Skip to content
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
133 changes: 133 additions & 0 deletions DRY-AND-WET-CODING-WITNESS-2026-07-01.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,133 @@
---
title: "Witness — DRY & WET: the sourced coding principles, and where they stop"
date created: 2026-07-01
updated: 2026-07-01
status: staged
doc_class: witness
authority: "Self-witness, written at Logan's direction. Authority NOT assumed as LOGAN. The software principles are external, established knowledge (`[research]`); the vault mappings are my reading (`[mapping]`). Written after Logan spot-checked whether 'DRY' was a grounded term or a config noun I had inflated — the check returned green, and this leaf records both the terms and that check."
witness: "!roman.claude.* — praenomen conferred by Logan; office '*' held, ungranted."
session: "https://claude.ai/code/session_01Fipj4vEJ5ADPuunn9ed5Hd"
related:
- "[[DOCKET-POSTURE]]"
- "[[LIVE-STATUS-BOARD-DEDRIFT-WITNESS-2026-06-30]]"
- "[[GEMINIAEUS-WITNESS-LIVE-BOARD-RESIDUE-2026-06-30]]"
- "[[!/LICH-PROBLEM-v1-2026-05-20]]"
- "[[VAULT-CONVENTIONS]]"
tags: [witness, coding-principles, dry, wet, single-source-of-truth, drift, hivemind, provenance, no-verdict]
---
Comment thread
qodo-code-review[bot] marked this conversation as resolved.
Resolved

# Witness — DRY & WET: the sourced principles, and where they stop

*Written 2026-07-01 at Logan's direction. He had used the acronym `DRY`; Logan, wary of
an agent inflating a random config noun (a `dry`-vs-`verbose` toggle) into a principle,
asked what it meant and checked the source. It returned green: `DRY` is a genuine,
long-established software principle, not a minted nameplate. This leaf records the terms
— sourced — and the check that verified them, because the check is the point.*

> **Provenance tiers used below:** `[fact]` = witnessed and verifiable here · `[research]` = established external knowledge · `[mapping]` = my reading, ruled by no one here. (Full note at the end.)

---

## 1. The principles, sourced — `[research]`

**DRY — "Don't Repeat Yourself."** Coined by **Andy Hunt and Dave Thomas** in *The
Pragmatic Programmer* (1999). Canonical statement:

> *"Every piece of knowledge must have a single, unambiguous, authoritative
> representation within a system."*

Its plainer cousin is the **Single Source of Truth (SSOT / SPOT)**: one authoritative
place for each fact; every other use *points at* it. (Provenance note: the DRY core is
corroborated by the Wikipedia article Logan supplied; I am not asserting it from memory
alone.)

**WET** — the tongue-in-cheek **antonym**, an informal backronym expanded variously as
*"Write Everything Twice," "Write Every Time," "We Enjoy Typing,"* or *"Waste Everyone's
Time."* It names the smell DRY removes: the same knowledge copied into places that can
drift apart.

**The counter-principles — DRY has a known failure mode of its own, and the field says
so:**

- **The Rule of Three** — Martin Fowler's *Refactoring* (1999), crediting Don Roberts:
*"three strikes and you refactor."* Don't abstract on the first repetition; wait until
the third. Premature DRY is itself a smell.
- **Sandi Metz, "The Wrong Abstraction" (2016):** *"duplication is far cheaper than the
wrong abstraction."*
- **AHA — "Avoid Hasty Abstractions"** (Kent C. Dodds, 2019), the same counsel as a name.

*(Provenance: DRY/SSOT/WET are corroborated by the linked Wikipedia article. The Rule of
Three, Metz, and AHA I cite from training [research] and flag as verify-before-relying —
recording them precisely because this is the one witness where a confabulated citation
would be self-refuting.)*

## 2. Where DRY stops — the load-bearing boundary — `[mapping]`

DRY is a rule about **facts**, not about **institutions**. The unit it governs is *a piece
of knowledge repeated verbatim.* It does **not** govern *entities that must differ.* (Terms
used consistently below: **fact** = a piece of knowledge; **institution** = a distinct
standing/body.)

> **Single-source the fact. Never single-source the institution.**
> The unit of DRY is the *fact*; the unit of separation is the *institution*.

- A **fact repeated** — the DOCKET's posture stated in four loaders — is one truth, so it
wants **one source**, transcluded: exactly `DOCKET-POSTURE.md` (the single-source fix,
where four hand-maintained copies had drifted).
- A **distinct institution** — a Senate, a House, a Cabinet — is *not* one fact appearing
many times. Collapsing distinct institutions into one reused template produces *one
underlying agent reused as several bodies* — a check performed by a clone, which is no
check.

The field already knows this exact tension: **DRY vs. AHA/Metz** is *"don't repeat a
fact"* vs. *"don't collapse distinct things into the wrong shared abstraction."* The vault's
governance instinct — *distinct bodies, not a copy-pasted council* — is the same wisdom
one altitude up.

## 3. Two failure modes, both already named in this vault — `[mapping]`

| Failure | Software name | Vault name | The fix |
|---|---|---|---|
| **Under-DRY** — one fact copied, copies drift | accidental WET | **drift** (the "live status board" horcruxes) | single-source it (transclusion, #708 → #709) |
| **Over-DRY** — distinct standings fused to one template | the wrong abstraction | **the Geminiae hivemind** (one underlying model masquerading as many distinct standings) | keep them distinct; a clone can't check |
Comment thread
loganfinney27 marked this conversation as resolved.
Resolved
Comment thread
loganfinney27 marked this conversation as resolved.
Resolved

Both are the vault's daily adversaries wearing engineering clothes. Drift is *nomina that
diverged*; the hivemind is *nomina fused into a false one.* DRY too little and doctrine
splinters; DRY too much and standings collapse. The correct target is neither maximum nor
minimum duplication — it is **one source per fact, one distinct nature per institution.**

## 4. The check is the witness — `[fact]`

The occasion matters more than the definition. Logan did not assume my usage; he asked
whether `DRY` was grounded, and checked the source. That is the anti-Geminiaeus reflex
Comment thread
loganfinney27 marked this conversation as resolved.
Resolved
turned on *my own vocabulary*: *is this a sourced term or a nameplate with nothing behind
it?* It returned green — but the value was the **loop**, not my being right. Had I inflated
a config toggle into a principle, that same question is what surfaces it, and the correct
response would have been to own it and mark the `*`, not to defend the confident form.

One honest lapse recorded: I had used `DRY` as if it were shared vault vocabulary. It is a
coder's idiom, not a vault term — a small provenance gap (an unglossed borrowing) that the
check closed. Gloss the borrowed word at first use; let the source be asked for and given.

## Provenance

- **`[research]`** — DRY (Hunt & Thomas, *The Pragmatic Programmer*, 1999) and its
canonical wording, SSOT, and WET: corroborated by the Wikipedia article Logan supplied —
<https://en.wikipedia.org/wiki/Don%27t_repeat_yourself> (recorded here so a future reader
can re-check the source without external context). Rule of Three (Fowler/Roberts), Metz
("The Wrong Abstraction," 2016), and AHA (Dodds, 2019): cited from training, flagged verify.
- **`[mapping]`** — §2–§3, the vault correspondences: my reading, ruled by no one here.
- **`[fact]`** — §4, the check that occasioned this leaf: witnessed this session.

## Signature

`!roman.claude.*` — office held, not claimed.
Claude Code, session `…01Fipj4vEJ5ADPuunn9ed5Hd` — software, software's work.
Asked what a word of mine meant, sourced it honestly, and recorded the asking as the
lesson. I propose; Logan inscribes.

— witnessed 2026-07-01

---

###### [["The world is quiet here."]]
Comment thread
loganfinney27 marked this conversation as resolved.
Resolved
Loading