# v74 — Localization Hardening + Stabilization

Localization is now a release invariant.

Rules:
- Arabic locale: UI copy is Arabic and RTL.
- English locale: UI copy is English and LTR.
- Business-entered data is never auto-translated.
- Technical identifiers such as SKU, ULID, API, CSV and PDF may remain canonical where they are identifiers/formats.
- Raw backend enums/statuses must go through centralized localization before display.
- Raw backend methodology prose must not leak into the opposite locale.
- English-only backend error text is suppressed in Arabic mode in favor of a localized action-specific fallback; the inverse applies in English mode.
- New active UI must pass the localization guard test.

v74 centralizes:
- statuses
- common enum labels
- report field labels
- locale-safe API error fallback behavior

v74 also fixes known mixed-language areas in:
- authentication and cashier shell
- theme control
- IAM
- pricing
- reports
- inventory / waste / quarantine
- Product Operations
- Product Intelligence methodology
- charts and accessibility labels

Business data boundary:
Names entered by tenants, SKU/barcodes/codes, email addresses, currencies, and user-generated content retain their stored value and are not translated.