@torrent-tv/proxy 2.87.21 → 2.88.0

package/services/encode/quality/drain-threat.js

@@ -1,47 +1,47 @@
-/**
- * @file Whether a viewer's buffer will run dry before another output could have
- * the piece they need (roadmap item 98).
- *
- * THE RULE, stated by the user 2026-09-27: a move to a smaller output is
- * prepared when, on the buffer's present trend, the viewer's buffer ends
- * sooner than the piece they need would be closed on the new output. A buffer
- * that falls without that threat moves nothing — a trend down is ordinary
- * between two segments and while a cushion is being spent.
- *
- * WHEN IT IS JUDGED. On each report of the viewer, and the next judgement comes
- * with their next report. So the question asked is whether the buffer lasts
- * until the new piece would be ready IF THE MOVE WAITED FOR THAT NEXT REPORT:
- *
- *     secondsToEmpty <= secondsToReady + reportGapSec
- *
- * where `secondsToEmpty` is what they hold divided by how fast it drains, and
- * `reportGapSec` is the interval between their last two reports. Every term is
- * measured; nothing here is chosen.
- *
- * A time to ready that is not known is not assumed to be short: with nothing to
- * show there is time left, a draining buffer is a threat.
- */
-
-/**
- * @param {object} params
- * @param {number} params.bufferedSec - What the viewer holds now.
- * @param {number | null} params.slope - The buffer's trend, seconds of film per
- *   second; negative while it drains. Null when not yet known.
- * @param {number} params.reportGapSec - The interval between the viewer's last
- *   two reports.
- * @param {number | null} params.secondsToReady - How long the new output would
- *   take to close the piece the viewer needs; null when not known.
- * @returns {{ threat: boolean, secondsToEmpty: number | null }}
- */
-export function drainThreat({ bufferedSec, slope, reportGapSec, secondsToReady }) {
-  if (!Number.isFinite(slope) || slope >= 0) {
-    return { threat: false, secondsToEmpty: null };
-  }
-  const held = Number.isFinite(bufferedSec) && bufferedSec > 0 ? bufferedSec : 0;
-  const secondsToEmpty = held / -slope;
-  if (!Number.isFinite(secondsToReady) || secondsToReady < 0) {
-    return { threat: true, secondsToEmpty };
-  }
-  const gap = Number.isFinite(reportGapSec) && reportGapSec > 0 ? reportGapSec : 0;
-  return { threat: secondsToEmpty <= secondsToReady + gap, secondsToEmpty };
-}
+/**
+ * @file Whether a viewer's buffer will run dry before another output could have
+ * the piece they need (roadmap item 98).
+ *
+ * THE RULE, stated by the user 2026-09-27: a move to a smaller output is
+ * prepared when, on the buffer's present trend, the viewer's buffer ends
+ * sooner than the piece they need would be closed on the new output. A buffer
+ * that falls without that threat moves nothing — a trend down is ordinary
+ * between two segments and while a cushion is being spent.
+ *
+ * WHEN IT IS JUDGED. On each report of the viewer, and the next judgement comes
+ * with their next report. So the question asked is whether the buffer lasts
+ * until the new piece would be ready IF THE MOVE WAITED FOR THAT NEXT REPORT:
+ *
+ *     secondsToEmpty <= secondsToReady + reportGapSec
+ *
+ * where `secondsToEmpty` is what they hold divided by how fast it drains, and
+ * `reportGapSec` is the interval between their last two reports. Every term is
+ * measured; nothing here is chosen.
+ *
+ * A time to ready that is not known is not assumed to be short: with nothing to
+ * show there is time left, a draining buffer is a threat.
+ */
+
+/**
+ * @param {object} params
+ * @param {number} params.bufferedSec - What the viewer holds now.
+ * @param {number | null} params.slope - The buffer's trend, seconds of film per
+ *   second; negative while it drains. Null when not yet known.
+ * @param {number} params.reportGapSec - The interval between the viewer's last
+ *   two reports.
+ * @param {number | null} params.secondsToReady - How long the new output would
+ *   take to close the piece the viewer needs; null when not known.
+ * @returns {{ threat: boolean, secondsToEmpty: number | null }}
+ */
+export function drainThreat({ bufferedSec, slope, reportGapSec, secondsToReady }) {
+  if (!Number.isFinite(slope) || slope >= 0) {
+    return { threat: false, secondsToEmpty: null };
+  }
+  const held = Number.isFinite(bufferedSec) && bufferedSec > 0 ? bufferedSec : 0;
+  const secondsToEmpty = held / -slope;
+  if (!Number.isFinite(secondsToReady) || secondsToReady < 0) {
+    return { threat: true, secondsToEmpty };
+  }
+  const gap = Number.isFinite(reportGapSec) && reportGapSec > 0 ? reportGapSec : 0;
+  return { threat: secondsToEmpty <= secondsToReady + gap, secondsToEmpty };
+}

package/services/encode/quality/link-budget.js

@@ -1,254 +1,254 @@
-/**
- * @file What a viewer's link can carry, and what an output asks of it.
- *
- * PURE FUNCTIONS: the viewer's connection enters as a NUMBER — how many megabits
- * the last report showed — and never as a reference to a viewer. Nothing here
- * holds anybody, reads anything or can be asked at the wrong moment.
- *
- * WHAT A LINK CARRIES IS A LOAD, NOT A PICTURE (roadmap item 97, step 11). The
- * viewer receives the picture AND the soundtrack they chose — inside the
- * picture's own stream, or as a separate one — so both are added before the
- * link is asked. And each part says how much its figure can be trusted:
- *
- * 1. `known` — a bound the encoder is held to: a re-encode's `maxrate`, or the
- *    constant rate a re-encoded soundtrack is produced at;
- * 2. `estimated` — a figure the FILE states: a copied picture's average rate,
- *    a copied soundtrack's stated rate. An average is not a bound, so a load
- *    built on it is admitted only by estimate and is never confirmed;
- * 3. `unknown` — no figure at all: a hardware encoder given no limit, a copied
- *    soundtrack whose rate nothing states.
- *
- * A load is as trustworthy as its least trustworthy part.
- */
-
-import { maxrateKbpsFor, nominalKbpsFor } from "../args.js";
-
-/**
- * How much of a measured link may be spent on the stream.
- *
- * The rest is what the link does when it is not being perfect: transport
- * framing, retransmission, the other tabs, the moment somebody else in the
- * house starts something. A load sized to the whole reading stalls on the
- * first of those.
- *
- * Chosen, not measured, and written as a constant rather than dressed up as a
- * measurement. What would replace it is a reading of the transport's own
- * overhead and of how far a link's throughput varies over a session, which
- * nothing takes.
- */
-export const LINK_SAFETY = 0.8;
-
-/** How far a figure in a load can be trusted. */
-export const PEAK_CLASS = Object.freeze({
-  KNOWN: "known",
-  ESTIMATED: "estimated",
-  UNKNOWN: "unknown"
-});
-
-/** What a link was found to do with a load. */
-export const LINK_VERDICT = Object.freeze({
-  FITS: "fits",
-  ESTIMATED_TO_FIT: "estimated to fit",
-  DOES_NOT_FIT: "does not fit",
-  NO_SAFE_BOUND: "no safe bound",
-  NO_MEASUREMENT: "no measurement"
-});
-
-const CLASS_ORDER = [PEAK_CLASS.KNOWN, PEAK_CLASS.ESTIMATED, PEAK_CLASS.UNKNOWN];
-
-/**
- * @typedef {{ mbps: number | null, peakClass: string }} LoadPart
- * @typedef {{ video: LoadPart | null, audio: LoadPart | null, totalMbps: number | null, peakClass: string }} Load
- */
-
-/**
- * @param {number | null} mbps
- * @param {string} peakClass
- * @returns {LoadPart}
- */
-function part(mbps, peakClass) {
-  const figure = Number.isFinite(mbps) && mbps > 0 ? mbps : null;
-  // A figure that is missing is no figure, whatever the part claimed to be.
-  return { mbps: figure, peakClass: figure === null ? PEAK_CLASS.UNKNOWN : peakClass };
-}
-
-/**
- * The picture part of a load, read from an output's identity.
- *
- * An encoder given no limit states no bound, and its part is `unknown` — unless
- * this host has seen outputs of exactly this mode carry a peak (roadmap item
- * 97, step 14), and then that peak is its figure, as an ESTIMATE: it is what
- * was seen, not a bound anything holds the encoder to.
- *
- * @param {import("../output/OutputSpec.js").OutputSpec} spec
- * @param {number | null} sourceMbps - The average rate the source file states.
- * @param {number | null} [observedPeakMbps] - The largest peak outputs of this
- *   mode were seen carrying on this configuration, or null.
- * @returns {LoadPart | null} Null for an output that carries no picture.
- */
-export function videoLoadOfSpec(spec, sourceMbps, observedPeakMbps = null) {
-  if (!spec?.video) {
-    return null;
-  }
-  const encode = spec.video.encode;
-  if (!encode) {
-    return part(sourceMbps, PEAK_CLASS.ESTIMATED);
-  }
-  const rateControl = encode.rateControl;
-  if (rateControl) {
-    return part(rateControl.maxrateKbps / 1000, PEAK_CLASS.KNOWN);
-  }
-  return Number.isFinite(observedPeakMbps) && observedPeakMbps > 0
-    ? part(observedPeakMbps, PEAK_CLASS.ESTIMATED)
-    : part(null, PEAK_CLASS.UNKNOWN);
-}
-
-/**
- * The picture part of a load for a height nothing has produced yet, the way
- * it WOULD be produced here: at the nominal limit on the software encoder,
- * with no limit on a hardware one, and as the source's own stream when the
- * picture is copied at the source's height.
- *
- * The nominal limit is the one of the FRAME the height would be encoded at,
- * chosen by its area (`nominalKbpsFor`) — the same row the output is then
- * opened in, so the offer and the output cannot price one step differently.
- *
- * @param {{ sourceHeight: number, copiesAtSource: boolean, sourceMbps: number | null, encoderKind: string }} picture
- * @param {{ width: number, height: number }} frame - The frame this height is
- *   encoded at for this source.
- * @returns {LoadPart}
- */
-export function videoLoadForFrame(picture, frame) {
-  const sourceHeight = Math.round(Number(picture?.sourceHeight) || 0);
-  if (picture?.copiesAtSource === true && frame.height === sourceHeight) {
-    return part(picture?.sourceMbps ?? null, PEAK_CLASS.ESTIMATED);
-  }
-  if (picture?.encoderKind !== "software") {
-    return part(null, PEAK_CLASS.UNKNOWN);
-  }
-  return part(maxrateKbpsFor(nominalKbpsFor(frame)) / 1000, PEAK_CLASS.KNOWN);
-}
-
-/**
- * The picture part of a load for a software encode at a nominal limit nothing
- * has produced yet. The bound is the one `softwareRateControlFor` would give
- * it, so an output opened at that limit carries exactly this load.
- *
- * @param {number} nominalKbps
- * @returns {LoadPart}
- */
-export function videoLoadOfLimit(nominalKbps) {
-  return part(maxrateKbpsFor(nominalKbps) / 1000, PEAK_CLASS.KNOWN);
-}
-
-/**
- * The soundtrack part of a load: the track THIS viewer chose.
- *
- * @param {{ transcode: boolean, bitrateKbps: number | null } | null} track -
- *   Null when the viewer receives no sound from this output or beside it.
- * @param {number} transcodeKbps - What a re-encoded track is produced at.
- * @returns {LoadPart | null}
- */
-export function audioLoadOf(track, transcodeKbps) {
-  if (!track) {
-    return null;
-  }
-  if (track.transcode === true) {
-    return part(transcodeKbps / 1000, PEAK_CLASS.KNOWN);
-  }
-  return part((Number(track.bitrateKbps) || 0) / 1000, PEAK_CLASS.ESTIMATED);
-}
-
-/**
- * The whole load: the parts added, and the class of the least trustworthy.
- *
- * @param {LoadPart | null} video
- * @param {LoadPart | null} audio
- * @returns {Load}
- */
-export function loadOf(video, audio) {
-  const parts = [video, audio].filter(Boolean);
-  const peakClass = parts.reduce(
-    (worst, one) => (CLASS_ORDER.indexOf(one.peakClass) > CLASS_ORDER.indexOf(worst) ? one.peakClass : worst),
-    PEAK_CLASS.KNOWN
-  );
-  const totalMbps = peakClass === PEAK_CLASS.UNKNOWN
-    ? null
-    : parts.reduce((sum, one) => sum + one.mbps, 0);
-  return { video, audio, totalMbps, peakClass };
-}
-
-/**
- * What a link that measured `linkMbps` does with a load.
- *
- * 1. nothing measured the link — every class is admitted, since the link gave
- *    no ground to refuse, and nothing is confirmed;
- * 2. an `unknown` load against a measured link — `no safe bound`, refused:
- *    there is no number to compare;
- * 3. otherwise the load is compared with the usable share of the link: a
- *    `known` load that passes FITS and is confirmed, an `estimated` one that
- *    passes is admitted by estimate and is NOT confirmed.
- *
- * @param {number | null} linkMbps
- * @param {Load} load
- * @returns {{ verdict: string, admitted: boolean, confirmed: boolean, linkMbps: number | null, load: Load }}
- */
-export function linkCouldCarry(linkMbps, load) {
-  const measured = Number.isFinite(linkMbps) && linkMbps > 0;
-  const answer = (verdict, admitted, confirmed) => ({
-    verdict,
-    admitted,
-    confirmed,
-    linkMbps: measured ? linkMbps : null,
-    load
-  });
-  if (!measured) {
-    return answer(LINK_VERDICT.NO_MEASUREMENT, true, false);
-  }
-  if (load.peakClass === PEAK_CLASS.UNKNOWN) {
-    return answer(LINK_VERDICT.NO_SAFE_BOUND, false, false);
-  }
-  if (linkMbps * LINK_SAFETY < load.totalMbps) {
-    return answer(LINK_VERDICT.DOES_NOT_FIT, false, false);
-  }
-  return load.peakClass === PEAK_CLASS.KNOWN
-    ? answer(LINK_VERDICT.FITS, true, true)
-    : answer(LINK_VERDICT.ESTIMATED_TO_FIT, true, false);
-}
-
-/**
- * How strongly an answer is admitted, for ordering: a load whose figure is a
- * bound before one whose figure is an average, and both before one with no
- * figure (which is only ever admitted with nothing measured). Zero for a
- * refusal.
- *
- * @param {{ admitted: boolean, load: Load }} answer
- * @returns {number}
- */
-export function admissionRank(answer) {
-  if (!answer.admitted) {
-    return 0;
-  }
-  return 3 - CLASS_ORDER.indexOf(answer.load.peakClass);
-}
-
-/**
- * The figures behind an answer, as plain values a log line or a report can
- * carry.
- *
- * @param {{ verdict: string, linkMbps: number | null, load: Load }} answer
- * @returns {object}
- */
-export function linkAnswerFigures(answer) {
-  return {
-    verdict: answer.verdict,
-    linkMbps: answer.linkMbps,
-    linkSafety: LINK_SAFETY,
-    videoMbps: answer.load.video?.mbps ?? null,
-    videoClass: answer.load.video?.peakClass ?? null,
-    audioMbps: answer.load.audio?.mbps ?? null,
-    audioClass: answer.load.audio?.peakClass ?? null,
-    totalMbps: answer.load.totalMbps,
-    peakClass: answer.load.peakClass
-  };
-}
+/**
+ * @file What a viewer's link can carry, and what an output asks of it.
+ *
+ * PURE FUNCTIONS: the viewer's connection enters as a NUMBER — how many megabits
+ * the last report showed — and never as a reference to a viewer. Nothing here
+ * holds anybody, reads anything or can be asked at the wrong moment.
+ *
+ * WHAT A LINK CARRIES IS A LOAD, NOT A PICTURE (roadmap item 97, step 11). The
+ * viewer receives the picture AND the soundtrack they chose — inside the
+ * picture's own stream, or as a separate one — so both are added before the
+ * link is asked. And each part says how much its figure can be trusted:
+ *
+ * 1. `known` — a bound the encoder is held to: a re-encode's `maxrate`, or the
+ *    constant rate a re-encoded soundtrack is produced at;
+ * 2. `estimated` — a figure the FILE states: a copied picture's average rate,
+ *    a copied soundtrack's stated rate. An average is not a bound, so a load
+ *    built on it is admitted only by estimate and is never confirmed;
+ * 3. `unknown` — no figure at all: a hardware encoder given no limit, a copied
+ *    soundtrack whose rate nothing states.
+ *
+ * A load is as trustworthy as its least trustworthy part.
+ */
+
+import { maxrateKbpsFor, nominalKbpsFor } from "../args.js";
+
+/**
+ * How much of a measured link may be spent on the stream.
+ *
+ * The rest is what the link does when it is not being perfect: transport
+ * framing, retransmission, the other tabs, the moment somebody else in the
+ * house starts something. A load sized to the whole reading stalls on the
+ * first of those.
+ *
+ * Chosen, not measured, and written as a constant rather than dressed up as a
+ * measurement. What would replace it is a reading of the transport's own
+ * overhead and of how far a link's throughput varies over a session, which
+ * nothing takes.
+ */
+export const LINK_SAFETY = 0.8;
+
+/** How far a figure in a load can be trusted. */
+export const PEAK_CLASS = Object.freeze({
+  KNOWN: "known",
+  ESTIMATED: "estimated",
+  UNKNOWN: "unknown"
+});
+
+/** What a link was found to do with a load. */
+export const LINK_VERDICT = Object.freeze({
+  FITS: "fits",
+  ESTIMATED_TO_FIT: "estimated to fit",
+  DOES_NOT_FIT: "does not fit",
+  NO_SAFE_BOUND: "no safe bound",
+  NO_MEASUREMENT: "no measurement"
+});
+
+const CLASS_ORDER = [PEAK_CLASS.KNOWN, PEAK_CLASS.ESTIMATED, PEAK_CLASS.UNKNOWN];
+
+/**
+ * @typedef {{ mbps: number | null, peakClass: string }} LoadPart
+ * @typedef {{ video: LoadPart | null, audio: LoadPart | null, totalMbps: number | null, peakClass: string }} Load
+ */
+
+/**
+ * @param {number | null} mbps
+ * @param {string} peakClass
+ * @returns {LoadPart}
+ */
+function part(mbps, peakClass) {
+  const figure = Number.isFinite(mbps) && mbps > 0 ? mbps : null;
+  // A figure that is missing is no figure, whatever the part claimed to be.
+  return { mbps: figure, peakClass: figure === null ? PEAK_CLASS.UNKNOWN : peakClass };
+}
+
+/**
+ * The picture part of a load, read from an output's identity.
+ *
+ * An encoder given no limit states no bound, and its part is `unknown` — unless
+ * this host has seen outputs of exactly this mode carry a peak (roadmap item
+ * 97, step 14), and then that peak is its figure, as an ESTIMATE: it is what
+ * was seen, not a bound anything holds the encoder to.
+ *
+ * @param {import("../output/OutputSpec.js").OutputSpec} spec
+ * @param {number | null} sourceMbps - The average rate the source file states.
+ * @param {number | null} [observedPeakMbps] - The largest peak outputs of this
+ *   mode were seen carrying on this configuration, or null.
+ * @returns {LoadPart | null} Null for an output that carries no picture.
+ */
+export function videoLoadOfSpec(spec, sourceMbps, observedPeakMbps = null) {
+  if (!spec?.video) {
+    return null;
+  }
+  const encode = spec.video.encode;
+  if (!encode) {
+    return part(sourceMbps, PEAK_CLASS.ESTIMATED);
+  }
+  const rateControl = encode.rateControl;
+  if (rateControl) {
+    return part(rateControl.maxrateKbps / 1000, PEAK_CLASS.KNOWN);
+  }
+  return Number.isFinite(observedPeakMbps) && observedPeakMbps > 0
+    ? part(observedPeakMbps, PEAK_CLASS.ESTIMATED)
+    : part(null, PEAK_CLASS.UNKNOWN);
+}
+
+/**
+ * The picture part of a load for a height nothing has produced yet, the way
+ * it WOULD be produced here: at the nominal limit on the software encoder,
+ * with no limit on a hardware one, and as the source's own stream when the
+ * picture is copied at the source's height.
+ *
+ * The nominal limit is the one of the FRAME the height would be encoded at,
+ * chosen by its area (`nominalKbpsFor`) — the same row the output is then
+ * opened in, so the offer and the output cannot price one step differently.
+ *
+ * @param {{ sourceHeight: number, copiesAtSource: boolean, sourceMbps: number | null, encoderKind: string }} picture
+ * @param {{ width: number, height: number }} frame - The frame this height is
+ *   encoded at for this source.
+ * @returns {LoadPart}
+ */
+export function videoLoadForFrame(picture, frame) {
+  const sourceHeight = Math.round(Number(picture?.sourceHeight) || 0);
+  if (picture?.copiesAtSource === true && frame.height === sourceHeight) {
+    return part(picture?.sourceMbps ?? null, PEAK_CLASS.ESTIMATED);
+  }
+  if (picture?.encoderKind !== "software") {
+    return part(null, PEAK_CLASS.UNKNOWN);
+  }
+  return part(maxrateKbpsFor(nominalKbpsFor(frame)) / 1000, PEAK_CLASS.KNOWN);
+}
+
+/**
+ * The picture part of a load for a software encode at a nominal limit nothing
+ * has produced yet. The bound is the one `softwareRateControlFor` would give
+ * it, so an output opened at that limit carries exactly this load.
+ *
+ * @param {number} nominalKbps
+ * @returns {LoadPart}
+ */
+export function videoLoadOfLimit(nominalKbps) {
+  return part(maxrateKbpsFor(nominalKbps) / 1000, PEAK_CLASS.KNOWN);
+}
+
+/**
+ * The soundtrack part of a load: the track THIS viewer chose.
+ *
+ * @param {{ transcode: boolean, bitrateKbps: number | null } | null} track -
+ *   Null when the viewer receives no sound from this output or beside it.
+ * @param {number} transcodeKbps - What a re-encoded track is produced at.
+ * @returns {LoadPart | null}
+ */
+export function audioLoadOf(track, transcodeKbps) {
+  if (!track) {
+    return null;
+  }
+  if (track.transcode === true) {
+    return part(transcodeKbps / 1000, PEAK_CLASS.KNOWN);
+  }
+  return part((Number(track.bitrateKbps) || 0) / 1000, PEAK_CLASS.ESTIMATED);
+}
+
+/**
+ * The whole load: the parts added, and the class of the least trustworthy.
+ *
+ * @param {LoadPart | null} video
+ * @param {LoadPart | null} audio
+ * @returns {Load}
+ */
+export function loadOf(video, audio) {
+  const parts = [video, audio].filter(Boolean);
+  const peakClass = parts.reduce(
+    (worst, one) => (CLASS_ORDER.indexOf(one.peakClass) > CLASS_ORDER.indexOf(worst) ? one.peakClass : worst),
+    PEAK_CLASS.KNOWN
+  );
+  const totalMbps = peakClass === PEAK_CLASS.UNKNOWN
+    ? null
+    : parts.reduce((sum, one) => sum + one.mbps, 0);
+  return { video, audio, totalMbps, peakClass };
+}
+
+/**
+ * What a link that measured `linkMbps` does with a load.
+ *
+ * 1. nothing measured the link — every class is admitted, since the link gave
+ *    no ground to refuse, and nothing is confirmed;
+ * 2. an `unknown` load against a measured link — `no safe bound`, refused:
+ *    there is no number to compare;
+ * 3. otherwise the load is compared with the usable share of the link: a
+ *    `known` load that passes FITS and is confirmed, an `estimated` one that
+ *    passes is admitted by estimate and is NOT confirmed.
+ *
+ * @param {number | null} linkMbps
+ * @param {Load} load
+ * @returns {{ verdict: string, admitted: boolean, confirmed: boolean, linkMbps: number | null, load: Load }}
+ */
+export function linkCouldCarry(linkMbps, load) {
+  const measured = Number.isFinite(linkMbps) && linkMbps > 0;
+  const answer = (verdict, admitted, confirmed) => ({
+    verdict,
+    admitted,
+    confirmed,
+    linkMbps: measured ? linkMbps : null,
+    load
+  });
+  if (!measured) {
+    return answer(LINK_VERDICT.NO_MEASUREMENT, true, false);
+  }
+  if (load.peakClass === PEAK_CLASS.UNKNOWN) {
+    return answer(LINK_VERDICT.NO_SAFE_BOUND, false, false);
+  }
+  if (linkMbps * LINK_SAFETY < load.totalMbps) {
+    return answer(LINK_VERDICT.DOES_NOT_FIT, false, false);
+  }
+  return load.peakClass === PEAK_CLASS.KNOWN
+    ? answer(LINK_VERDICT.FITS, true, true)
+    : answer(LINK_VERDICT.ESTIMATED_TO_FIT, true, false);
+}
+
+/**
+ * How strongly an answer is admitted, for ordering: a load whose figure is a
+ * bound before one whose figure is an average, and both before one with no
+ * figure (which is only ever admitted with nothing measured). Zero for a
+ * refusal.
+ *
+ * @param {{ admitted: boolean, load: Load }} answer
+ * @returns {number}
+ */
+export function admissionRank(answer) {
+  if (!answer.admitted) {
+    return 0;
+  }
+  return 3 - CLASS_ORDER.indexOf(answer.load.peakClass);
+}
+
+/**
+ * The figures behind an answer, as plain values a log line or a report can
+ * carry.
+ *
+ * @param {{ verdict: string, linkMbps: number | null, load: Load }} answer
+ * @returns {object}
+ */
+export function linkAnswerFigures(answer) {
+  return {
+    verdict: answer.verdict,
+    linkMbps: answer.linkMbps,
+    linkSafety: LINK_SAFETY,
+    videoMbps: answer.load.video?.mbps ?? null,
+    videoClass: answer.load.video?.peakClass ?? null,
+    audioMbps: answer.load.audio?.mbps ?? null,
+    audioClass: answer.load.audio?.peakClass ?? null,
+    totalMbps: answer.load.totalMbps,
+    peakClass: answer.load.peakClass
+  };
+}