E.4.PFAD - Principle-Framework Architecture Decision

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

E.4.PFAD:1 - Problem frame

Use this pattern when an author is choosing among a new or revised principle framework, a contribution to an existing framework, another product that is not a framework, a thinner publication or access route, and no new maintained product now, and that choice will settle an identity, edition, relation, intended-use, or publication decision that later work must use. The decision may concern the public field and first use, framework edition, dependencies, initial pattern placement or relations, the kind and identity or change rule of a non-framework product, or the publication or access consequence. Another author or reviewer must need the answer and its rationale for later action.

E.4.DPF may return this question before a public product name, PatternIDs, or pattern bodies exist. Open it when preliminary exact subtraction leaves exact ownership unresolved, a coherent connected remainder, or a material field, relation, source or refresh, publication, or access consequence that later work must use. Treat an incoming carrier or organizing cue—for example, a role account, MethodDescription, file, Card, mantra, shared topic, or missing authoring artefact—as evidence for that comparison. Decide the outcome from exact candidate contributions and the later-used consequence.

Here product has the Plain management meaning declared in E.4:4.1; it is not a technical kind. When a non-framework product is selected, the answer names its direct subject, kind, and the identity, current-state, provision, publication, availability, or other relations used by the decision. Name a maintenance relation only when that stronger claim separately obtains and changes the answer. If a kind or relation that can change the answer is unresolved, keep it as an explicit decision question and do not invent U.Product.

A proposed new or substantially revised DPF also needs an answer about its field boundary. That answer says who can first use the framework and for what, which connected problem families and useful results it covers, what the current FPF and admitted DPFs already provide, and what remains uncovered. It compares serious alternatives, tests one representative case that crosses problem families, states where the evidence runs out, and names the change that will require a refresh. Together these must support one independently usable pattern language. One pattern or a narrow authoring slice is not a DPF merely because it has a broad title or a coherent carrier.

If a cheap search, curated reading route, useful contribution to an existing framework, suitable non-framework product, or stop answers the immediate need without settling such a boundary, use that result and stop. Reserve the framework-architecture DRR for a later-used boundary.

When the architecture question is live, use E.4.PFAD to state the framework-specific content of one ordinary E.9 DRR. The decision-maker selects the answer during decision Work; that DRR is its decision record. This practitioner-facing profile locates the questions and answer content inside the DRR, while acceptance remains a separate decision.

E.4.PFAD:2 - Problem

Framework authors repeatedly need to decide whether a recurring practitioner problem calls for a new framework, an existing framework contribution, another product such as a programme, service, or evidence package, a thinner access result, or no new maintained product now. When a framework is selected, later work needs its public field promise, first-edition boundary, problem-family coverage, first patterns and their material relations, representative use, important omissions, and publication or access consequence. They also need the selected FPF edition, the Core claims required for each receiving use, and any other framework-edition dependencies. Generic decision prose can hide those choices.

A small coherent authoring slice creates a common false positive: its few current patterns and neat structure are mistaken for a field-scale pattern language. Source diagrams create another: one list or hierarchy is copied into the DPF although Methods, Work, subjects, descriptions, capabilities, providers, and cultural change may have different structures. A large framework-specific form creates the opposite problem by making proposal, acceptance, DRR, edition, authoring, quality review, and publication look like one extra decision object.

The useful result is one readable answer whose framework consequences and limits are visible without adding a second decision stage or making cheap exploratory work produce decision paperwork.

E.4.PFAD:3 - Forces

E.4.PFAD:4 - Solution

E.4.PFAD:4.1 - Decide whether the architecture question is open

Ask whether choosing a framework, a non-framework product, a thinner route, an existing-framework contribution, or stop will settle at least one decision used by later authoring or review:

• the public field promise, a first use that does not depend on unpublished authoring context, or the problem-family coverage of a proposed DPF;

• an intended or existing framework edition;

• a dependency on a selected FPF edition for named Core claims and a receiving use, or on another identified framework edition;

• initial pattern placement or a material relation among those patterns that changes the architecture;

• the direct subjects and identity or change rules for a continuing programme, an admitted service, or a separate editioned result, plus any maintenance relation when later work separately claims or uses it;

• a publication or access consequence; or

• for a proposed DPF Suite, the ecosystem use, which product series may belong, constitution, inclusion and removal rules, identity through change, source return, later-review and retirement conditions, exposure choice, any separate DPF Suite Reference product decision, and any maintenance relation only when separately claimed.

When E.4.DPF supplies a preliminary contribution account, use the contributions rather than their carrier as the input. If same-situation comparison shows that exact current FPF or admitted-DPF owners carry every action, first result, return, and source or refresh duty—or exact external results supply them—and no decision above remains, take the smallest useful result and close without PFAD. For connected Methods, establish the supplied whole use through E.4.CM:4.4 before treating the constituent suppliers as a complete answer. Keep the architecture question open when exact ownership remains unresolved, a coherent action-bearing remainder survives, or closure would erase a material relation or field or refresh responsibility that changes a later use. Carrier form, role label, shared topic, naming and PatternID state, and counts locate candidate contributions. The four dispositions and later-used consequence determine whether to close or select an outcome.

If none of these decisions and no receiving use are present, take the exploratory result as the answer. If one is present, the decision-maker selects a framework, non-framework product, thinner route, existing-framework contribution, or stop during decision Work, and one E.9 DRR records that answer. The cheap exit and the architecture decision are alternative entry outcomes.

For every product alternative, use product only as the first management cue. Then compare the direct subjects at the same grain: the exact framework or package episteme, System, service arrangement, Method, programme description, carrier, or other admitted result, and the relations that later work will rely on. Use a quality-management, service-management, publication, or content-management scheme as a probe; use the FPF direct-subject patterns to settle the kind. If an unresolved kind can change the selected answer, keep the product proposed and make that kind the next decision question.

E.4.PFAD:4.1.1 - Choose a contribution through attainable use and its whole cost

Use this construction when the architecture choice turns on who can benefit from the proposed content, how that benefit becomes available, or which uses must change together. It supplies grounds for the compact answer in :4.2; it is not another record or an obligatory study before a small repair. If one adequate route already answers the question without a consequential boundary choice, keep the :4.1 cheap exit.

Start with a useful occasion, not a catalogue of possible mistakes. Describe a recurring or credibly anticipated situation, the result worth obtaining, the presently available way of obtaining it, and the consequential difficulty. Distinguish frequency of the occasion, frequency of failure conditional on that occasion, severity of its consequence, and the proposed remedy’s effect. A frequently encountered task can justify assistance even when mistakes are rare; a rare severe failure can justify protection without being called frequent. Source occurrence is evidence that a method was described, not that the intended population needs another description. State which claims are observed, source-supported, estimated or still hypotheses. Use C.11.DUA to decide which uncertainty is worth investigating.

Describe the actual reader and performer arrangement. Separate general and domain education, skill in the relevant work, familiarity with FPF/DPF distinctions, and the means available on this occasion. Preparation can differ for recognizing, performing, checking, adapting and teaching the same Method. Use only differences that could change the arrangement; do not classify people or models by a single rank of intelligence.

These are common arrangements, not an exhaustive role taxonomy or three mandatory editions. A learner can later need only a reminder; an experienced practitioner can need the full reasoning when conditions change. If the intended result is the person’s independent judgement, keep that result in the comparison instead of substituting a good answer delivered by the tool. If the intended result is obtaining the answer from a provider, do not require every recipient to learn the provider’s Method.

Follow the route by which the benefit could occur. Begin with the work question the user recognizes, without silently supplying the PatternID. Follow discovery, recovery of the task-relevant distinctions and connections, selection of a suitable action, performance of the Method or obtaining the required result from a capable provider, and the receiving use of that result. E.11 and A.15.11 supply discovery and occasion; C.2.8 makes recovery relative to preparation, accessible material and effort. Find the first necessary contribution that the proposed arrangement cannot supply. It can be a missing method connection, inaccessible explanation, unacquired skill, unavailable assistance, or a result that does not serve the intended work. These require different repairs.

If the contribution cannot be found or its necessary structure cannot be recovered in this arrangement, the promised benefit through that route is unavailable; search and reading can still consume resources. This is a causal constraint on value, not a conversion of structural information into money. More extractable structure can also be irrelevant to the task. Distinguish an observed failed attempt, a reasoned walkthrough and an untested access assumption; unknown performance is not a measured zero.

Construct serious alternatives for the same result. Compare the available public method and no addition with the consequential alternatives: repairing discovery; a reminder or checklist for an already possessed capability; a developed pattern or composite; a pedagogical companion; external competent provision; or distinct variants with a small shared supplier. A checklist can prompt a judgement without teaching how to make it. A companion can supply explanation, practice and feedback without creating a different subject Method. E.8:4.2.1.1 assigns their respective content and tests whether teaching can be constructed without inventing the Method. A short route and a full explanation can belong to one architecture.

For one composite Method or connected independently usable Methods, including alternatives, compare the whole receiving use and its necessary connections under E.4.CM:4.4, not only the available constituents. For an unprepared audience, make learning or external help a real alternative with its costs and availability; do not hide it in an assumed expert reader. Conversely, do not force experts through elementary instruction that changes no relevant result. A result beyond the present capability and budget can warrant a different allocation, staged development or an honest unmet use.

Compare the whole cost of obtaining and sustaining useful application. Use a stated horizon and plausible number of users and occasions. Include only terms that can change the decision, but include them where they fall: developing and maintaining the contribution; finding and interpreting it; learning, practice and feedback; obtaining help; adapting and integrating it into work; recurring application, checking and recovery; and responding to consequential changes. For AI-supported work, computation and retrieval are costs alongside human prompting, checking, waiting and correction. Include displaced work and the cost of retaining a capability or dependence when those matter. A free document is not free use.

C.11.CRC compares complete configurations rather than summing isolated pattern merits. C.11.DUA compares added work with attainable gain; C.19.2 limits how much decision apparatus is worth applying. A useful ordinary calculation is one-time acquisition and setup, plus recurring cost over the horizon, plus consequential change and retirement costs. Count shared development once at the chosen population boundary; state any allocation to individuals. Do not charge an author’s investment in full to every reader, count the same correction both as expected loss and paid rework, or amortize over an invented infinite audience. Human time, compute, money, delay and serious non-compensable losses need not collapse into one score. Use ranges or break-even conditions when a point estimate would pretend to knowledge.

Compare acceptable result quality and constraints together with these costs. Lower cost for a different or inadequate result is not a saving. Past learning is part of the available starting capability; further learning can create benefits beyond this one application. State those future uses as supported expectations or hypotheses. Identify who pays and who benefits: an aggregate saving can leave the intended user unable to afford adoption. A rare but unacceptable hazard is handled under its governing obligation, not traded away by a small average cost.

Challenge sharing by an independent change. Vary one premise that matters to one intended use or audience: a domain criterion, teaching prerequisite, available tool, or required outcome. Determine which explanations, interfaces, dependent uses and release obligations would actually need to change. Common ideas, reused Method descriptions and a common running implementation create different dependencies. Separate files alone do not create independent change; a shared carrier alone does not prove shared semantics.

Compare a shared supplier with a bounded interface, independent variants, and limited deliberate duplication. Retain sharing where it saves enough construction and maintenance while preserving useful independence. Split where accumulating exceptions, migration, compatibility or coordinated releases defeats that gain. C.31 supplies the relevant modularity and reuse characteristics; select those that answer the live question instead of maximizing reuse or requiring every coordinate. E.4.PFR handles edition dependence when that relation actually obtains.

Spend the next effort where it could change the choice. Try the smallest consequential uncertainty: can the intended reader find the contribution; can the intended reader recover a necessary connection from the available public explanation; can a user complete a supported case; does a changed premise force unrelated variants to change? If the reader must invent a consequential connection, retain that construction as a proposed contribution and compare its useful effect and cost; the private invention does not establish that the public explanation supplies the connection. A prospective benefit is a hypothesis that can justify a bounded experiment, not a realized return. Separate a content-recovery probe from an effectiveness or learning trial. In a teaching claim, assisted performance and later independent performance can require different observations. Stop expanding the analysis when the available distinction already selects the next action, or when further inquiry costs more than the decision it could improve.

Return the selected contribution and arrangement, the useful result, preparation and support on which it relies, material costs and uncertainty, the chosen sharing boundary, and the observation that could reverse the decision. These are grounds for the existing :4.2 answer, not an extra universal form.

E.4.PFAD:4.2 - State the compact framework answer

When the architecture question is open, the framework-specific part of the DRR states:

1. the intended practitioner, public field name and promise, recurring problem, and bounded architecture question;

2. the selected outcome: a new or revised framework edition, a contribution to an existing framework, a non-framework product, a thinner publication or access route, or no new maintained product now; for a non-framework product, also the direct subject kind and the identity, current-state, provision, publication, availability, or maintenance relations actually used by the decision;

3. its field boundary: who can first use it without unpublished authoring context and for what; the connected problem families and useful results; what the current FPF and admitted DPFs already provide and what remains uncovered; serious alternatives, such as splitting or merging the proposed framework, using existing sources directly, contributing to one existing framework, composing exact contributions across several admitted DPFs and FPF, selecting a programme or service, selecting a separate evidence-package episteme, or selecting no new maintained product now; the limits of evidence; and what change will require a refresh;

4. the selected problem-family pattern sets, first patterns and their material relations, representative cross-problem application, and important omissions;

5. which structures of Methods, their use, support and development change the answer, and how the identified Methods, descriptions, patterns, direct subjects and managed result boundaries fit together. When those structures do not line up one-for-one, use a completed C.32.MWA synthesis; use E.23.CDI only when capability development for a named Work family changes the answer;

6. the existing or intended framework-edition boundary; the selected FPF edition, the Core claims this framework relies on and the receiving use that requires each claim; and any other framework-edition dependencies required by this answer, under E.4.PFR;

7. the sources to revisit for each important claim, whether the evidence supports, suggests, or only motivates it, the limits of that evidence, and the publication or access consequence; and

8. material alternatives, accepted costs or losses, practical consequences, the first authoring action or stop, and the reopen condition. For each framework-edition dependency, distinguish a change of the relied-on edition from a material change to a relied-on claim, and identify the dependent use that each reopens.

For any opened question, record only candidate contributions that can change the outcome. For each, state whether an exact current FPF or admitted-DPF owner carries its action, first result, return, and source or refresh duty; an exact external result supplies it; an action-bearing remainder survives; or ownership remains unresolved. A shared topic or broad framework name can locate candidates; assign a disposition only after exact contribution comparison.

When professional Method coverage can change point 5, the same compact framework answer projects five connected claim groups from points 1, 3, 4, 5, 7, and 8. The projection remains content of that one answer and its one ordinary E.9 DRR. Fill each group to the grain that changes first use, using a representation suited to the case, before DPF authoring:

1. Practice truth and first use: identify every bounded practice claim or promised practice contribution by its exact subject and scope, mark that claim—not the answer as a whole—as obtaining or possible-future, and state practitioner, recurring or anticipated difficulty, sought result, first use, stop or wrong-turn return, qualification window, and receiving decision. Include only non-use boundaries admitted by F.19’s grounded-contribution test.
2. Project and Method positions: name the direct project subjects, use and environment, materially different solution forms, and Methods under their actual operational, system-change, solution, Method-of-interest, or Method-development relations. Keep incumbent Work, development or trial Work, candidate-practice Work, and intended Work distinct.
3. Selected structures and correspondences: include only the Method, Work, subject, transformation-flow, capability/provider, description, contribution, Method-development, and cultural structures whose correspondence, conflict, or non-isomorphism changes the answer.
4. Pressures and evidence: keep constraints, conflicts, failures, environment or interest changes, and observed, source-supported, estimated, contradicted, and missing links distinct from causal history and temporal unfolding.
5. Contribution, subtraction, gaps, and reopen: apply the four dispositions above, then name each receiving pattern and domain filling still needed, honest omissions and gaps, and the observation that reopens the architecture.

Within an existing-framework contribution or another answer that selects no new framework, distinguish one existing DPF receiving a contribution from a use that composes exact contributions supplied by several DPFs and FPF. Record the latter as cross-DPF composition under the selected one of the five outcomes, optionally exposed through a role-centred view or access route. Decide Suite membership separately.

One answer may contain several bounded practice claims with different truth status. Every selected practice question names the claim or claims it consumes, so an obtaining incumbent-practice claim can coexist with a possible-future candidate-practice claim without backdating the candidate or erasing current incumbent coverage. Independently obtaining A.13 agency claims and actual development or trial Work keep their own status; neither proves that the candidate practice obtains. Public coverage is another claim and remains limited to the exact obtaining or prospective contribution and later package evaluation.

For an obtaining practice claim, name actual recurring difficulties and representative actual Work. A precise Agent-performer branch first supplies A.13’s core: the exact admitted System, local agential system-role kind and criterion, classification, obtaining assignment, and needed scope, working situation, and window. Add the agency-characteristic profile only when a Grade, autonomy or profile claim, a criterion-dependent characteristic, or a named assurance use consumes it. A.15.1 then independently admits the actual Work from its performance history, Method, extent, and containment. Only after admission does F.6 supply any precise assignment-bound attribution through that same obtaining assignment. State evidence limits; a missing F.6 relation leaves admitted Work intact and only the attribution unresolved.

For a possible-future practice claim, name intended use, incumbent Work or Method and observed problem evidence, candidate Methods and architecture, realization conditions, a planned representative trial, expected acceptance and failure observations, and reopen conditions. Any incumbent-practice or actual trial-Work claim stays independently obtaining when supported, but the candidate-practice Work, candidate-practice Agents, and current candidate-practice coverage remain unasserted until their own conditions obtain.

For every selected question, name its receiving pattern and the exact bounded practice claim or claims whose values change first use. If a required group or claim-to-question binding is absent at that grain, return a bounded PFAD gap to the architecture decision for completion before DPF authoring. A completed C.32.MWA synthesis is used only when several selected structures do not line up one-for-one, and E.23.CDI only when capability development changes the answer.

The answer is one identified claim-bearing episteme under C.2.1 and is recorded, with its rationale, in one ordinary E.9 DRR. The decision-maker selects that answer during decision Work. An authorized decision-maker accepts, redirects, rejects, or reopens it through a separately identified acceptance decision. Record that accepting decision separately and hand its exact accepted answer to E.4.DPF.

Common practice questions include:

If another question changes the answer, name it and the pattern that handles it. A required contribution, transformation, performed Work, capability, provider contribution, or cultural change belongs to its own pattern; B.1.5’s complete relation predicate supplies Method parthood. For a DPF Suite answer, an architecture decision takes effect to constitute the continuing collection. It selects the ecosystem use, which product series may belong, inclusion and removal rules, identity through change, alternatives, practical consequences, and the reopen condition. The same E.9 DRR records that answer. Publication and availability of the first or a later edition are separate occurrences. A maintained-Suite claim separately identifies the maintenance relation, capable System, any commitment that actually obtains, and the refresh response. For any current or available Suite claim, apply E.4:4.2’s direct currentness, availability, and source-return conditions, including return to every product-series state presented as current. When the answer selects exposure, choose an independent Suite route, a bounded projection in a current DPF Suite Reference edition with source return, or a neutral combined carrier. Constituting and including the Reference product series, admitting its editions, publishing them, making them available, maintaining them, and refreshing their answers remain separate decisions and claims. Record a proposed result use or future constraint as such; apply E.4.PFR after edition-level case facts establish a dependency or compatibility relation.

For an existing-framework contribution, non-framework product, thinner route, or stop, state only the parts needed to explain that outcome and the later-used decision. A selected product still names its direct subjects and the relations used; a proposed product with an unresolved kind says so. The selected outcome determines which field-assessment or package content applies.

When the architecture keeps, merges, removes, reuses, or omits a load-bearing contribution, record the E.8:4.1.3 same-situation disposition and the action or result that changed. Preserve a difference when it changes action at comparable effort without adding an unsupported or needless burden. A narrower label or example alone is not that difference.

When the answer treats a promised problem family as covered by a result supplied from outside the framework, name the exact result, its exact relied-on content and direct kind, supplying product and exact edition or current state, receiving use, practical discovery route, and every currentness or availability condition that can change that use. State that the result remains external, and state maintenance only when it changes the receiving use. When the receiving use requires an availability or compatibility result, name that exact result and its exact basis. If those facts are absent, or the result does not answer the promised use, record a gap or omission rather than relabelling the result as framework content, a MethodDescription, or source evidence. When the selected keep, merge, removal, profile, external reliance, or omission materially changes the stable set for a promised problem family, obtain a current E.4.DPF.DA D12DomainProblemFamilyCoverageAdequacy result for the resulting exact DPF or LPF edition. Reuse a matching current result while that exact resulting edition and basis are unchanged. Do not require proof that a revisit occurred.

Keep the ordinary E.9 grounds, sources, affected loci, rationale, and consequences in the same DRR. Add naming, quality, admission, currentness, or package details only when they change this answer or a named later use requires them. Each added claim remains under the pattern that defines, constrains, or tests it and appears in PFAD for that named use.

E.4.PFAD:4.3 - State initial pattern relations directly

When an initial pattern relation changes the selected architecture, state the relation and its participants as an ordinary assertion. For example: Pattern A frames the recurring problem; Patterns B and C specialize its reusable move for two stated situations. Use the pattern that defines or constrains each relation function.

For the architecture answer, use the direct assertion under the relation’s own predicate. Add an optional E.4.PFR row when a named maintenance use needs that representation.

E.4.PFAD:4.4 - Keep the answer, DRR, authoring, and publication distinct

The decision-maker selects the answer during decision Work. The E.9 DRR records that answer and its rationale. An authorized decision-maker accepts, redirects, rejects, or reopens it through a separately identified acceptance decision. Later authoring follows an accepted answer. A framework edition is the pattern-language episteme assembled from accepted sources; any maintenance relation obtains separately. Publish or project claims about these objects in the selected form—for example, an ADR-like document, site, or PDF—and identify its presentation carrier separately when needed.

When the answer uses C.32.MWA or E.23.CDI, keep each proposed Method distinct from the pattern that describes it, the Work in which it is performed, the result of that Work, the framework answer, the DRR, and the resulting edition. Use a proposal or evidence locator to find the supporting material.

Use C.32.PAD only when the question is an exact project architecture decision about a named composite project Work, and use C.32.ADR only to project that project decision. For an ordinary framework answer, publish the selected decision episteme or a reader-specific projection through E.17 and E.24.PUB. None of these is a mandatory stage of principle-framework authoring.

E.4.PFAD:5 - Archetypal Grounding

E.4.PFAD:5.1 - Positive DPF

A systems-management group considers a public DPF for recurring problems in service launch, cross-team coordination, incident response, and feedback-based improvement. A broad FPF route covers several shared distinctions, and an admitted neighboring DPF covers one specialist branch, but neither gives this practitioner group a coherent first use across the four problem families. The field-boundary assessment compares a new DPF with direct FPF-and-source use, a guide, contribution to the neighboring DPF, two existing DPF edition series, and no new maintained product now. It favors one DPF because a representative service-launch case needs patterns from several problem-family sets together and has an independent edition, change, and refresh boundary, including its later-review rule.

The source accounts organize Methods, dated Work, service and equipment subjects, descriptions, provider capabilities, and cultural change differently. A completed C.32.MWA result makes those correspondences and conflicts readable; the selected problem families and relations supply the DPF structure. The E.9 DRR records the public promise, selected problem-family sets and material relations, representative case, omitted procurement and certification questions, the sources to revisit, which claims the evidence supports, suggests, or only motivates, the publication consequence, first authoring action, and reopen condition. It also identifies the selected FPF edition and any other relied-on framework editions, their required claims and receiving uses. Capability development does not change this case, so E.23.CDI is absent. PFR rows and proposal locators serve their conditional representation and discovery uses.

E.4.PFAD:5.2 - Exploratory access result

Existing FPF and source material answer the immediate need through a curated route. The inquiry closes with that route because no later author or reviewer needs a settled framework boundary.

E.4.PFAD:5.3 - Decision-level access result

A team needs a settled choice among a DPF, an access route, and stop because later work depends on the rationale. The architecture question is therefore open. The team selects the access route and no new framework edition. One E.9 DRR records the selected access consequence, stop, and condition for reconsidering the answer.

E.4.PFAD:5.4 - Non-framework programme product

A cross-domain inquiry need recurs, but practitioners do not need another pattern language. The decision compares a DPF, an inquiry-programme product, a separate inquiry evidence-package episteme, a curated route, and no new maintained product now. It selects the programme because named users need continuing access to inquiry Methods, bounded-project intake, and result return. The answer records the programme’s admitted direct subject and relations under their owning patterns. Any maintenance relation is a separate claim.

Its first usable version is a current programme-description episteme that names the users and questions, inquiry Methods, project intake, result return, access, change, and retirement rules. Name any provider System, maintenance relation, accepted commitment, or admitted service state when it independently obtains and changes the answer. Each bounded inquiry project requires independent A.15.1 admission as a separate Work occurrence, and each returned result is a separate episteme. A subject pattern may instead admit the programme itself as a System or another exact arrangement, in which case the answer names it. A bounded project may end while the managed programme continues and evolves. The inquiry evidence package remains its own editioned episteme.

E.4.PFAD:5.5 - DPF Suite and Reference

Three separately constituted DPF product series already cover one recurring practitioner use. The live architecture question is Suite constitution and membership for their shared ecosystem use. When one architecture decision takes effect, it constitutes the continuing Suite collection, states its ecosystem use, defines which product series may belong, selects inclusion and removal rules and identity through change, and chooses source return and exposure. Its E.9 DRR records that answer and the initial inclusion decisions. Each DPF edition still belongs to its own product series under that series rule. The answer separately decides whether a DPF Suite Reference product series is constituted and included and states its edition-admission, source-return, later-review, and retirement rules. Publication and availability are separate occurrences. A maintenance relation, maintaining System, or commitment is recorded only when it separately obtains. A Reference edition may then give a problem-led cross-DPF answer; Suite constitution and membership remain with the Suite decision. Record a proposed cross-DPF result use as such, then apply the edition-level dependency or compatibility predicates when their case facts exist. A later author returns an inclusion or removal proposal to the Suite decision, which settles that membership.

E.4.PFAD:5.6 - Existing framework

A local practice framework already has an accepted architecture answer and a source record. Reopen that answer when its selected edition boundary, dependencies, initial pattern architecture, or publication or access consequence changes. A change of a relied-on edition or a material change to a relied-on claim first reopens the affected use; reconsider the architecture answer when that affects one of those architectural choices. An example or publication-carrier edit that leaves those choices intact follows the existing answer.

E.4.PFAD:5.7 - Candidate recognition before product form

These cases apply the same contribution comparison at a carrier- or count-shaped entry.

E.4.PFAD:5.8 - Same useful Method, different costs of access and learning

A framework author must make a method for recurring case reviews available to a group. Participants must identify the case question, obtain reasons for proposed changes, and let the case author accept or reject those changes with reasons. The method’s operative distinctions and conditions are already settled. The architecture question is how participants with different preparation can learn and keep using it.

Alternative A puts a complete tutorial and the use instructions into one required reading route. Alternative B retains the complete method account for competent readers, supplies a beginner companion with practice and feedback, and exposes a short reminder for participants who already possess the capability. Both alternatives must achieve the same case-review result; B’s shorter reminder alone would be an inadequate third alternative for beginners. Direct qualified facilitation and continuing with the existing method remain alternatives if their availability or cost changes the choice.

For an illustrative cohort calculation, stipulate adequate results for A and B, 20 participants and 20 occasions per participant. All numbers below are assumed human minutes, not measured effects or a universal valuation of time.

B spends more on authoring and initial learning, but saves 2,000 minutes over this stipulated horizon. Its additional shared cost is 1,200 minutes and its per-participant saving is 160: it breaks even at 7.5 participants, so eight or more favor B on this cost alone. For five participants, A costs 2,475 and B 2,875; B then costs 400 more. Different performance, expected use or support availability can reverse either answer. The author first needs evidence about the recurring burden or likely audience, not another abstract declaration that reuse is desirable.

Now a teaching example proves unsuitable for beginners. With B, revise that companion’s example and the teaching result it supports; retain the unchanged operative method and expert reminder. If the rule for deciding which reasons warrant a revision changes, inspect all dependent forms. A common publication may still need a release, but unchanged subject content need not acquire a pedagogical exception. This change probe explains the selected boundary; separate file names alone would not.

An AI that supplies a polished case answer has not thereby taught a participant to evaluate reasons. If independent human judgement is the promised result, add the needed practice and feedback and test it. If obtaining a competent case answer is the promised result, direct AI or human provision can instead be compared on that result and its costs.

Cheap exit. A prepared participant repeatedly overlooks one already understood question, and an existing available checklist already prompts it at the right occasion. Keep that route unless a consequential defect appears. No new pattern, companion, audience survey or cost table is needed.

E.4.PFAD:6 - Bias-Annotation

Scope: limited. This profile decides a later-used architecture boundary for an FPF-grounded framework, adjacent result or service, thinner route, existing-framework contribution, or stop. It does not provide a universal product ontology, a service-design Method, a publication taxonomy, or a mandatory decision form for exploration.

The first drift is form-first decision making: a team starts from a schema, row, ADR heading, or status field and assumes that filling it has settled the architecture. Start from the reader’s problem, alternatives, later-used boundary, and practical consequence instead.

The second drift is machinery-first entry: proposal, dependency, quality, naming, and publication apparatus appears before the reader knows whether a framework decision is needed. Keep that apparatus conditional on its own receiving use.

The third drift is slice-as-product: the small set currently being authored receives a broad DPF name before its field promise, several problem families, representative first use, omissions, and refresh boundary have been tested. Treat the slice as a seed or existing-framework contribution until the field-boundary assessment supports a DPF.

The fourth drift is architecture-by-layout: source rows, levels, chapters, or diagrams become the product structure. Recover the Methods, Work, subjects, descriptions, capabilities, providers, cultural change, and their actual relations first; use C.32.MWA when several structures must be reconciled.

The fifth drift is relation-by-representation: a table row or reference list is treated as the relation it records. State the relation directly; add a representation only when a named maintenance or checking use needs it.

E.4.PFAD:7 - Conformance Checklist

When audience access, whole cost or shared change can alter the architecture answer, apply :4.1.1: keep the promised result fixed, identify relevant preparation and support, compare an adequate available alternative, and expose the decisive cost or change dependency. This is a conditional substantive comparison, not a required numerical score.

E.4.PFAD:8 - Common Anti-Patterns and How to Avoid Them