Status labels: ■ Built in the repository (checked files or tested library code; not a product you can run yet) ◧ Designed in a document, no code yet ○ Planned: listed in the catalog, not built ? Not decided △ Not in the catalog yet (a gap) – No claim made
What goes wrong with “Local-first data” in a shop, in particular regarding lost device?
◧ Designed C0184
Data lives on the shop's own devices by default; nothing requires a cloud account. This is a design principle in the README. Plan for: lost device; shared device; backup location. It is designed in a document and no code exists.
Open this answer on its own page
How risky is “No telemetry” to get wrong, especially around crash reports?
◧ Designed C0185
The README states no telemetry or data collection. Code-level guards exist in the repository (no network crates, no demo keys, no remote assets, no default network binding), but they cannot prove a future app. The launcher's one disclosed opening request is the stated exception and carries no account data. Plan for: crash reports; the opening request reveals a network address; optional features that need a network. It is designed in a document and no code exists.
Open this answer on its own page
Which edge cases come up around “Roles and permissions”, especially around shared logins?
○ Planned C0186
Giving cashiers, supervisors and owners different rights, and limiting risky actions. Plan for: shared logins; temporary permissions; owner lockout. In the catalog it is a plan; no screen or product is built.
Open this answer on its own page
What should I plan for with “Audit trail”, in particular regarding clock changes?
○ Planned C0187
An append-only record of edits, voids, discounts and logins is planned. Separately, a tamper-evident, hash-chained log of removal decisions exists as tested code: it is local, sent to nobody, and detects edited or deleted entries, but someone who rewrites the whole chain could change it. Plan for: clock changes; log size; deleting the log. In the catalog it is a plan; no screen or product is built.
Open this answer on its own page
Is “Encryption at rest” supported today, considering key storage?
? Not decided C0188
Protecting the database file on a lost device. The core has no encryption at rest today; whether to adopt SQLCipher is an open decision that weighs key loss against theft. Plan for: key storage; performance; recovery if a key is lost. It is not decided.
Open this answer on its own page
How does Simca approach “Backups”, with a focus on untested backups?
○ Planned C0189
Local backups to drives or another device, with restore tests. Plan for: untested backups; backup of a corrupt file; offsite copy. In the catalog it is a plan; no screen or product is built.
Open this answer on its own page
Is “Restore” supported today, considering version mismatch?
○ Planned C0190
Rebuilding a device from a backup, checking integrity first. Plan for: version mismatch; partial restore; restoring over newer data. In the catalog it is a plan; no screen or product is built.
Open this answer on its own page
Is “Device pairing” supported today, in particular regarding unknown device?
○ Planned C0191
Trusting a second device on the shop network. Plan for: unknown device; revoking a lost phone; time sync. In the catalog it is a plan; no screen or product is built.
Open this answer on its own page
Where does “Signed files” fit in Simca, with a focus on key ceremony not done?
■ Built (repo) C0192
Everything the launcher installs is signed: packages, update manifests, entitlements, the catalog snapshot and price documents. A missing signature or an empty checksum is an error, never skipped. Ed25519 is used in the packaging code with test keys only; no production key exists in the repository and the key ceremony is not done. A revocation list and a trusted key set exist as code. Plan for: key ceremony not done; key loss; revocation travels with launcher updates. It exists only as a file in the repository, not as a product feature.
Open this answer on its own page
What goes wrong with “Data minimisation” in a shop, especially around customer id thresholds?
○ Planned C0193
Only collecting what the sale needs, with customers' details optional. Plan for: customer id thresholds; export requests; retention. In the catalog it is a plan; no screen or product is built.
Open this answer on its own page
Where does “PINs and sessions” fit in Simca, in particular regarding shoulder surfing?
○ Planned C0194
Quick staff logins that time out safely. Plan for: shoulder surfing; shared pin; lock on idle. In the catalog it is a plan; no screen or product is built.
Open this answer on its own page
Why does “Secure updates” matter for a small business, with a focus on failed update?
■ Built (repo) C0195
Updating without a mandatory internet connection, with rollback. In the launcher library: a signed manifest with expiry and a rollback-protection sequence, a mandatory checksum, a list of blocked versions, staged download then rename, successor files (the running program is never overwritten), real rollback to a previous generation and automatic rollback after a failed first start, and install from a signed file. Tested with a mock fetcher; the HTTP fetcher and screens are not built. Plan for: failed update; offline update media; no http fetcher yet. It exists only as a file in the repository, not as a product feature.
Open this answer on its own page
Why does “Software supply chain” matter for a small business, considering dependency vulnerabilities?
◧ Designed C0196
Building and signing releases and checking dependencies. A process plan in the repo, not a result. Plan for: dependency vulnerabilities; reproducible builds; signing keys. It is designed in a document and no code exists.
Open this answer on its own page
What should owners check about “Security model and threat model”, in particular regarding external review not done?
◧ Designed C0197
A written security model exists in the repository: assets, attackers, eight rules (true removal, signed everything, local authority for offline use only, online features decided by the host, no hidden doors, staging discipline, break-in tests, an honest limit) and a table of what is built and what is not. A full threat model, an external security review and the host design are planned. Plan for: external review not done; host not built; a person who holds the code can attack the code. It is designed in a document and no code exists.
Open this answer on its own page
What should I plan for with “True removal of unselected elements”, with a focus on a removed element's data is a separate choice?
■ Built (repo) C0198
A package holds nothing for elements outside its spec: no code, assets, migrations or data tables, and hiding by a flag does not count. A packaging command proves the absence of files, menu entries, permissions and database tables for a profile, and the core tests check the database side. Plan for: a removed element's data is a separate choice; proven per profile, not for every possible spec; installer behaviour not tested. It exists only as a file in the repository, not as a product feature.
Open this answer on its own page
Why does “No hidden doors” matter for a small business, especially around applies to the code that exists?
■ Built (repo) C0199
Rules enforced by tests: no demo or default keys, no default network binding, no runtime CDN or remote asset loads, no founder or vendor accounts, no auto-start, and no hidden start path. Source scans run in all four workspaces. Plan for: applies to the code that exists; the flutter interface is not built; a person with the code can attack the code. It exists only as a file in the repository, not as a product feature.
Open this answer on its own page
Which edge cases come up around “Staging discipline”, thinking about rollout tooling not built?
◧ Designed C0200
Release rules from the security model: snapshots first, data never moves with code, a backup before any data-touching change, test on a copy, one stage per release, the previous build kept for real rollback, and ledger and customer data last. Parts exist as code (test on a copy, verified backup before removal); the rollout tooling does not. Plan for: rollout tooling not built; installer that leaves the data folder alone not proven; sensitive modules last. It is designed in a document and no code exists.
Open this answer on its own page
What goes wrong with “Removal audit chain” in a shop, considering someone who recomputes the whole chain can rewrite history?
■ Built (repo) C0201
Each removal decision is written to a local log in which every entry commits to the previous one, and the current state is cross-checked against the chain. It is tamper-evident, not tamper-proof; publishing the head hash outside the computer (for example in a printed or exported record) is a planned mitigation. Plan for: someone who recomputes the whole chain can rewrite history; no trusted time offline; local only. It exists only as a file in the repository, not as a product feature.
Open this answer on its own page