---
status: active
section: docs/plan
created: 2026-11-21
also: [development/standards/plan.md, development/wiki/work/measures.md]
---

<a id="conflicts"></a>
# Cross-plan conflicts

> Source: `development/wiki/archive/old/plans/conflicts.md` (archived 2026-11-21).
> One table, one home; measures links here.

> Cross-plan tensions identified from interview decisions (2026-07-23).
> **Living document — partly open:** the resolutions below hold, and the
> “New conflicts identified” section plus the S10 contract test remain open.
> Updated 2026-07-23 with D01-D23 strategic decisions (Central URLs primary, staging workflow, Railway monorepo, no TTL, HTTP status codes, error detail field).

## Hard conflicts

### H1: Tabbed app (Q2) × Masks F2 (Q4)

**Files**: `../browsy/spec.md`
**Status**: Resolved — global mask bar above tabs

The tabbed layout has no natural place for mask switching. Masks affect every tab's content.

**Resolution**: Global mask selector bar above the tab strip. Adds ~40px vertical space. All tabs read current mask from shared state.

### H2: Weighted ranking (Q6) × Self-selected curators (Q7)

**Files**: `../browsy/spec.md`
**Status**: Resolved — curator bonus with decay

Trust-weighted ranking means higher-stage users have more influence. But curators are self-selected with community veto.

**Resolution**: Curator status grants ×3 weight bonus that decays 10%/week. Full resolution requires the trust layer to work first.

### H3: Multi-repo (SD01) × Full test suite (SD13)

**Files**: `standards.md`
**Status**: Resolved — shared CI infrastructure across repos

**Resolution**: Shared test runner script invoked from any repo. CI config duplicated per repo with shared workflow template.

### H4: PR merge deploy (C70) × central/ has both backend + apps (C72)

**Files**: `central-backend.md`, `plans/index.md`
**Status**: Resolved — directory split ✅ Implemented 2026-08-01

The `central/` branch contains both the FastAPI backend (needs Railway deploy) and modified app versions (need flove.org deploy, same as main). A PR could change both, but the deploy pipeline is PR→Railway for backend. App changes in `central/` would need a separate deploy path.

**Resolution**: Split `central/` into three directories:
- `addons/backend/` — FastAPI + Turso, deployed by Railway (path filter in `railway.toml`)
- `central/apps/` — Central app HTML files, deployed to flove.org via GitHub Pages
- `addons/` — Shared libs (`libs/`), read-only for both pipelines

Railway CI watches `backend/` only. GitHub Pages CI watches `apps/` only. A single PR can touch both, but deploy pipelines are independent. `shared/` is bundled into both at deploy time.

## Soft tensions

### S1: Hybrid abuse (Q8) × Deferred infrastructure

**Files**: `../browsy/spec.md`
**Status**: Resolved — Tier 1 defined

Tier 1 = self-selection + community veto + simple report form. Tier 2 (jury) deferred.

### S2: Profily-integrated MyNet (Q9) × Tabbed app (Q2)

**Files**: `../browsy/spec.md`
**Status**: Accepted — discoverability trade-off

MyNet tab is prominent (tab #3 of 5). Social features are the headline content.

### S3: Masks F2 (Q4) × Deferred crypto recovery (Q5)

**Files**: `../browsy/spec.md`
**Status**: Resolved — recovery code v1

Recovery code displayed once at mask creation, user writes down. Lost code = new mask only.

### S4: Full offline (SD04) × Privacy-first analytics (SD02)

**Files**: `standards.md`
**Status**: Accepted — analytics degraded offline

Plausible needs network. When offline, analytics don't fire.

### S5: Anonymous device ID (C51) × data permanence

**Files**: `central-backend.md`
**Status**: Accepted — user must claim data before clearing browser

Auto-sync uses a device-generated UUID. If the user clears browser data (localStorage, OPFS), the UUID is lost and server data becomes orphaned. The claim flow (email) fixes this, but only if the user claims before clearing data.

**Resolution**: Document in Central modal: "Your data is linked to this device. Set an email to keep it if you clear your browser data."

## Coherent clusters

### C1: "Decide Now" (Q6 + Q7 + Q8)

All three overrode "Defer" suggestions. Tightly coupled: ranking needs curators, curators need abuse system.

### C2: "Privacy-First" (Q4 + Q9)

Early masks + integrated social = privacy-by-design in the social layer.

### C3: "Simple Wins" (Q2 + Q10)

Tabbed app (simple) + defer vizy XR/VR (don't add complexity).

### C4: "libSQL Everywhere" (C64 + C65 + C66 + CA37)

Same engine across SoloRich, Central, Decentral, DecentralRich. One API (`@libsql/client`), one schema, trivial SoloRich→Central migration (change connection string from local file to remote).

## Resolved this session

| Conflict | Resolution |
|----------|------------|
| H2: ranking × curators | ×3 bonus, 10%/week decay |
| H4: deploy pipeline ambiguity | Split `central/` into `backend/` + `apps/` + `shared/` |
| S1: abuse × deferred | Tier 1 = self-selection + veto + report form |
| S3: masks × recovery | Recovery code, lost = new mask |
| C16: milestone scope | JSON distro + trusty cleanup + app links |
| C07: deep link direction | One-way, trusty → maty/evily |
| C14: deep link params | `?from={app}&item={id}` |
| C11: deep link mechanism | URL params |
| C26: deep link HTML | Separate div below item |
| C04: MongoDB × SQLite/Turso | No MongoDB — Turso for serverless SQLite, PostgreSQL for CentralRich |
| C15: central/solo branch name | Renamed to `central/` |
| C18: local persistence | libSQL via `@libsql/client` (sql.js WASM + OPFS), fallback localStorage |
| D07: central/apps vs main/apps | Central apps are their own, share names only — two different things |
| D10: Central app serving | Same origin (`flove.org/api/*` proxy), no cross-origin issues |
| D20: URL hierarchy | Central URLs become primary; Solo URLs on main maintained as fallback |
| NEW: Browsy = main nav | Browsy replaces Central nav, gets HTML from Central or solo fallback |
| NEW: Central = data bank | Central enriches browsy navigations + frontend apps |
| NEW: Trust visibility | Trust acceptances need visibility parameter: first circle, second circle, social groups, public |
| NEW: Webhook exports | Generic webhooks, manual trigger, user-configurable auth |
| NEW: CRDT conflict resolution | Replace last-write-wins with CRDT-style merge for trust data |
| NEW: Unified browsy storage | Extension storage for all browsy data (trust, profiles, webhooks, panel state) |
| NEW: Trust sync endpoint | Separate `POST /api/trust/sync` — not batched with profile/app data |
| NEW: Per-app schema validation | Validate `data` against `collect-schemas.json` on server, reject malformed |
| NEW: Versioned ping | `/api/ping` returns version + feature flags for browsy feature detection |
| H7: Q16 (6 groups) × Q19 (4 groups) | 4 groups wins, Q16 removed |
| S7: Delegation in trust logic × v1 scope | Delegation is v2, direct vouches only in v1 |
| S8: "Closest" × "Close" naming | "Close" wins, the old browser-extension plan historical |
| S9: Friends × Close group | Same thing — friends list = Close group member list |

### S10: Per-app schema validation × undefined app datafields

**Files**: [`central-backend.md`](../archive/old/plans/central-backend.md), [`index.md`](../archive/old/plans/index.md)
**Status**: Unblocked 2026-11-15 — schemas exist; the contract test is still open.

Per-app schema validation was blocked until each app declared its datafields
(questy Q010). Every app with a source now declares its `flove-fields/1` tree in
`tools/routing.json` (`extract-fields.py --check` passes, 0 `partial`), so PEN001
is done. The dependency now runs **PEN001 → PEN005**: schemas enable the
`collect()` contract test; the test itself is not written yet.

**Resolution**: the datafields came first; validating `collect()` against them
is the remaining step in [`../pendings.md`](../archive/old/pendings.md) §Questy Session
2026-07-29. Work filter:
[docs · work](https://flove.org/development/index.html#work?tab=docs&view=r2&subtab=work&category=docs&section=work).

## New conflicts identified

### H5: Browsy main nav × Central nav

**Files**: `../browsy/spec.md`, `central-backend.md`
**Status**: Resolved — Browsy replaces Central nav

Browsy is now the main navigation app. Central apps get their nav from browsy.

**Resolution**: Browsy loads HTML from Central (or solo fallback). Updates via addons.

### H6: Offline trust × Central enrichment

**Files**: `../browsy/spec.md`, `central-backend.md`
**Status**: Resolved — Hybrid mode

Web of trust can be fully offline. Central provides optional enrichment.

**Resolution**: Browsy computes trust locally. Central can provide additional data when online.

### S6: Webhook exports × Privacy

**Files**: `../browsy/spec.md`, `central-backend.md`
**Status**: Resolved — User-configurable

Exports are manual, user controls what gets sent where.

**Resolution**: Manual trigger, user-configurable auth, visibility parameter for trust data.

### H7: Q16 (6 groups) × Q19 (4 groups)

**Files**: `../browsy/spec.md`
**Status**: Resolved — 4 groups wins

Q16 said 6 groups (3 defaults + 3 custom). Q19 said 4 groups (Close, Public, Private + 1 custom). Later decision supersedes earlier.

**Resolution**: 4 groups total, flat, no nesting. Q16 removed from decisions table. Q19 is authoritative.

### S7: Delegation in trust logic × v1 scope

**Files**: `../browsy/spec.md`, `../browsy/spec.md`
**Status**: Resolved — Delegation is v2

Trust logic spec included delegation as a core feature. Browsy plan says v1 = direct vouches only, no chains.

**Resolution**: Delegation, chain depth, key rotation all marked v2 in trust logic spec. v1 is direct vouches only. Friends list = Close group.

### S8: "Closest" naming × "Close" naming

**Files**: `../browsy/spec.md`, `../browsy/spec.md`, `../browsy/spec.md`
**Status**: Resolved — "Close" wins

the old browser-extension plan uses "Closest" (Q130-Q145, 8 references). Keys spec and trust logic spec use "Close". Later specs are authoritative.

**Resolution**: "Close" is the standard name. the old browser-extension plan references are historical, not updated.

### S9: Friends concept × Close group

**Files**: `../browsy/spec.md`, `../browsy/spec.md`
**Status**: Resolved — They're the same thing

"Friends" is the human term. "Close group" is the technical term. A friend = someone with a direct vouch relationship. The friends list IS the Close group member list.

**Resolution**: Friends section added to both specs. Friends = Close group, no separate data structure.

### H5b: Shared code loader (CA01-CA42) × self-contained apps

**Files**: `../../frontend/frontend.md`, `adoption.md`, `contract.md`
**Status**: Resolved — inlining wins (2026-11-15)

The shared-code plan wanted apps to load `flove.css` / `flove.js` /
`flove-appy.js` from a shared folder behind a browsy-gift gate, which made an app
depend on a runtime fetch and on a tier. The trunk rule is one self-contained
file per app.

**Resolution**: every app inlines the copy it uses (`floveStore`/`floveFields`/
`flovePort` everywhere; `flove.js`, `flove-appy.js`, `flove-theme.js`,
`flove.css` where needed) and `addons/code/` is a reference set shipped
in the offline package, never loaded. No app fetches shared **code**; the only
external files are the frozen vendor fonts/icons (`/addons/vendor/…`)
and Freed's local `vendor/leaflet/leaflet.{css,js}`. The tier gate and the
enrichment loader are retired.

### H6b: Tier model (§13.1) × trunk scope

**Files**: `2026-11-14-app-tier-planned.md`, `debates/proposals/frontend-tiers.md`, skills
**Status**: Resolved — tiers are demo-only (2026-11-15)

The tier model drives the blogy/dealy/keys/appy sketches but was never wanted in
this trunk.

**Resolution**: no app carries a tier attribute, switcher, gate or field. The
mini → super variants stay as demo sketches; §13.1 and the skills mark that
scope explicitly.

### CR26: Authy L3 — Telegram presence (2026-11-17 plan) × Telegram verified (pre-build spec)

**Files**: `2026-11-17-appy-authy-mynet.md`, `../browsy/pre-build/02-authy-levels.md`
**Status**: Resolved 2026-11-20 — the pre-build ladder owns L3; phone and bio stay as L4/L5

The 2026-11-17 plan put **presence** at L3 (bare bot link, risa handoff
included) and **verified** (nonce + DM) at L4, with phone L5 and bio L6. The
pre-build spec `02-authy-levels.md` makes **verified** (risa signed claim or
nonce + DM) the L3 MyNet and publish gate, keeps a bare link at L1, and adds
optional email L2 and phone L4.

**Resolution**: the owner keeps **L3 = Telegram verified**; publishing moves to
L3 with it. Phone stays **L4** and bio **L5** from the plan, with no new content
surface; L6 is retired. Both sources agreed that a bare link opens nothing.
The owner ladder is `../browsy/spec.md` §3.3; the plan and all Work rows now
follow it.
