#Question-Relative Source Selection

F.1 · Stable · Architectural (A) · Normative unless explicitly marked informative · Part F - The Unification Suite (U-Suite): Concept Sets, SenseCells, and System-Role Kinds and Assignments

#About this pattern

This is a generated FPF pattern page projected from the published FPF source. It is canonical FPF content for this ID; it is not a FPF Reference product feature page.

#How to use this pattern

Read the ID, status, type, and normativity first. Use the content for exact wording, the relations for adjacent concepts, and citations to keep active work grounded without pasting the whole specification.

Type: Architectural (A) Status: Stable Normativity: Normative unless explicitly marked informative

One-sentence summary. Select the smallest inspectable set of exact sources whose claims, limits, rivals, or counterexamples can change one stated answer or use.

Historical title (informative). Domain-Family Landscape Survey. Aliases (informative). Source cut; question-relative source selection.

Use F.1 when the receiving question is clear, more than one source may change its answer, and the source cut is not yet justified. Start from candidate sources and exact editions, ask how each could change the answer or its limits, and retain the smallest set that covers the distinct contributions that matter.

#Keywords

• receiving question
• intended use
• exact source and edition
• answer-changing source role
• finite source cut
• SourceCutNote
• reopen condition.

#Relations

#Content

#Problem frame

Use F.1 when the receiving question is clear, more than one source may change its answer, and the source cut is not yet justified. Start from candidate sources and exact editions, ask how each could change the answer or its limits, and retain the smallest set that covers the distinct contributions that matter.

The first useful result is one SourceCutNote : U.Episteme. Before treating the note as that episteme, identify the receiving question as one exact project entity. That question is the note's EntityOfConcern; the intended use remains in the ClaimGraph rather than becoming a second subject. Name the effective U.ReferenceScheme explicitly: it must resolve the question, the cited source-edition designators, and the answer-changing role claims. The ClaimGraph also names deliberate exclusions, known limits or source gaps, and reopen conditions. If the question or scheme is still missing, keep an ordinary working note and return that missing value instead of claiming a C.2.1 episteme. A one-screen source note may represent the same episteme when it carries the same claims; its compact form neither admits a source nor contains its local meaning.

What this buys. Later term recovery, comparison, synthesis, or direct source use begins from an inspectable reason for each source rather than from one influential canon, a bibliography quota, a domain label, or an opaque search score.

Not this pattern when. If one already identified source closes the question, use its claim directly or use F.0.1 for one source-local meaning. Use F.9 for an actual relation between recovered local senses and F.0.2 for a bounded conceptual comparison or synthesis. Use G.2 for a broader refreshable SoTA pack. When the receiving work requires an exhaustive, protocol-defined, or appraisal-bearing evidence review, use its domain method; F.1 does not replace that search, appraisal, or reporting discipline.

Primary result. One question-relative SourceCutNote, not a source container, meaning container, admission token, or assurance result.

#Problem

Without a question-relative source cut:

1. Word drift. Familiar terms silently change meaning across sources and editions.
2. Canon lock. One influential standard is mistaken for the whole answer.
3. Recency blur. An old edition remains load-bearing merely because it was used first.
4. Category bleed. Designed structures, performed occurrences, system roles, statuses, permissions, and measurements are merged before their source claims are inspected.
5. Source bloat. A large reading list hides which sources can actually change the current answer.
6. Search substitution. Rank, similarity, or family membership is mistaken for relevance, truth, or completeness.

#Forces

#Solution — select by answer-changing role

#Minimal vocabulary

• Exact source and edition. The identified publication, canon, corpus, practice source, or other claim-bearing basis being considered.
• Receiving question and use. The independently identified question that the source-selection claims concern, plus the decision, pattern contribution, comparison, or action the answer will serve. The question is the SourceCutNote's EntityOfConcern; the use is claim content.
• Answer-changing role. The inspected claim, limit, rival explanation, counterexample, transfer condition, material non-fit, or other stated contribution by which a source can change the answer.
• Source cut. The finite set of sources retained for that question and use.
• SourceCutNote. One C.2.1 episteme identified by its source-selection ClaimGraph, the exact receiving question as EntityOfConcern, and the named effective ReferenceScheme under which that question, source editions, and role claims are read.
• Domain family. An informative discovery label, such as workflow, provenance, services, sensing, types, or control. It carries no semantics and cannot admit, merge, or replace a source.
• Local search policy. An optional named way to prioritize candidate inspection using search terms, descriptors, citations, embeddings, distances, active learning, or portfolio readings. Its output is attention guidance, not a source-selection verdict.

Source selection, source identity, source-local meaning, source truth, source adequacy, cross-source relation, synthesis, reliance, and assurance are separate questions.

#Source-selection method

Step 1 — State the receiving question and use. Say what answer is needed, what later use it serves, and which source difference could change that answer.

Step 2 — Name candidate sources by exact edition. Use identified sources whose relevant claims can be inspected. A discipline or domain-family label may help discovery but is not a source.

Step 3 — Inspect the answer-changing role of each candidate. Ask what claim or limit from that edition can change the answer. If that role depends on a disputed expression, pause selection only long enough to use F.0.1's ordinary branch: name the exact source, edition, and passage, and say in plain language what the expression means there. Then return to source selection. Do not run this branch for every candidate. Look deliberately for intended-use contributions, rival explanations, action-changing counterexamples, transfer limits, and material non-fit. One source may serve several roles.

Step 4 — Take the smallest sufficient cut. Retain sources that cover distinct answer-changing roles. Exclude a candidate when no inspected claim from it changes the stated question or use. Do not optimize for a fixed count, diversity score, canonical status, or exhaustive appearance.

Step 5 — Record exclusions, gaps, limits, and reopen conditions. Name deliberate exclusions and why they do not change this answer. State a load-bearing source gap rather than treating an unavailable source as irrelevant. Say which question, use, edition, rival, counterexample, or transfer-boundary change would reopen the cut.

Step 6 — Return one SourceCutNote. Identify it under C.2.1 by three values: the ClaimGraph produced by Steps 1–5, the exact receiving question from Step 1 as EntityOfConcern, and the named effective ReferenceScheme that resolves the question, cited editions, and role claims. Keep the intended use in the ClaimGraph. Use a one-screen representation when it helps the receiver hold the result in view. Detailed source analysis stays with the receiving comparison, synthesis, or evidence method. Step 7 — Recover only the meanings the work needs. After the cut is stable, create an F.17 local-sense cell only for a retained expression that later reuse, a claim, a named receiver, or an actual relation needs. Do not postpone a meaning needed by Step 3 to this stage, and do not manufacture a cell for every retained source.

#When the cut will support a SoTA claim

A source cut used to claim SoTA has a stricter role test because source relevance and source currentness do not establish the best-known answer. E.8:11 owns the FPF definition of SoTA and the meanings of its comparison roles; F.1 neither redefines SoTA nor selects the winning line. For each retained source, record one of those roles in plain wording: best-known-line candidate, serious current rival, failure or counterexample evidence, official or popular comparator, lineage only, or identity/currentness only.

The first three roles can supply answer-changing evidence. An official or popular comparator may stay only when its named defect is necessary to the comparison. Roles are assigned by contribution, not institution: an official standard or widely used practice may instead be a best-known-line candidate when its substantive answer wins against the serious alternatives. Officiality, prevalence, maintenance, citation, freshness, or academic praise contributes nothing to that rank. An official catalogue, publisher page, or registry entry may verify the source's identity, edition, publication date, or maintenance state; it establishes neither truth, adequacy, nor SoTA rank.

For a SoTA claim, disable the generic one-source cheap exit unless the one source is itself a current critical synthesis that compares the serious alternatives for the named question and the author can state why no known action-changing rival or counterexample remains hidden. Otherwise retain the necessary rival and failure evidence or return an unresolved source gap. Do not manufacture confidence from a one-source cut.

The SourceCutNote records the E.8:11 roles and the missing comparison, but it does not itself select the best-known line. Use F.0.2 when an actual cross-source synthesis claim is required. Use G.2 only when a broader refreshable evidence pack is justified; a bounded comparison does not require that apparatus by default.

#The SourceCutNote

The note's exact EntityOfConcern is the independently identified receiving question from Step 1. It is not the note, its file or one-screen form, the retained-source list, or a question-and-use bundle. The intended use remains part of what the note claims. Name the effective ReferenceScheme explicitly and make sure it resolves the question, every exact source-edition reference, and every answer-changing role statement. If either the question or scheme is unresolved, return that gap and keep the text as an ordinary working note.

Its ClaimGraph states:

• the receiving question and intended use;
• every retained source and exact edition;
• the answer-changing role of each retained source;
• for a SoTA-supporting cut, each retained source's plain role and any unresolved rival, counterexample, or synthesis gap;
• deliberate exclusions and their reasons;
• known limits and load-bearing source gaps;
• the distinction between designed and performed material when it affects the answer; and
• the conditions that reopen the cut.

A one-screen source note is a compact representation of the same episteme when it carries the same claims. If its ClaimGraph differs, it is another C.2.1 episteme rather than a second F.1 result kind. A short form does not establish source truth, adequacy, or local meaning.

#Optional search assistance

Use search aids only under a named local policy and only when they help find a suspected omission, rival, counterexample, or near-duplicate. When the result matters, keep the searched corpus, source editions, model or ranking method, scale or threshold, and intended interpretation recoverable.

Inspect the underlying source claims before changing the cut. A descriptor match, citation count, embedding distance, LLM answer, rank, threshold, active-learning choice, or portfolio score never admits, excludes, merges, or replaces a source by itself.

#Invariants

1. Every retained source has an inspected answer-changing role for the stated question and use.
2. The cut is finite, inspectable, and revisable; no universal count establishes sufficiency.
3. Exact sources and editions remain recoverable.
4. Source-local meanings remain local; F.1 does not merge or relate them.
5. The result contains no F.9 relation, synthesis conclusion, truth verdict, reliance decision, or assurance claim.
6. Domain families and search readings guide discovery only.
7. Designed descriptions and performed occurrences stay distinct when that source difference matters.
8. A changed relied premise reopens the affected cut claims; an unrelated edition-number change does not.
9. One already identified sufficient source is the ordinary cheap exit; for a SoTA claim, the stricter critical-synthesis condition in F.1:4.2.1 applies.
10. A protocol-defined evidence review continues to follow its domain method.

#Self-checks

• Answer-change test. What can each retained source change in the answer or action? If nothing, exclude it or state the missing role.
• Rival-and-limit test. Is a known rival explanation, counterexample, or transfer limit still hidden? Add the source that makes it inspectable or return the source gap.
• One-source test. Does one already identified source close the ordinary question? If yes, stop without additional F.1 steps. If the cut will support a SoTA claim, take this exit only when that source is a current critical synthesis of the serious alternatives and no known action-changing rival or counterexample remains hidden.
• SoTA-role test. Has each retained source been classified under one of the comparison roles defined in E.8:11? If not, classify it there before using the cut for a SoTA claim; do not recreate the role meanings in F.1.
• Rank test. Did official status, popularity, maintenance state, date, or a catalogue check raise a source's SoTA role? Remove that inference; keep the check only as source identity/currentness evidence.
• Gap test. If the best-known line cannot be established, does the note return the missing rival, counterexample, or synthesis need as a source gap rather than naming the newest available source?
• Locality test. Have two sources been treated as saying the same thing merely because they use the same word? If so, use the ordinary F.0.1 branch on the exact passages before deciding their roles, then return to selection.
• Search-policy test. Did a score decide membership before source claims were inspected? Undo that decision.
• Memory test. Can the receiver hold the cut and each source's role in view? Remove non-changing material or split genuinely different questions.
• Reopen test. Can the note say what future change would require selection again?

#Archetypal Grounding

#Three heterogeneous source cuts

#Role assignment, performed work, sensing, and execution

Candidate sources include BPMN 2.0 (designed workflow structures), PROV-O (performed activities and provenance), ITIL 4 (service vocabulary), ODRL 2.2 (permissions, prohibitions, and duties), SOSA/SSN (observations and results), and IEC 61131-3 (control-program execution). Retain only those whose exact claims change the receiving question.

The cut keeps false friends visible: a BPMN participant is not an RBAC role; a PROV activity is not a BPMN process; an SOSA observation is an act, not a status; an ITIL incident is not automatically a plant fault. F.1 exposes these source roles but does not settle their relations.

#Methods, types, and measurement

SPEM 2.0 or ISO 24744 may change the reading of Method and MethodDescription; OWL 2 and formal concept analysis may change kind reasoning; SOSA/SSN and ISO 80000-1 may change measurement and quantity claims. The cut preserves the source-fixed differences rather than making method, concept, or measurement globally uniform.

#Control, actuation, and services

Control-theory sources, IEC 61131-3, ISA-95, ITIL 4, and SOSA/SSN can contribute different claims about controller design, program execution, integration, service commitments, and observations. A current question may need only a subset. Actuation is not a service promise, and incident is not a plant fault merely because the words occur near operational work.

#Minimal worked source cut

The project brief already identifies this receiving question: Can one contribution use “process” for both a workflow description and a performed occurrence? The exact question, not the note or source list, is the SourceCutNote's EntityOfConcern. The project names its effective ReferenceScheme Workflow and occurrence source cut, August 2026; that scheme fixes the three editions below and the ordinary-English role statements. The intended use stays in the ClaimGraph.

Receiving use: decide which subjects the later pattern must keep distinct.

Retain:
- OMG BPMN 2.0.2 (January 2014), §10.1: its workflow-structure claim changes the design-description side.
- W3C PROV-O Recommendation (30 April 2013), §3.1: its Activity claim supplies the performed-occurrence contrast.
- W3C SOSA/SSN Recommendation (19 October 2017), §4.3.2.2: Observation supplies an action-changing counterexample—an act that follows a Procedure and yields a Result, not a workflow graph.

Exclude for now:
- thermodynamic-process literature: no inspected claim changes this stated use.

Known limit: service commitments are outside this cut.
Reopen if: the use adds physical transformation or service commitments, a relied edition changes the relevant claim, or a known counterexample no longer fits.
Search policy: none needed for this bounded question.

The resulting SourceCutNote is identified by that ClaimGraph, the stated receiving question, and the named reading scheme. Because this cut turns on the difference between a designed workflow and a performed occurrence, the project recovers any disputed process reading through ordinary F.0.1 while inspecting source roles, before stabilizing the cut. It adds an F.17 cell afterward only if later reuse, a claim, a named receiver, or an actual relation needs one.

#Portable first-hour case — function-to-module allocation

“First hour” means a compact entry slice, not a deadline or completeness claim.

Receiving question. Which current FPF sources must a Systems Engineering author use to prepare a first function-to-module allocation teaching slice without collapsing functions into modules? The project identifies this sentence as one exact question before treating the note as a C.2.1 episteme; that question is its EntityOfConcern.

Receiving use. Prepare a slice that helps a practitioner generate and compare candidate bearers, modules, allocations, interfaces, and trade-offs instead of copying functions into a component list or selecting familiar bearers first. This use stays in the note's ClaimGraph.

Effective ReferenceScheme. FPFCoreReferenceScheme, applied to the FPF August 2026 edition and the exact section locators below.

Retained sources and answer-changing roles. All five are exact passages in that edition.

1. A.6.F §§4.2 and 4.5. Separates the possible subjects hidden by function-like wording and requires function, bearer, flow, module allocation, and interface claims to remain distinct. This blocks the one-function-one-component shortcut.
2. A.6.M §§4 and 4.3. Treats module use as claim content over exact holons, allows many-to-many or still-unallocated functional claims, and requires an actual interface specification when compatibility or substitution is claimed. This changes what an allocation row must show.
3. C.30.TFS-REL §4.2. Relates functional structure to selected transformation-flow structure without identifying them. This exposes candidate flow topology, crossings, and correspondence limits that can change bearer placement.
4. C.31 §4.5. Makes function-module alignment, interface burden, and flow-boundary alignment separate characteristics rather than one modularity score. This changes the trade-offs the comparison must expose.
5. C.32 §§4–5. Starts candidate synthesis from functional demand and candidate bearers, keeps materially different configurations visible, and records expected gain, known loss, constraints, and source-return conditions. This prevents the source cut from pretending to choose the architecture.

Deliberate limits. The cut claims neither an exhaustive survey of allocation algorithms nor one module taxonomy or cross-sector optimum. It does not decide the final DPF pattern identity. The campaign-specific guide and research-source pilot remains in FPF-DPF-CLAIM-PLACEMENT-CAMPAIGN/PILOT-SYSTEMS-ENGINEERING-FUNCTION-TO-MODULE-ALLOCATION.md; it is not a hidden dependency of this portable example.

First result. One SourceCutNote whose ClaimGraph contains the five roles, use, limits, and reopen conditions; whose EntityOfConcern is the stated question; and whose effective scheme is FPFCoreReferenceScheme for FPF August 2026. The later comparison must expose unsupported capabilities, unallocated functions, unresolved interfaces, alternatives, trade-offs, and accepted losses.

Reopen when. A relied FPF passage changes, the question expands to external allocation algorithms or another domain, or the intended DPF use changes the required answer or boundary.

#Readable reasoning moves

• Retain: “This exact source stays because this inspected claim changes the answer in this way.”
• Exclude: “No inspected claim from this source changes the stated use; record the exclusion and what would reopen it.”
• Return a gap: “This unavailable or uninspectable source may be load-bearing; the cut remains limited here.”
• Keep meanings local: “Selection puts both sources in view; it does not say their terms are identical or related.”
• Use search assistance: “The ranking tells us what to inspect next; the source claim decides whether it enters the cut.”
• Reopen: “This question, use, relied claim, rival, counterexample, or transfer boundary changed, so select again.”

#Didactic distillation

State and independently identify the question; keep the later use in the claims. Keep a source because an inspected claim can change the answer, expose a rival or counterexample, or mark a transfer limit. Return one finite SourceCutNote whose ClaimGraph carries those roles, exclusions, limits, and reopen conditions, whose EntityOfConcern is the exact question, and whose named effective scheme resolves the question and exact editions. Search scores may guide inspection; they do not admit or exclude a source. Stop when additional candidates do not change the answer.

#Bias-Annotation

• Gov: Authority, popularity, international status, or sponsorship does not decide membership; the receiving question and inspected claims do.
• Arch: The method keeps source roles explicit and the active cut small without claiming exhaustive coverage.
• Onto/Epist: A source, edition, source claim, local meaning, source-cut episteme, and representation remain different.
• Prag: Direct use of one identified source is the cheap exit. Search assistance earns its cost only when it improves candidate inspection.
• Did: A one-screen representation protects working memory; detailed analysis stays with the receiving work.
• Scope: Practitioners use F.1 to select sources, not to establish truth, adequacy, semantic relations, synthesis, reliance, or assurance.

#Conformance Checklist

#Static checks

• SCR-F1-S01 (Question and use). The receiving question is independently identified as the note's exact EntityOfConcern; the intended use and the source difference that can change the answer remain in its ClaimGraph.
• SCR-F1-S02 (Exact sources). Every retained source and edition is recoverable.
• SCR-F1-S03 (Answer-changing roles). Every retained source has an inspected role; exclusions and source gaps are explicit.
• SCR-F1-S04 (Finite and inspectable). The cut can be held in view without dropping a known answer-changing source merely to satisfy a count.
• SCR-F1-S05 (C.2.1 identity and one result). The SourceCutNote's ClaimGraph, exact receiving-question EntityOfConcern, and effective ReferenceScheme are all recoverable. The scheme resolves the question, exact source editions, and role claims. A file, source list, one-screen form, or question-and-use bundle supplies none of these values by form and does not become another result kind.
• SCR-F1-S06 (Locality). No cross-source equivalence, merge, or relation is asserted by selection.
• SCR-F1-S07 (Temporal honesty). Designed and performed source claims remain distinct when the difference affects the answer.
• SCR-F1-S08 (Family neutrality). No meaning, relation, or membership decision relies on a domain-family label.
• SCR-F1-S09 (Named search policy). Any material search aid has a recoverable policy, corpus, method or model edition, and interpretation.
• SCR-F1-S10 (No algorithmic gate). A search reading changes the cut only after inspection of source claims.
• SCR-F1-S11 (Reopen conditions). The result says which question, use, edition, rival, counterexample, or transfer-boundary change reopens it.
• SCR-F1-S12 (Domain-method boundary). A required systematic or appraisal-bearing review remains with its domain method.
• SCR-F1-S13 (SoTA source roles). A cut used for a SoTA claim assigns every retained source one of the comparison roles defined in E.8:11, records it in plain wording, and says which retained contributions can change the answer. F.1 does not restate or supersede the E.8 definition.
• SCR-F1-S14 (No authority or currentness laundering). Official status, popularity, maintained status, citation count, publication date, freshness, or academic praise does not promote a source into the best-known line. An official or widespread source may still earn that role from its substantive comparison. Catalogue and publisher pages support identity/currentness only unless their substantive claims independently enter the comparison.
• SCR-F1-S15 (SoTA one-source guard and gap). A one-source SoTA cut is used only when that source critically synthesizes the serious alternatives and no known action-changing rival or counterexample remains hidden. Otherwise the cut retains the necessary comparison or returns a source gap.

#Regression checks

• RSCR-F1-E01 (Edition change). Recheck only answer-changing claims that used the changed edition.
• RSCR-F1-E02 (Use change). Recheck which sources, rivals, counterexamples, and limits matter when the question or use changes.
• RSCR-F1-E03 (New false friend). Add a concise warning when recurrent wording confusion changes the receiving answer.
• RSCR-F1-E04 (Bounded cut). Remove non-changing sources or split genuinely different questions when the active cut can no longer be inspected together.
• RSCR-F1-E05 (Search drift). A changed taxonomy, corpus, model, descriptor, scale, distance, threshold, or rank cannot silently change membership.

#Common Anti-Patterns and How to Avoid Them

#Consequences

The source cut becomes small enough to inspect, while a decisive rival, counterexample, or transfer limit cannot be discarded merely to satisfy a count. Later work can recover why each source was present, which edition was used, what the cut deliberately omitted, and what change reopens it.

The cost is explicit judgment. The author must say what each source can change and disclose a load-bearing gap instead of treating absence as evidence against a claim. Unknown rivals can still be missed, so citation chasing, expert leads, search aids, or a domain review method may remain necessary. Their outputs are candidates until source claims are inspected.

The cut is not evidence that its sources are true, sufficient, representative, mutually compatible, or safe to rely on. Those conclusions belong to the receiving synthesis, domain evidence method, F.9 relation, A.10 reliance account, or B.3 assurance claim.

#Changes that reopen or revise a cut

1. A source edition changes a relied claim, limit, or distinction.
2. The receiving question or intended use changes.
3. A new rival explanation, action-changing counterexample, or transfer limit appears.
4. A formerly retained source no longer changes the answer.
5. A language edition changes distinctions used by the result; treat editions separately unless the source supplies a mapping that preserves them.
6. Search-policy drift reveals a candidate but never changes membership without source inspection.
7. A later F.9 relation is proposed; keep source selection unchanged unless that proposal changes the source roles themselves.

#Rationale

Source selection for conceptual work is neither statistical sampling nor a race to the largest bibliography. The receiving question determines which source differences matter. A source belongs in the active cut when an inspected claim can change the answer, expose a rival or counterexample, or mark where transfer fails.

This keeps the useful part of purposive and iterative sampling—selection for a stated analytic purpose—without importing another field's population, richness scale, saturation claim, or count convention. It also keeps the useful part of systematic-search and AI-assisted screening—recoverable searches and efficient candidate discovery—without treating search coverage, rank, or reporting conformance as source adequacy.

#SoTA-Echoing

These comparisons change the source-selection method and its stops; they do not make a SourceCutNote a synthesis result or assurance claim.

#Relations

• E.8:11 owns the canonical SoTA definition, comparison-role meanings, and positive comparison contract. F.1 prepares only the question-relative source cut and records those roles for the receiving comparison.

Builds on.

• C.2.1 identifies the SourceCutNote from its ClaimGraph, the exact receiving question as EntityOfConcern, and its effective ReferenceScheme, and distinguishes that episteme from its representation.
• F.0.1 supplies a plain local reading during selection when a candidate's answer-changing role depends on a disputed expression.
• F.17 supplies a durable local-sense cell after selection only when later reuse, a claim, a named receiver, or an actual relation needs one.
• A.11 supplies the parsimony test; finiteness does not replace relevance with a fixed count.
• E.15 supplies affected-premise reopening across source editions.

Related later uses.

• F.0.2 can compare or synthesize the exact sources, roles, limits, and reopen conditions returned here.
• F.9 defines any actual relation between distinct recovered local senses; F.1 asserts none.
• G.2 supplies the broader refreshable SoTA-pack method when that heavier result is required.
• Use A.10 for reliance and B.3 for assurance; source selection establishes neither.
• An author of a subject pattern or DPF may use the source cut directly when an inspectable source basis is needed but no cross-source synthesis is current.

#F.1:End

Last Updated: 2026-09-10 — upstream FPF commit a87d0ef4 (github.com/ailev/FPF)