Skip to content

curriculum(SEF5): competency-derived testing, design and review progression - #162

Merged
christophersesugh merged 1 commit into
mainfrom
curriculum/sef5-competency-derived-engineering-practice
Nov 18, 2026
Merged

christophersesugh merged 1 commit into
mainfrom
curriculum/sef5-competency-derived-engineering-practice

Conversation

@christophersesugh

Copy link
Copy Markdown
Collaborator

M1 — SEF5 competency-derived software engineering practice review

Pinned main baseline: 23567c85e02c39ba4ac1773fdc4e5727e19c4471, after approved SEF4 #160.
Review authority: temporary review protocol v1.0, deliberately unmerged draft PR #150, M0 inventory #152, binding OEE specs and the curriculum design guide.
Scope: curriculum/SEF/SEF5.md and its two directly related regression suites (tests/test_sef5_design_quality.py, tests/test_sef5_migration.py). No other course, program catalog, specification, architecture decision or public audit file changed. The PR targets main directly.

1. Independent exit-competency reconstruction

A learner who has completed SEF2 must demonstrate the ability to make a deliberate, correct, reversible and independently reproducible engineering change to a real small Python CLI. This includes:

  1. Track precise source revisions, resolve a genuine conflict and revert an unwanted change without discarding unrelated work.
  2. Derive independently expected results, reproduce and repair bugs, test actual subprocess/persistence behavior and distinguish flaky test fixtures from product failure.
  3. Recognize substantive component responsibilities, dependency direction, public API/error/side-effect contracts and proportionate design alternatives; refactor without silently changing documented behavior.
  4. Run local reproducible quality checks, reason about CI privilege and untrusted input, and avoid inventing hosted CI execution.
  5. Write an actionable, accessible, evidence-backed design review; respond to a genuine independent reviewer without treating a mock review as actual external feedback.
  6. Reuse one versioned application and previously assessed evidence in a clean-start handoff and defend a meaningful unfamiliar change.

Comparative authority, not an imposed outline or accreditation standard:

Boundary: SEF2 owns programming, Python function/object basics and a beginner CLI. SEF5 owns quality of change and the evidence behind engineering decisions, not a second beginning-programming course. SEF9 owns project-level workflow, professional ethics, licensing, privacy, stakeholder accessibility and AI-assisted engineering governance. SEF10 independently integrates the complete program in a different bounded capstone. SEF5 teaches a readable, accessible review artifact of its own code but does not take over SEF9's wider professional-practice scope.

2. Outcome and assessment identity

The 15 current published outcome definitions (SEF5-O1–O15) are byte-for-byte identical. The seven SEF5-AS1–AS7 assessment contracts (each Work, Criteria, Evidence and reviewer variation) are also byte-for-byte identical. No new outcomes, assessments, examination, compulsory application or readiness status is introduced. Existing draft, v0.2.0, 10–12-week provisional duration, SEF2-only prerequisite, Python 3.11+ and offline-capable/no-paid-account route are unchanged.

In particular, the existing AS4 20-run observation is retained as a specific published assessment obligation, not imposed as a universal per-course template. AS6 and AS7 already permit reuse of the same CLI code and prior evidence. The genuine independent reviewer required for the final handoff is not silently weakened into self-review, an imaginary peer, or a simulated hosted PR. Conversely, actual hosted CI and hosted Git accounts are not required.

3. Module-by-module realignment: mechanisms, dependency order and outcomes

The former course had 9 modules / 74 numbered lessons. A majority of the earlier Maintainable Software, Design Boundaries, Code Maintenance, Software Quality and Tested CLI content repeated the same interface/review/change concepts at differing textual depth without a new assessment capability. Similarly, testing and reliability were split across old M2 and M5, and a number of lessons were just isolated nouns (“Assertions,” “Refactoring,” “Logging,” “Code review”). No pedagogical evidence justified each repeated teaching boundary. A separate standalone quality module is warranted, but repeated initial explanations of coupling, interfaces or test layers are not.

The replacement has 6 competency-owned modules, 51 explicitly described lessons, distributed 8/10/11/7/8/7. These counts are incidental to the ability and dependency reconstruction: there is no desired count or enforced numerical variation.

Module Owned teaching capability and all major lesson dispositions
M1 — Git Change Control and Reversible Collaboration (8 lessons) Versioned CLI context, staging/index/ignore policy, meaningful diffs, branch/merge relationships, actual conflict resolution, revert versus destructive history rewrite, and clearly simulated local pull-request review. Absorbs old M1 Git concepts and old M7's redundant change-history introduction. Direct preparation for AS1.
M2 — Independent Tests, Causal Debugging and Determinism (10) Combine old M2 and M5 around a single test-and-cause discipline: independent contract oracles, unittest unit/real subprocess integration separation, negative input partitions, observed pre-fix failure, hypothesis-testing repair, temporary fixture state, injected clock/seed/order, actual flaky-test diagnosis, existing AS4 repeated-run exercise, and an application defect that corrected tests must still reject. Supports AS2 and AS4 without inventing a per-module test quota.
M3 — Responsibilities, Contracts and Behavior-Preserving Refactoring (11) Consolidate old M3, M4 and M7 design/interface material: map real parser/domain/storage calls; diagnose cohesion/coupling and dependency direction from source; choose between simple functions and composition; explicit API/error/side-effect/mutation contracts; JSON/CLI compatibility; characterization tests before refactoring; fake/test double plus actual adapter subprocess tests; one bounded refactor; safe diagnostic logs; and reviewable alternatives. Supports AS3 and the existing additional design evidence of AS7.
M4 — Reproducible Quality Gates and CI Trust Boundaries (7) Retain the distinct CI ownership from old M6 and sharpen it: local quality entry point, correct failure propagation, clean-checkout/Python/dependency reproduction, separate failing test and static-check fixtures, reviewed least-privilege workflow, the untrusted code versus privileged secrets boundary (including pull_request_target), and honest local/hosted evidence distinction. Supports AS5.
M5 — Quality, Accessible Review and Maintainer Handoff (8) Merge old M7/M8 and repeated review lessons into a substantive quality judgment: multidimensional risk rather than metrics theater; diff/architecture inspection; defensible design alternatives; evidence-specific severity and comments; reproducible README; text/semantic Markdown accessibility; actual vs simulated reviewer provenance; compatibility, rollback and disclosed limits. Supports review evidence for AS3, AS6 and AS7 without duplicating the final product.
M6 — Integrated CLI Delivery and Independent Defense (7) Reuse the same course application with a public behavior/version contract, real multi-process/failure tests, module/adapter correctness, reproducible quality gates and recorded red/green evidence, reviewed diff and correction, genuine independent reader reproduction, and an unfamiliar bounded technical defense. Supports AS6/AS7; no second application.

Every module has one nonempty mapping to existing published outcomes, distinct lesson tasks with falsifiable/observable cases, and optional formative practice. There are no invented resource or platform prerequisites.

4. Assessment evidence and validity

Existing assessment Intended actual learning/evidence Change
AS1 (O1/O2/O10) Real Git index/history, branch merge/conflict and reversible fault handling in one throwaway repository; focused review summary M1 prepares it; assessment unchanged
AS2 (O3/O4/O12) Causal failure reproduction, independently expected results, unit and real CLI adapter tests, fixed regression M2 prepares it; unchanged
AS3 (O5/O6/O10/O12) Characterization-before-refactor, compatible parser/domain/storage boundaries, safe logs, reviewed diff/response M3 and M5 prepare it; unchanged
AS4 (O3/O4/O7) Two genuinely unstable test mechanisms, corrected deterministic fixtures, real 20-run result and a still-detected seeded product defect One M2 progression; unchanged
AS5 (O8/O9/O12) Independently red test and static-check results, actual nonzero local gate, safe minimal-permissions CI design and clean-checkout reproduction M4 prepares it; unchanged
AS6 (O1/O3/O4/O5/O6/O7/O8/O9/O10/O11/O12) One maintainable tested CLI with a real independent review and source/quality/bug evidence M6 integration; unchanged
AS7 (O3/O5/O6/O10/O11/O12/O13/O14/O15) The same AS6 CLI, real responsibility map, two alternatives, explicit contract, fake and actual integration, before/after impact and accessible quality review M3/M5/M6 prepare it; unchanged

The outcomes O13–O15 are independently assessable by the existing AS7, not by a new ornamental module quiz. Source code, real output/status/persistence bytes, reviewer feedback and qualified claims remain inspection requirements. Structural coverage alone is not competency evidence.

5. Confirmed findings and dispositions

  • F1 — major: duplicated test and diagnosis teaching. Old M2 and M5 separately introduced test levels, boundary cases, failure and flaky state. Resolve: one lesson chain in M2 from independent oracles to test layers, controlled failure, deterministic isolation and regression.
  • F2 — major: broad design and review overlap. Old M3/M4/M7/M8/M9 repeatedly discussed cohesion, coupling, interfaces, logging, contract compatibility and reviewer feedback. Resolve: separate one design/contract/change block (M3) from an independently useful quality-review block (M5), then integrate existing evidence in M6.
  • F3 — major: unobservable terse lessons. The early syllabus named “Assertions”, “Code smells” and similar nouns without an observable mechanism or counterexample. Resolve: each lesson names a distinct action, observation, boundary or failure and an explicit intended feedback opportunity.
  • F4 — major: fixed-module-count regression coupling. Both test_sef5_design_quality.py and test_sef5_migration.py hard-coded nine modules; the former also froze titles. Resolve: test sequential structure regardless of count, mapped/current outcomes, individual described lessons, genuine prerequisite chain (Git and tests → refactoring → gates and review → final independently reviewed handoff), version/history and no fake hosted CI or paid gate.
  • F5 — major: fake-only testing risk. A replaceable adapter should not be claimed correct simply because its in-memory test double passes. Resolve: reinforce integration evidence with actual subprocess/JSON state and an independently failing case; no new assessment imposed.
  • F6 — major: trust and evidence-provenance risk. Local workflow simulation must not be called a hosted CI run or authentic independent peer review. Resolve: stage the two contracts separately and require accurate observed provenance, secret/permission reasoning and actual external reproduction where AS6/AS7 already require it.
  • F7 — investigate, learner readiness/workload. The 10–12-week range is provisional, especially for learners new to Git and rigorous code review. Do not claim tested or verified status without real learner time and reviewer evidence.

6. Cross-course effects, historical compatibility, verification

SEF2 continues to own basic Python, classes, validated file persistence and the beginner CLI; SEF5 reuses that actual prior artifact and increases engineering evidence rather than starting another CLI project. SEF9 continues to own planning, ethical responsibility, data stewardship, licenses, stakeholder communication and AI-assistant governance; local review quality is already a legitimate SEF5 competency, not a duplicate SEF9 professional curriculum. SEF10's capstone remains separate and requires actual prerequisite closure.

Verified against pinned main:

  • Entire Learning outcomes section: byte-for-byte identical, 15 IDs.
  • Entire Required summative assessments through Completion decision: byte-for-byte identical, seven complete published assessment contracts.
  • All fifteen outcomes map to explicit modules, and all seven assessment IDs still map to the required outcomes.
  • Six sequential modules, 51 individually described lessons, no prescribed distribution, no new public review/audit file.
  • Updated two targeted tests to enforce actual preparatory competencies and safety/portability of evidence, not cosmetic counts.

CI is to be verified after PR creation. Academic text and green CI are not substitutes for independent learner review, provenance or pedagogical validation.

7. Maintainer approval, DCO, and next course

The GitHub connector writes unsigned API commits and cannot personally attest to the contributor's DCO. Before merging, the maintainer should verify the review and all checks, and squash these commits into one personally signed-off local commit in a clean checkout:

git fetch origin
git switch curriculum/sef5-competency-derived-engineering-practice
git pull --ff-only origin curriculum/sef5-competency-derived-engineering-practice
git reset --soft 23567c85e02c39ba4ac1773fdc4e5727e19c4471
git commit -s -m "curriculum(SEF5): competency-derived engineering practice and review"
git push --force-with-lease origin HEAD

Keep temporary draft PR #150 unmerged; it is the review protocol, not an integration branch. This course PR should merge to main only after the required DCO and CI checks are green and the maintainer approves. Next review in M1 sequence after the merge: SEF8 — Mathematics for Numerical Computing & AI.

Signed-off-by: Christopher S. Aondona <christohybrid185@gmail.com>
@christophersesugh
christophersesugh force-pushed the curriculum/sef5-competency-derived-engineering-practice branch from 2cde443 to f50540a Compare November 18, 2026 12:47
@christophersesugh
christophersesugh merged commit c200af1 into main Nov 18, 2026
7 checks passed
@christophersesugh
christophersesugh deleted the curriculum/sef5-competency-derived-engineering-practice branch November 18, 2026 12:47
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant