Protocol: how the pool of sources was built

This file records how the architecture-design review was searched, screened and scored, so that someone else can repeat the search. The rules shared by all reviews are in the evidence index (method). All searches of this review ran on 2026-10-29. The terms it uses (architecture, architectural knowledge, viewpoint) are defined in architecture-descriptions; decision records are in design-decisions.

1. Why grey literature

Of the seven questions in Garousi 2019 Table 4, one yes would justify grey literature; all seven are yes for this topic (the rule).

# Question Answer for this topic
1 Complex, not solvable from formal literature alone? yes: styles, methods and the monolith debate come from practice
2 Lack of volume or quality of evidence? yes: few controlled experiments on structure (Zhang 2012 found 11)
3 Context important? yes: team size, domain and scale decide which structure fits
4 Corroborate research with practice? yes
5 Challenge practice with research, or the reverse? yes: practice moves between microservices and monoliths ahead of the evidence
6 Would a synthesis serve both communities? yes
7 Large practitioner volume? yes: the microservices survey alone has 1,502 respondents (Loukides 2020)

2. Questions

Id Question
Q1 Which design methods exist, and what do they share?
Q2 What do modularity principles (information hiding, coupling, cohesion) predict, measured how?
Q3 Do styles and patterns help maintenance, measured?
Q4 How do quality attributes drive the structure (tactics, trade-offs)?
Q5 How are architectures evaluated before building?
Q6 How much design up front, in agile practice?
Q7 How does structure follow the organisation (Conway, mirroring)?
Q8 What do erosion and architectural debt cost, and how are they found?
Q9 Monolith, modular monolith, microservices: what is the evidence and the practice?
Q10 What is measured when an architecture or design is handed to a model that writes the code?
Q11 How do architects reason while designing, and what biases affect them?

Adjusted after the frame: Q10 was added as the brief's strand "designing for a model as implementer". Q11 was added in the second pass. Its strand, design reasoning, had not been searched in the first pass.

3. Existing reviews (Garousi Guideline 2)

  • Searched first, with strings R1 to R12 below, top hits by citation count read by title.
  • Found: 14 reviews, maps or surveys. Of these, 9 were read by abstract, 3 by a summary of the abstract, 1 by a summary line, and 1 was located only. One more (Tofan 2014) is reviewed in design-decisions and linked.
  • A tertiary study of software architecture in general was not found: R2 returned 2 hits, a study of systems of systems and a study of sustainability.
  • They are placed in review.md strand 0.

6. Stopping rule

  • Fixed before searching: effort bounded at N = 10 results per string, with saturation at k = 5. Five consecutive results adding no new concept end a string; G3 ended this way.
  • Formal: effort bounded (top 7 to 10 per review string); the leads were checked one by one.
  • Q/A: the parameters are in the subsection above; the run ended after AR15.

7. Selection

  • Include: a source on the choice of structure of one software system, its method, principle, evaluation, change or cost; English; readable on the search date.
  • Exclude:
    • documenting the architecture (architecture-descriptions), decision records (design-decisions), distributed runtime behaviour (system-design), specification languages (algorithm-design);
    • function-level planning for code generation (algorithm-design);
    • vendor posts selling a product.
  • Grey sources must also pass section 8.

8. Quality of grey sources (Garousi Guideline 11, Table 7)

  • Criteria, tier scale and threshold (10 of 20): as in the evidence method.
  • Tiers as applied here: books, magazines, standards, vendor guidance and reports are tier 1; project documentation and Q/A sites are tier 2; blogs are tier 3.
  • Under the threshold a source is only an example of practice; no claim rests on it.
  • A company page is its owner's claim about itself; the review says so.
  • Books not opened are listed but not scored; no claim rests on them.

9. Extraction and synthesis

  • Extraction and synthesis follow the evidence method; the synthesis runs by question, Q1 to Q10.

10. Limits

  • The limits shared by every review are in the evidence method.
  • One reviewer; no second rater was used.
  • arXiv counts are missing; OpenAlex counts were taken after the search, the same day.
  • No abstract was reached for 11 works (located only, or a summary line). For 18 more it was read in a search summary, not in the source. A number read in such a summary is marked "(secondary)".
  • Forward snowballing was not done.
  • Q/A threads: one scorer.