ECSS Large Packet Transfer — Downlink Resources (space-systems/ecss/e7041-resources)
Use when the task is the resource step of the large message downlink of
ECSS-E-ST-70-41C clause 6.13.3.2 — deciding how many simultaneous large
message transactions the subservice reserves capacity for, how many octets
each one holds while it is in progress, and whether a stated downlink demand
fits inside that declaration.
Domain quick reference
- A downlink transaction is not free while it runs. From the moment its first
part is generated until its last part leaves, it holds a transaction slot,
a transaction identifier and the source data behind the parts still to be
sent. The resource question is therefore about simultaneity, not about the
total volume sent over an orbit.
- Reserved octets are counted in whole parts. The subservice reads and sends
part-sized units, so a message that leaves a partly filled final part still
reserves that whole part, and a resource budget built from raw message
sizes is optimistic by up to one part per transaction.
- The peak is what sizes the subservice, and the peak is found by sweeping
the transaction timeline, not by averaging. Two long transfers that merely
touch at their ends never coexist; two short ones that overlap by a second
do, and that second is the sizing case.
- A transaction that ends exactly when another begins hands over its slot
cleanly. Treating that instant as an overlap inflates the slot count by one
on every back-to-back schedule and hides the real peak elsewhere.
- Admission is ordered. When the declared resources are short, it matters
which arrival is refused: the subservice admits in arrival order and the
refusal falls on the transaction that finds no free slot or no free buffer,
which is rarely the largest one.
Workflow
- Validate the declared resources: whole positive transaction slots, a whole
positive buffer pool in octets and a positive part size.
- Validate every demand entry: a positive message size and a transaction
window whose end is not before its start.
- Convert each message into its reserved footprint by rounding its size up
to a whole number of parts.
- Sweep the windows as an ordered event list, releasing a transaction before
admitting one that starts at the same instant, and record the peak
simultaneous transaction count and the peak reserved octets together with
the instants at which they occur.
- Replay the demand in arrival order against the declared slots and buffer
pool, marking each transaction admitted or refused and naming which of the
two resources was exhausted.
- Report the peaks, the admission outcome and every finding: a peak
transaction count above the declared slots, a peak reservation above the
declared pool, or a single message whose footprint no empty pool can hold.
Pitfalls
- Sizing the buffer pool from the largest message instead of from the peak of
overlapping messages. One large transfer is the easy case; three medium
ones in flight together is what exhausts the pool.
- Counting reserved octets from the message length. The final part is
reserved whole, so the footprint is the part-rounded size and the
difference is one part per concurrent transaction at the worst moment.
- Treating an end instant and a start instant at the same time as an overlap.
The slot is released first, and the inflated count sends the sizing to the
wrong transaction.
- Reporting only that the demand does not fit. The useful answer names which
resource ran out first, because adding slots and adding buffer octets are
different design changes with different costs.
- Assuming a refused transaction frees nothing. A refusal never reserves, so
the transactions behind it in arrival order are assessed against the
unchanged remaining resources rather than against a spent pool.
Behavior contract (gate 3)
The resource validation, part-rounded footprint, timeline sweep, peak
detection, ordered admission and the assembled resource assessment are
exercised by the gate 3 contract test: scripts/test_e7041_resources.py
against scripts/e7041_resources_logic.py (stdlib unittest, offline). Run:
python3 scripts/test_e7041_resources.py
Compliance
- ECSS standards are freely downloadable (ESA); cite the source and
paraphrase per standards-map.yaml.
- compliance: STANDARDS-REF, gated: false.
1---2name: e7041-resources3description: "Size the on-board resources a large packet transfer downlink subservice reserves under ECSS-E-ST-70-41C clause 6.13.3.2: turn each queued large message into the whole parts it buffers, sweep the transaction timeline for the peak of simultaneous transactions and peak reserved octets, then admit or refuse each arrival against the declared transaction slots and buffer pool. Use when the number of simultaneous large message downlinks, the octets reserved per transaction or the buffer pool of a service 13 subservice has to be justified or reviewed. Refuses an unordered transaction interval and a zero slot count. Trigger: ecss, e-st-70-41c, pus-service-13, large-message-downlink-resources, concurrent-transaction-slots, downlink-buffer-pool, transaction-slot-admission, peak-reserved-octets."4license: Apache-2.05compliance: STANDARDS-REF6standards:7 - id: ecss8 reference-only: true9gated: false10domain: space-systems11pack: space-systems12compatibility: "agentskills.io SKILL.md; any SKILL.md host (Claude Code, Hermes, OpenClaw)"13metadata:14 source: https://skillmd.com/skills/ashfordeou/e7041-resources15 domain: space-systems16 subdomain: ecss17 tags: [ecss, e-st-70-41c, pus-service-13, e7041-resources, large-message-downlink-resources, concurrent-transaction-slots, downlink-buffer-pool, transaction-slot-admission, peak-reserved-octets]18 version: 0.1.019 author: Aero Agent Skills20---2122# ECSS Large Packet Transfer — Downlink Resources (space-systems/ecss/e7041-resources)2324Use when the task is the resource step of the large message downlink of25ECSS-E-ST-70-41C clause 6.13.3.2 — deciding how many simultaneous large26message transactions the subservice reserves capacity for, how many octets27each one holds while it is in progress, and whether a stated downlink demand28fits inside that declaration.2930## Domain quick reference3132- A downlink transaction is not free while it runs. From the moment its first33 part is generated until its last part leaves, it holds a transaction slot,34 a transaction identifier and the source data behind the parts still to be35 sent. The resource question is therefore about simultaneity, not about the36 total volume sent over an orbit.37- Reserved octets are counted in whole parts. The subservice reads and sends38 part-sized units, so a message that leaves a partly filled final part still39 reserves that whole part, and a resource budget built from raw message40 sizes is optimistic by up to one part per transaction.41- The peak is what sizes the subservice, and the peak is found by sweeping42 the transaction timeline, not by averaging. Two long transfers that merely43 touch at their ends never coexist; two short ones that overlap by a second44 do, and that second is the sizing case.45- A transaction that ends exactly when another begins hands over its slot46 cleanly. Treating that instant as an overlap inflates the slot count by one47 on every back-to-back schedule and hides the real peak elsewhere.48- Admission is ordered. When the declared resources are short, it matters49 which arrival is refused: the subservice admits in arrival order and the50 refusal falls on the transaction that finds no free slot or no free buffer,51 which is rarely the largest one.5253## Workflow54551. Validate the declared resources: whole positive transaction slots, a whole56 positive buffer pool in octets and a positive part size.572. Validate every demand entry: a positive message size and a transaction58 window whose end is not before its start.593. Convert each message into its reserved footprint by rounding its size up60 to a whole number of parts.614. Sweep the windows as an ordered event list, releasing a transaction before62 admitting one that starts at the same instant, and record the peak63 simultaneous transaction count and the peak reserved octets together with64 the instants at which they occur.655. Replay the demand in arrival order against the declared slots and buffer66 pool, marking each transaction admitted or refused and naming which of the67 two resources was exhausted.686. Report the peaks, the admission outcome and every finding: a peak69 transaction count above the declared slots, a peak reservation above the70 declared pool, or a single message whose footprint no empty pool can hold.7172## Pitfalls7374- Sizing the buffer pool from the largest message instead of from the peak of75 overlapping messages. One large transfer is the easy case; three medium76 ones in flight together is what exhausts the pool.77- Counting reserved octets from the message length. The final part is78 reserved whole, so the footprint is the part-rounded size and the79 difference is one part per concurrent transaction at the worst moment.80- Treating an end instant and a start instant at the same time as an overlap.81 The slot is released first, and the inflated count sends the sizing to the82 wrong transaction.83- Reporting only that the demand does not fit. The useful answer names which84 resource ran out first, because adding slots and adding buffer octets are85 different design changes with different costs.86- Assuming a refused transaction frees nothing. A refusal never reserves, so87 the transactions behind it in arrival order are assessed against the88 unchanged remaining resources rather than against a spent pool.8990## Behavior contract (gate 3)9192The resource validation, part-rounded footprint, timeline sweep, peak93detection, ordered admission and the assembled resource assessment are94exercised by the gate 3 contract test: scripts/test_e7041_resources.py95against scripts/e7041_resources_logic.py (stdlib unittest, offline). Run:96python3 scripts/test_e7041_resources.py9798## Compliance99100- ECSS standards are freely downloadable (ESA); cite the source and101 paraphrase per standards-map.yaml.102- compliance: STANDARDS-REF, gated: false.