# Beta Reader Protocol
Status: ready to run after the editorial branch is merged or shared as a review branch
Version under review: PostgreSQL 18 edition
Canonical source: `Manuscript/`
## Purpose
The beta is not a final proofread. It tests three different promises:
1. Primary readers can build a physical and operational model of PostgreSQL without already being database specialists.
2. PostgreSQL experts can trust the literal mechanism beneath the metaphors.
3. A reader can start the companion environment, complete the core labs, and recover from mistakes without undocumented help.
Do not combine these audiences into one undifferentiated score. A passage can be delightful to a primary reader and still be technically unsafe; it can also be impeccable to an expert and exhausting to the intended audience.
## Cohorts
Recruit 8–12 readers in each cohort. A useful minimum viable beta is five completed readers per cohort, but publication confidence should be marked lower when a cohort falls below eight.
### A. Primary audience
Backend or application engineers who:
- can write joins, aggregates, and basic DDL;
- have used PostgreSQL in an application;
- do not already identify as PostgreSQL internals or performance experts.
Primary question: does the book create durable capability without turning into a reference manual on the main road?
### B. Expert red-team
Include a mix of PostgreSQL DBAs, performance engineers, contributors, extension authors, and production operators. Ensure that the cohort contains at least:
- one storage/MVCC/WAL reviewer;
- one planner/executor reviewer;
- one HA/replication reviewer;
- one security/RLS reviewer.
Primary question: does any vivid simplification teach a false operational model?
### C. Teaching and usability
Self-directed learners, technical educators, workshop authors, or documentation specialists.
Primary question: can a reader predict, observe, explain, repair, and generalize without hidden setup knowledge or prematurely revealed answers?
## Review package
Give every participant the same package:
- the exact branch or immutable commit under review;
- `Manuscript/00 - Introduction/Index.md` as the entry point;
- the PostgreSQL 18 version contract and environment verification command;
- the chapter assignment and expected completion date;
- this protocol and the feedback template below;
- a private channel for security-sensitive or potentially embarrassing technical defects.
Do not silently update the review branch while a cohort is reading it. If a P0 correction must land, record the old and new commit and tell affected readers which version they used.
## Chapter assignment strategy
Every chapter needs at least:
- three primary-audience completions;
- two expert completions in the relevant subsystem;
- two setup/usability completions for any chapter containing core labs.
Oversample the highest-risk chapters:
| Chapter | Extra review needed |
| --- | --- |
| 2 — Physical Storage & MVCC | storage/MVCC expert and primary-reader metaphor recall |
| 4 — Query Planning & Execution | planner expert and pacing/drop-off observation |
| 5 — Durability & Transactions | WAL/recovery expert and two-session lab verification |
| 6 — Resource Management & Processes | process/memory/cache expert |
| 7 — Wait Events & Concurrency | production diagnostician and optional-tool usability |
| 8 — Scaling & Architectural Coordination | HA/replication reviewer and vendor-claim audit |
| 9 — Identity & Access Control | PostgreSQL security/RLS reviewer |
| 10 — Summary & Epilogue | readers who completed at least four earlier chapters |
## Per-chapter feedback form
Collect these answers immediately after each assigned chapter:
1. Reader cohort and relevant experience.
2. Commit reviewed.
3. Chapter and time spent.
4. Capability gain, 1–7: “How much more capable do you feel after this chapter?”
5. First confusion: “What was the first point where your mental model became uncertain?”
6. Trust break: “Did any sentence make you doubt the book's technical authority? Quote or link it.”
7. Removal: “What would you delete, shorten, or move to the field guide?”
8. Lab result: completed without help / completed with help / could not complete / no core lab.
9. Undocumented prerequisite, if any.
10. Prediction quality: “Before the explanation, could you make a falsifiable prediction?”
11. Transfer: “Name one unfamiliar PostgreSQL situation you can now reason about.”
12. Metaphor boundary: “Which metaphor was most useful, and where could it mislead?”
13. Defects with file path, heading, quoted phrase, expected model, and evidence.
## End-of-book feedback form
Ask only readers who completed the agreed route:
1. Recommendation score, 0–10.
2. “What single thing prevents you from giving this book a 10?”
3. “Describe PostgreSQL's central design tendency one week after finishing, without reopening the book.”
4. “Walk through a read from connection to returned rows.”
5. “Walk through a write from statement to commit acknowledgement and crash recovery.”
6. “A query is slow. What evidence distinguishes working, waiting, and blocking?”
7. “Which chapter felt most like a manual rather than a narrative, and where did that shift begin?”
8. “Which idea will change something you do at work?”
9. “Who is this book not for?”
## Defect severity and triage
### P0 — stop-the-line
Use P0 when feedback identifies:
- a false production mechanism;
- unsafe operational or security advice;
- a contradiction in MVCC, WAL, process, cache, replication, or visibility models;
- a core lab that fails from the declared clean PostgreSQL 18 environment;
- a version or vendor claim presented as universal when it is not.
Pause new review assignments for the affected material. Reproduce the issue, cite a primary source, correct the literal model, scan for duplicates, add a regression check where possible, and notify readers who saw the old text.
### P1 — trust, learning, or reproducibility blocker
Examples: misleading boundary, obsolete non-core behavior, an optional tool that looks mandatory, serious pacing friction, or an undocumented prerequisite.
P1 issues must be resolved or explicitly accepted with a publication-impact note before release.
### P2 — polish
Examples: repetition, navigation, terminology, typo, awkward transition, or optional elaboration. Batch these after P0/P1 decisions so the team does not polish text that will move or disappear.
## Weekly synthesis
Maintain one row per finding with:
| Field | Meaning |
| --- | --- |
| ID | stable `BETA-###` identifier |
| Cohort | primary / expert / teaching |
| Source | reader and review commit |
| Location | canonical Markdown file and heading |
| Severity | P0 / P1 / P2 |
| Observation | what the reader experienced or disputed |
| Evidence | lab output, primary source, or repeated-reader pattern |
| Decision | fix / move / clarify / accept / defer |
| Owner | person responsible for resolution |
| Regression guard | test, validator rule, or explicit reason none is possible |
| Status | open / verifying / closed |
Summarize patterns by cohort. Three independent readers stumbling at the same transition is a structural signal even when no sentence is factually wrong.
## Release decision
The Markdown edition is ready for publication authoring only when:
- no known P0 issue remains;
- all core PostgreSQL 18 lab checks pass from a clean checkout;
- expert disagreement concerns nuance or editorial choice rather than basic mechanics;
- Chapters 4 and 7 do not produce a recurring narrative-to-manual drop-off;
- primary readers can reconstruct the central mental model a week later;
- at least one security reviewer accepts the Chapter 9 threat boundaries;
- at least one HA reviewer accepts the Chapter 8 failure and replication boundaries;
- the most common answer to “what prevents a 10?” is polish or desired additional scope, not trust or reproducibility.
Record the final decision as **release**, **release with named follow-ups**, or **hold**. Include the reviewed commit, cohort completion counts, all accepted P1 risks, and the exact PostgreSQL/extension/vendor version contract.