| 證據與範圍 | 200 · 4 | Hypatia 接著量出這種變動的代價，在 Starlink 與 Kuiper 上中位數每 200 秒 4 次路徑變更。 |
| 證據與範圍 | 2024 · 2025 | 2024 到 2025 年的前沿則把路由的單位從衛星換成它剛好飛過的那片地面。 |
| 證據與範圍 | 定性範圍 | SHORT 是這最後一步在控制平面上的陳述，而 TinyLEO 的地理區段任播是同一團隊一年後在資料平面上的陳述。 |
| 證據與範圍 | 2023 | 這條脈絡也往回延伸：由重疊作者完成的 MobiCom 2023 Starlink 自駕星座研究確立了機動每天會改寫鄰居集合數十次，這是 SHORT 拒絕只依賴軌道預測的實證理由。 |
| 證據與範圍 | 定性範圍 | 熟悉地面網路的讀者會認出這一招是把定位符與辨識符分離倒過來用，因為在這裡穩定的是定位符（格子），移動的是辨識符（衛星）。 |
| 證據與範圍 | 2026 · 10 · 29 · 54 | Crossref 在 2026 年 10 月 29 日計得 54 篇引用，視為下限。 |
2. The problem, measured2. 問題，用量測說話
Attachment changes, geometric link churn and random operational events affect routing separately. A stable geographic namespace removes motion-driven updates while retaining local responses to disruption.
接入變更、幾何鏈路變動與隨機營運事件各自影響路由，穩定地理命名空間吸收運動帶來的更新，並保留局部擾動處理。
| Item項目 | Values數值 | Conditions and source條件與來源 |
|---|---|---|
| Evidence and scope | 2 · 4 · 2023 | The slides list three instability sources [69, slides 7–12]. Daily maneuver charts cover Starlink Shells 2–4 from January–July 2023 [69, slide 17]. Exact values remain in the original chart for source review. The fact base instead uses the MobiCom 2023 digest for operation counts. |
| Evidence and scope | Qualitative scope | The structure of the argument is what matters. |
| Orbital configuration | 25° · 550 km · 4.49 | Ground attachment follows predictable motion. At a 25° mask, a 550 km satellite remains visible for at most 4.49 minutes. Attachments therefore change hundreds of times daily. Satellite-identity destination entries can become stale within minutes. |
| Orbital configuration | Qualitative scope | ISL changes arise from latitude-dependent inter-plane spacing and drift between shells at separate altitudes. These produce the path changes measured by Hypatia. |
| Evidence and scope | Qualitative scope | Random changes include drag compensation, collision avoidance, hardware failures and partial deployment. These require observations beyond predicted ephemerides. |
| Reliability and accuracy | Qualitative scope | At megaconstellation scale, OLSR propagates every link-state change to every node. AODV and DSR flood requests for new destinations. Prediction-based schemes depend on satellites following published orbits [69, slides 15–17]. |
| Evidence and scope | Qualitative scope | The calc table converts two fact-base rates into churn events per day to make the first two sources concrete. |
| 證據與範圍 | 2 · 4 · 2023 · 1 · 7 | 投影片以三種變動來源[69, slides 7–12]以及一張 Starlink 第 2 至第 4 殼層 2023 年 1 月到 7 月每日機動次數的圖來鋪陳問題[69, slide 17]，該圖的數值只能從圖上讀取，因此這裡留在原圖供查閱，事實庫中關於營運的數字改由 MobiCom 2023 導讀提供。 |
| 證據與範圍 | 定性範圍 | 重要的是論證的結構。 |
| 軌道配置 | 25° · 550 km · 4.49 | 第一個來源是地面到衛星的變動，這遵循可預測的時程：地球上一個固定點在 25° 遮罩下最多只能看到一顆 550 km 衛星 4.49 分鐘，所以它的連接每天會改變數百次，而儲存目的地服務衛星的路由表在寫入後幾分鐘內就會出錯。 |
| 證據與範圍 | 定性範圍 | 第二個來源是殼層內與殼層間的 ISL 變動，因為軌道面間距隨緯度擺動、相異高度的殼層彼此漂移而過，這就是 Hypatia 以路徑變更量到的幾何變動。 |
| 證據與範圍 | 定性範圍 | 第三個來源是隨機的部分，阻力驅動的席位保持、碰撞閃避、硬體故障，以及殼層都處於部分部署狀態的事實，這些均需星曆預測之外的觀測。 |
| 證據與範圍 | 定性範圍 | 投影片接著說明為什麼三種經典回應在巨型星座規模下都失敗：OLSR 這類主動式鏈路狀態協定必須把每個變化傳播到每個節點，AODV 與 DSR 這類被動式協定必須為每個新目的地洪泛請求，而基於預測的方案只能好到星隊服從其公布軌道的程度[69, slides 15–17]。 |
| 證據與範圍 | 定性範圍 | 下方計算表把兩個事實庫中的速率換算成每天的變動事件數，使前兩個來源具體化。 |
| Question問題 | Calculation算式 | Result結果 |
|---|---|---|
| Serving-satellite changes per day for a fixed ground point, if each pass lasts the 25° maximum若每次過頂都達 25° 遮罩下的最長時間，固定地面點每天的服務衛星變更次數 | \(\dfrac{1{,}440\ \text{min}}{4.49\ \text{min}}\) | ≈ 321 per daycomputed. A lower bound on attachment churn, since real passes are shorter than the maximum (4.49 min from the Chapter 4 sheet)計算值。為連接變動的下限，因為真實過頂短於最大值（4.49 分取自第 4 章參數表） |
| Shortest-path changes per day for a typical pair at Hypatia's median rate以 Hypatia 的中位數速率計算，典型衛星對每天的最短路徑變更次數 | \(4 \times \dfrac{86{,}400}{200}\) | ≈ 1,728 per daycomputed. 4 changes per 200 s from [6, Fig. 8]計算值。每 200 秒 4 次取自 [6, Fig. 8] |
3. The design3. 設計
SHORT combines a geometric invariant with a routing hierarchy. Sub-points move linearly in orbit-aligned coordinates (α, γ). The authors therefore use these cells instead of latitude–longitude, Uber H3 or Google S2 cells. They describe those alternatives as satellite-oblivious, with complex user-to-satellite mapping [69, slides 27–30]. For two same-shell satellites, Δα and Δγ stay constant during motion. Forwarding rules based on these differences therefore remain fixed [69, slides 31–32].
The hierarchy has three tiers [69, slide 22]. The terrestrial tier addresses geographic cells. Orbital shells form separate routing domains with cell-coordinate forwarding. Individual orbit rings provide stable one-dimensional neighborhoods. A terrestrial node's IPv6 address includes an external-network prefix, shell identifier, cell column α and row γ.
A terrestrial node identifier that is unique within its cell. Thus, a packet's destination address already says which shell to use and which cell to steer toward [69, slide 38]. Dynamics are then handled where they arise. ISL churn inside a shell is absorbed by local rerouting between satellites that know their coordinate offset. Thus, a broken link is routed around with updates confined to the neighborhood [69, slide 33].
Inter-shell routing applies when one shell covers only one endpoint. The authors expect this rarely, mainly at high latitudes. Earth remains the anchor, with opportunistic cross-shell shortcuts [69, slides 34–36]. Terrestrial routing addresses a cell, and its current satellite delivers the last hop. Figure A shows the hierarchy and invariant.
SHORT 的設計建立在一個幾何事實與一個組織選擇上。幾何事實是衛星的星下點，也就是它正下方的地面點，若以與軌道對齊的座標 (α, γ) 表示，會隨時間線性變化，這就是作者在這組座標中劃分地理服務格子，。經緯度格子、Uber 的 H3 六角格與 Google 的 S2 空間填充曲線則有相異的對應特性的原因，他們認為後三者都讓衛星與地面用戶之間的對應獨立於衛星軌道且較複雜[69, slides 27–30]。若同一殼層內的兩顆衛星各以其目前所在的格子標記，它們座標之差 Δα 與 Δγ 便在兩者前進時維持固定，而以這些差寫成的轉送規則在僅有軌道運動時維持固定[69, slides 31–32]。組織選擇是三層階層[69, slide 22]。
最上層是地面網路，它只把目的地視為一個地理格子。中間層是各軌道殼層，每個殼層是獨立的路由域，其衛星依殼層內的格子座標轉送。最下層是各自軌道，軌道內的環提供穩定的一維鄰域。地面節點的 IPv6 位址把這個結構嵌入為四段：用於路由到外部網際網路的網路前綴、用於殼層間路由的軌道殼層辨識碼、用於殼層內路由的格子欄 α 與列 γ，以及在其格子內唯一的地面節點辨識碼，所以封包的目的位址本身就說明了要用哪個殼層、要朝哪個格子前進[69, slide 38]。動態於是在它們產生之處被處理。
殼層內的 ISL 變動由知道彼此座標偏移的衛星之間的局部重繞吸收，斷裂的鏈路被繞過，更新侷限於鄰域內[69, slide 33]。殼層間路由只在單一殼層僅能覆蓋其中一個端點時使用，作者預期這主要發生在高緯度的來源或目的地，並形容在現實中很少見，此時以地球為錨點，並利用殼層之間的機會性捷徑[69, slides 34–36]。地面變動被推到最上層，因為地面側定址的是地理格子，遞送當下飛過該格子的任何衛星都能完成最後一跳。圖 A 畫出三層與這個恆定量。
Original figure note原始圖說出處
Drawn for this primer from Li et al., MobiCom 2024, authors' slides 22, 31–33 and 38. The right panel is a schematic of the time-invariance argument of slide 32.4. Evaluation setup4. 實驗設定
The available slides compare routing-update work across protocols and constellations. Absolute rates, latency values and detailed network parameters require the full paper.
現有投影片比較各協定及星座的路由更新工作量，絕對速率、延遲數值與細部網路參數則需完整論文才能確認。
| Item項目 | Values數值 | Conditions and source條件與來源 |
|---|---|---|
| Evidence and scope | Qualitative scope | What can be said about the setup is limited to what the slides show. |
| Evidence and scope | Qualitative scope | Slides compare SHORT with proactive OLSR, predictive TS-SDN, reactive AODV and geographic LBP. Tests use Starlink, Iridium and Telesat [69, slide 39]. Fact-base Part B classifies the evaluation as emulation, with details in the full paper. |
| Latency comparison | 15 | The slides leave constellation parameters, ISL count and rate, latency, acquisition and 15-second scheduling unspecified. Fact-base Part D limits quantitative claims to verified relative results. Absolute update rates and convergence times require the full paper. |
| Latency comparison | Qualitative scope | Slides compare AODV, TS-SDN, LBP and SHORT latency under "Near optimal routing" [69, slide 39]. Under "Resilient to ISL failures", hops and delay rise with failed-link count [69, slide 40]. |
| Evidence and scope | Qualitative scope | but their values can only be read off the charts and are left in the source figures. |
| Orbital configuration | Qualitative scope | The metric counts routing and switch-table entry updates per second. It measures control-plane work needed to keep forwarding state consistent. This tests the effect of destination naming, with routing algorithms treated separately. |
| Evidence and scope | 0 | In that chart the bars for LBP and SHORT are both labelled 0 on a logarithmic axis that starts at one update per second |
| Evidence and scope | 81 · 1,489 | so the chart requires source details to show how the 81 to 1,489 range divides among the baselines |
| Evidence and scope | 2024 | Other quantities require confirmation in the full MobiCom 2024 paper. This includes which baseline and constellation determine each endpoint of the range. |
| 證據與範圍 | 定性範圍 | 關於實驗設定能說的，限於投影片所呈現的內容。 |
| 證據與範圍 | 定性範圍 | 評估投影片在 Starlink、Iridium 與 Telesat 三個星座上，把 SHORT 與 OLSR（標為平面主動式）、TS-SDN（平面預測式）、AODV（平面被動式）和 LBP（地理路由）比較[69, slide 39]，事實庫 Part B 把評估列為模擬平台、細節在完整論文中。 |
| 證據與範圍 | 15 | 星座參數、每顆衛星的 ISL 數、ISL 速率、延遲模型、雷射建立延遲，以及如何建模 Starlink 的 15 秒重組，在投影片中都待補充，而事實庫 Part D 明確禁止引用待查證的 SHORT 絕對更新率、收斂時間或星座參數，所以本導讀僅使用已核對的相對數值。 |
| 證據與範圍 | 定性範圍 | 投影片也以「Near optimal routing」為標題展示 AODV、TS-SDN、LBP 與 SHORT 隨時間變化的延遲比較[69, slide 39]，並以「Resilient to ISL failures」為標題畫出路徑跳數與延遲對失效 ISL 數量的關係，兩者都隨失效鏈路增加而上升[69, slide 40]，但這些數值只能從圖上讀取，這裡保留原圖供查閱。 |
| 證據與範圍 | 定性範圍 | 本導讀能陳述的是度量本身，即每秒的路由與交換表項目更新數，它衡量維持轉送狀態一致所需的控制平面工作量，對一篇主張「使 LEO 路由變動的是命名空間，演算法則另行考量」的論文來說，這是自然的量。 |
| 證據與範圍 | 0 · 81 · 1,489 · 2024 | 在該圖中，LBP 與 SHORT 的長條都標為 0，而對數座標軸從每秒一次更新起算，所以這張圖需要原文細節才能顯示 81 到 1,489 倍的範圍如何分配到各基準線，讀者應把其他所有數量，包括哪個基準線與星座各自落在範圍的哪一端，視為需要在 MobiCom 2024 論文本身確認的內容。 |
5. What was measured5. 量到了什麼
| Claim宣稱 | Setup and caveat設定與附註 |
|---|---|
| Routing updates fall by 81–1,489× [69, slide 39]. The figure also compares routing-table and switch-table updates per second. Baselines include OLSR, TS-SDN, AODV, LBP, and SHORT across Starlink, Iridium, and Telesat.路由更新減少 81 到 1,489 倍，旁邊是 OLSR、TS-SDN、AODV、LBP 與 SHORT 在 Starlink、Iridium 與 Telesat 上每秒路由與交換表項目更新數的圖[69, slide 39]。 | The slide’s Evaluation Highlights figure reports relative improvement. Absolute rates and per-baseline results require the full paper. LBP and SHORT bars both carry a zero label on a logarithmic axis starting at one update per second. OLSR, TS-SDN, and AODV bars differ by orders of magnitude across the constellations. Assigning the range endpoints therefore requires the full paper.投影片標題「Evaluation Highlights」。僅相對數字，絕對速率與按基準線的拆分留給完整論文。LBP 與 SHORT 的長條在從每秒一次更新起算的對數座標軸上都標為 0，而 OLSR、TS-SDN 與 AODV 的長條在三個星座之間相差數個數量級，所以範圍端點的對應仍需完整論文。 |
| Latency described as near-optimal, with a latency-over-time comparison against AODV, TS-SDN and LBP [69, slide 39].延遲被描述為接近最佳，並有與 AODV、TS-SDN 和 LBP 隨時間變化的延遲比較[69, slide 39]。 | Slide figure. Values left in the source figures, and the reference optimum is definition pending in the slides.投影片圖。數值留在原圖，參考最佳值的定義仍待補充。 |
| Resilience shown as path hop count and delay against the number of failed ISLs [69, slide 40].韌性以路徑跳數與延遲對失效 ISL 數量呈現[69, slide 40]。 | Slide figure. Paths remain connected as failures increase hop count and delay. The slides leave failed-link laser reacquisition cost outside the stated model.投影片圖。路徑保持連通，跳數與延遲則隨失效數增加而上升，但就投影片所見，失效鏈路的雷射重新擷取成本列於模型之外。 |
6. Limits and open questions6. 範圍與待解問題
- Only the slides were available. The MobiCom 2024 paper itself awaited access for this primer. Thus, the digest's account of the method is a reading of the authors' 42-page slide deck, cited by page. The only result it uses is the relative 81 to 1,489 figure. Absolute update rates, convergence times, latency values and constellation parameters are deliberately omitted as the fact base requires.只有投影片可用。本教材取得仍待完成 MobiCom 2024 論文本身，所以本導讀對方法的描述是對作者 42 頁投影片的解讀，並逐頁標註出處，唯一使用的結果是 81 到 1,489 這個相對值。絕對更新率、收斂時間、延遲數值與星座參數依事實庫要求刻意省略。
- The range spans four baselines and three constellations. OLSR floods link state, AODV floods route requests, and TS-SDN uses flat prediction. Geographic LBP and SHORT both have plotted rates labeled 0 [69, slide 39]. Update rates differ by orders of magnitude. The slides leave endpoint-to-scenario mapping unspecified.主要範圍很寬，因為基準線與星座的性質相異。OLSR 主動洪泛鏈路狀態，AODV 被動洪泛路由請求，TS-SDN 被標為平面預測式，而 LBP 是地理路由基準線，其圖中的更新率與 SHORT 一樣標為 0[69, slide 39]，所以這個範圍分布在四個基準線與三個星座之上，它們的更新率相差數個數量級，而各端點對應的基準線與星座仍待投影片之外的資料補充。
- ISL count, rate and setup cost require the full paper. Three-versus-four-terminal wiring and acquisition costs affect real recovery. Chapter 2 gives acquisition times of 2–30 s. The reported resilience therefore describes the routing plane.這篇論文的 ISL 數量、速率與建立延遲都待確認。模擬是每顆衛星接三條還是四條雷射、給它們多少容量、重繞後的鏈路如何要付出第 2 章的 2 到 30 秒擷取時間，都需取得完整論文才能確認，所以韌性結果應被理解為路由平面的特性，，真實終端恢復時間則需額外證據。
- The time-invariance argument holds within a shell. Two satellites in separate shells have separate angular rates, so their cell offset drifts, which is why SHORT confines routing to one shell wherever possible. How often real traffic needs the inter-shell path, and what that costs, is requires details beyond the slides.時間恆定性的論證只在殼層內成立。相異殼層的兩顆衛星角速率相異，其格子偏移會漂移，這就是 SHORT 盡可能把路由限制在單一殼層內的原因。真實流量多常需要殼層間路徑、代價多大，需投影片之外的細節才能確認。
- The 15-second scheduler and the user link are outside the model as far as can be told, so the stability claim covers the space segment. The last hop still changes satellite on Starlink's schedule, which the terrestrial tier hides from routing but leaves present in the user’s latency.就目前所知，15 秒排程器與用戶鏈路都在模型之外，所以穩定性的宣稱涵蓋的是太空段。最後一跳仍依 Starlink 的排程更換衛星，地面層把它藏在路由之外，卻使用者仍會承受這部分延遲。
- Cell granularity is a design parameter whose value and effect on stretch remain unspecified in the slides. Coarser cells make the invariant more robust to small manoeuvres but lengthen the last hop within a cell.格子粒度是一個設計參數，其數值與對延展比的影響仍需投影片之外的資料。較粗的格子讓恆定量對小幅機動更穩健，卻會拉長格子內的最後一跳。
7. How it connects7. 與其他內容的關聯
- Section 5.4 places SHORT's geographic hierarchy between Hypatia path changes and TinyLEO anycast. It provides the control-plane part of Earth-fixed routing.第 5.4 節：SHORT 是該節標題中的「地理階層」，置於 Hypatia 的路徑變動與 TinyLEO 的任播之間，作為地球固定路由的控制平面一半。
- Section 5.6: the random manoeuvres that section documents are the reason SHORT handles operational changes through local rerouting.第 5.6 節：該節記錄的隨機機動，正是 SHORT 以本地重繞處理營運變動的原因。
- Section 5.7: the assumptions table marks SHORT's ISL count, rate and setup-delay treatment as unspecified in the slides, pending the full paper.第 5.7 節：假設表把 SHORT 的 ISL 數、速率與建立延遲處理標記為待取得完整論文，有待完整論文。
- Section 4.2: the 4.49-minute maximum pass in the parameter sheet is the input to this digest's estimate of attachment churn.第 4.2 節：參數表中 4.49 分鐘的最長過頂是本導讀估算連接變動的輸入。
- TinyLEO digest: its SRv6 geographic segment anycast is the data plane that SHORT's cell addressing implies. MobiCom 2023 digest: the manoeuvre statistics that motivate both.TinyLEO 導讀：其 SRv6 地理區段任播就是 SHORT 格子定址所隱含的資料平面。MobiCom 2023 導讀：促成兩者的機動統計。
- Handley 2018 recomputes every 10 ms, while CoNEXT 2019 motifs uses minute-by-minute snapshots. Both route over satellite identities. SHORT measures the resulting control work. Ali & Zhang add daily changes retained even by stable-link topologies.Handley 2018 與 CoNEXT 2019 motif：兩者都在拓樸快照上以衛星身分計算最短路徑，Handley 每 10 ms 一次、CoNEXT 每分鐘一次，這正是 SHORT 的更新度量所要暴露的成本。Ali 與 Zhang 則顯示即使穩定拓樸仍保有的每日鏈路斷裂。
Key numbers關鍵數字
| Number | 數字 | Meaning (setup) | 意義（設定） |
|---|---|---|---|
| 81–1,489× | Fewer routing/switch-table entry updates per second in the evaluation against OLSR, TS-SDN, AODV and LBP on Starlink, Iridium, Telesat [69, slide 39]. The only SHORT result this primer uses | 在 Starlink、Iridium、Telesat 上與 OLSR、TS-SDN、AODV、LBP 比較時每秒路由/交換表項目更新的減少倍數[69, slide 39]。本教材使用的唯一 SHORT 結果 | |
| 3 tiers | Terrestrial network, orbital shells, orbits [69, slide 22] | 地面網路、軌道殼層、軌道[69, slide 22] | |
| 4 fields + prefix | IPv6 address of a ground node: network prefix, orbital-shell ID, column α, row γ, terrestrial node ID unique within its cell [69, slide 38] | 地面節點的 IPv6 位址：網路前綴、軌道殼層 ID、欄 α、列 γ、在格子內唯一的地面節點 ID[69, slide 38] | |
| 4.49 min | Maximum pass at 25° for 550 km, the input to the churn estimate (Chapter 4 sheet) | 550 km 在 25° 下的最長過頂，變動估算的輸入（第 4 章參數表） | |
| ≈ 321 / day | Lower bound on serving-satellite changes for a fixed point [computed] | 固定地面點服務衛星變更次數的下限 [計算值] | |
| ≈ 1,728 / day | Shortest-path changes at Hypatia's median 4 per 200 s [computed] | 以 Hypatia 中位數每 200 秒 4 次計算的最短路徑變更 [計算值] | |
| 54 | Crossref citing works on 29 Oct 2026, a lower bound | 2026 年 10 月 29 日 Crossref 的引用數，為下限 | |
| full-paper detail pending | ISLs per satellite, ISL rate, latency model, setup delay, 15-s schedule, absolute update rates, convergence times | 每衛星 ISL 數、ISL 速率、延遲模型、建立延遲、15 秒排程、絕對更新率、收斂時間 |