jcoffey-dev
95dcb96086
The groundwork in #145 gave the app a language to serve. This gives it something to serve, and a way to measure the distance to the languages actually planned. The English text is the key. `t("Archive")` looks "Archive" up and returns the English when it is not there, which buys three things worth more than tidy symbolic keys: no English catalogue to keep in step with the code, a missing translation that degrades to readable English rather than to `mail.list.archive`, and an extraction step that is wrapping a string rather than inventing a name for it. Names are where extraction stalls, and 55 components is a lot of small naming arguments. The cost is that editing English copy orphans its translations, which is the right way round: the copy is the product, and a stale German sentence should fall back to the new English. `plural()` takes forms rather than (one, other), because two forms is an English assumption that does not survive phase two of the plan. Russian and Ukrainian need three, and choosing between them is not a question about the number 1. Intl.PluralRules knows the rule for every language the browser knows, so the catalogue supplies the forms and the runtime picks; a category the catalogue does not carry falls back to `other` rather than rendering undefined. Interpolation is named rather than positional for the same reason -- German moves the parts of a sentence around and means the same thing. Catalogues are dynamically imported, so a reader who never leaves English never downloads one, and English needs no fetch at all. `applyLang` sets the lang attribute before kicking the load, deliberately: lang is what stops Chrome offering to translate and should not wait on a network request to say something it already knows. `t()` is a plain function, not a hook, so the tree is keyed on a language version at the root and thrown away when the catalogue changes. Making every call site a subscriber would turn extracting a string from "wrap it" into "wrap it and add a hook", for an event that happens about once per account. NotificationsSettings is extracted end to end as the reference -- it covers all four shapes, being JSX text, translated attributes, a toast, and a sentence with a value interpolated into it. scripts/i18n-coverage.mjs counts what is left, because ~1,000 strings across 56 files is too many to eyeball in review or carry in anyone's head. It reports 20 wrapped and 925 remaining, and it deliberately does not count punctuation and separators as untranslated -- a floor no amount of work could reach would make the number useless. A progress report rather than a gate: --check exits non-zero, for once the number is low enough for that to mean something. ROADMAP.md said translations were "English-only for now" on a page whose stated purpose is things the answer is "no" to. It now says what is actually happening, carries the phase order, and says why Arabic, Hebrew and Persian are on neither list: RTL is a layout and bidi problem rather than a longer catalogue, and shipping it as though it were the same kind of work is how an RTL build ends up unusable with nobody saying so.
ihasmail
Immutable webmail for Stalwart Mail Server — a container with nothing to persist, and a Gmail-class client on top of it.
Mail, calendars, contacts, files and filters in a responsive single-page app
that works equally well on a desktop monitor and a phone. It talks only JMAP
(plus Stalwart's blob/upload/EventSource endpoints) — no IMAP, no SMTP, no
database, and with IMMUTABLE=1 no writable filesystem either. Everything
durable belongs to Stalwart; the container is disposable.
| 🌐 ihasmail.org | What it is, what it looks like, the full feature list |
| 📘 docs.ihasmail.org | Installing · Configuring · Using it · Shortcuts · Rebranding · Troubleshooting |
| 📋 FEATURES.md | Everything it does, feature by feature, with the capability each one needs |
| 🧪 KNOWN-ISSUES.md | What was verified live, and where Stalwart departs from a spec |
| 🛣 ROADMAP.md | What ihasmail does not do, and why |
This file is for people working on ihasmail. Everything about running it lives in the docs.
Screenshots
Taken against the built-in mock server (npm run dev:mock) with sample data — no real mailbox involved.
Inbox & conversation (dark)  |
Inbox & conversation (light)  |
Composer  |
Calendar  |
Contacts  |
Sieve filter builder  |
More, including the mobile layout, on ihasmail.org.
What's in it
- Mail — three-pane Gmail-style layout, conversation view, virtualised list, labels, undo, Gmail search operators and keyboard shortcuts, Sieve rules from a message's context menu, sanitised HTML with remote images blocked, read receipts, invitations and RSVP, multi-composer rich-text editing with signatures, scheduled send and undo send
- Calendar — JMAP Calendars / JSCalendar: month/week/day/agenda, recurrence, attendees and free-busy, colour categories
- Contacts — JMAP Contacts / JSContact: address books, groups, full editor, vCard import/export
- Files — JMAP FileNode: browse, upload, download, rename, move, delete
- Settings that follow the account, not the browser — kept in a
settings.jsonin the account's own JMAP Files, so ihasmail itself stays stateless - Runs read-only — one optional write path, and with it switched off the container needs no volume and no writable root.
IMMUTABLE=1is checked at startup rather than trusted, so a half-applied switch refuses to boot instead of failing quietly. See Running immutably - On a phone — swipe a message to archive or delete it (either direction, your choice), hold one to select it, hold a folder for its menu, pull the list to refresh, swipe back from a conversation
- Platform — installable PWA, Web Push with ihasmail closed,
mailto:handler, no credentials in the browser, strict CSP, SSRF-safe image proxy
The long version is on ihasmail.org; how to drive each one is in Using ihasmail.
Requires Stalwart 0.16 or newer
Sign-in refuses anything older, by name. 0.16 replaced the REST management API
with JMAP registry objects, changed the shape of FileNode, split its rights up
and moved configuration into the store; supporting both generations meant a
wrong guess had somewhere to fall back to, so it failed quietly — and that
reached production. With one supported generation a wrong guess is a loud error
on the first call.
- Still on 0.15? The last release that runs on it is tagged
stalwart-0.15-support. - Upgrading? stalwart-migrator does it in place, checkpointing every phase and validating afterwards. The live instance moved 0.15.5 → 0.16.19 with eight seconds of downtime and nothing lost.
Quick start (Docker)
cp .env.example .env
# edit: STALWART_URL=https://mail.example.com and APP_SECRET=$(openssl rand -base64 48)
docker compose up --build -d
# → http://localhost:8080 (put Caddy/nginx in front for TLS; see Caddyfile.example / nginx.example.conf)
Users sign in with their Stalwart mailbox credentials. An account with two-factor authentication needs an app password, created in Stalwart's own settings — Stalwart accepts a TOTP code only through an OAuth flow and offers no password grant, so no client holding a username and password can exchange them plus a code for a token.
Full instructions, TLS, and every environment variable: Installing · Configuring.
Running immutably
The server writes to exactly one path, the optional SESSION_FILE. Clear it
and there is nothing left to write, so the container can run with no writable
filesystem at all:
docker run --read-only --tmpfs /tmp -e IMMUTABLE=1 -e SESSION_FILE= ...
IMMUTABLE=1 is an assertion the server checks at startup rather than a switch
that changes what it does: it refuses to start if SESSION_FILE is still set,
or if the filesystem it is installed on turns out to be writable after all.
Without it the same misconfiguration is silent — sessions are held in memory
and persisting them is best-effort, so a read-only /data costs one warning at
the first sign-in and nothing else until the instance is replaced and everyone
is signed out.
That sign-out is the standing cost of this mode today, since sessions have nowhere to live across a restart. Removing it means moving the session upstream into a token Stalwart itself issues and can revoke, which is what the OAuth work in ROADMAP.md is for.
Architecture
browser ──(same-origin /api/*)──► ihasmail server (Node + Hono) ──(JMAP over HTTPS)──► Stalwart
React SPA • session cookie ⇄ Basic auth
JMAP client + stores • /api/jmap, /api/blob, /api/upload, /api/events (SSE), /api/image
web/— Vite + React 19 + TypeScript SPA.src/jmap(client, push, types),src/store(zustand: session, mail, compose, contacts, calendar, files, sieve, settings),src/views,src/lib(sanitiser, search parser, Sieve codec, locale-aware dates, vCard, …).server/— Node/Hono backend: authenticates against Stalwart's JMAP session endpoint, seals the credentials with a key derived from the cookie secret, proxies JMAP/blob/SSE, serves the SPA under a strict CSP.src/mock/is an in-memory fake Stalwart for development and demos.
Capabilities used: core, mail, submission, vacationresponse, sieve,
contacts(+parse), calendars(+parse), principals(+availability),
quota, blob, filenode, EventSource push, plus Stalwart's own
urn:stalwart:jmap (read-only). Features degrade gracefully when one is
missing.
Development
Requirements: Node ≥ 20.10 (22 recommended), npm ≥ 10.
npm install
npm run dev # real Stalwart (STALWART_URL in .env) — server :8080, Vite :5173
npm run dev:mock # built-in mock Stalwart ([email protected] / demo), mock on :8788
npm run dev:mock:no-future-release # mock that advertises FUTURERELEASE and drops every hold
npm run typecheck # tsc for both packages
npm test # vitest (web) + node:test (server)
npm run build # web/dist + server/dist
npm start # serve the production build
Open http://localhost:5173 in dev, or http://localhost:8080 for the production build. Running it for real is covered in Installing and Configuring.
The mock
An in-memory fake Stalwart 0.16 — enough JMAP to develop and demo against
without a real mailbox. It reproduces the things a naive fake would get wrong,
because each cost a live debugging session: urn:stalwart:jmap advertised
per-account rather than session-level, identity signatures capped at 2047
bytes, and CalendarEvent/set speaking Stalwart's vocabulary rather than
RFC 8984's. Two switches: MOCK_NO_FUTURE_RELEASE=1 advertises FUTURERELEASE
and then drops every hold; MOCK_NO_REGISTRY=1 omits the Stalwart capability so
the sign-in refusal can be tested.
Version numbers
ihasmail v2026.8.30+pr129 — the date of the commit this was built from, and
the pull request that commit arrived through. A commit that did not arrive
through one carries its short SHA instead: 2026.8.30+g1fa6578. It all comes
from git at build time; nothing writes a version into the tree, and
package.json sits at 0.0.0 because it is no longer the source of anything.
The date is the commit's own rather than today's, so rebuilding an old commit gives the version it had the first time.
node scripts/version.mjs # the version for the current checkout
docker build --build-arg IHASMAIL_VERSION="$(node scripts/version.mjs)" -t ihasmail:2026.8.30 .
.dockerignore excludes .git deliberately, so an image build cannot work this
out for itself — pass it in. Left out, the build reports 0.0.0, which is meant
to look wrong: a version with no +pr or +g means whoever built the image did
not pass one.
The version says nothing about Stalwart, deliberately. It used to: 2.16.x had
16 for the 0.16 generation it targeted, which leaves nowhere to go once
Stalwart reaches 1.0 — 2.1 sorts below the 2.16 already deployed, so every
image and About screen would read as a downgrade. Which Stalwart a build needs is
stated where it can be precise, in the badge at the top of this file and in
KNOWN-ISSUES.md, rather than compressed into one digit.
The pull request lives after the +, as build metadata, because it is
provenance rather than a rank: at the rate they merge here it climbs without
bound and says nothing about how new a build is. Everything after the + is
ignored when versions are compared, which is the right reading — two builds from
the same day differ in where they came from, not in age. Nothing here depends on
that comparison: images are pruned oldest-first by creation time, and a rollback
names a git ref.
Deploying
deploy.example.sh is a single-host Docker deploy: it
fetches, refuses anything held back by .deploy-hold, shows what is about to be
introduced and asks, rebuilds with the right version baked in, replaces the
container, waits for healthy, then prunes all but the newest
IHASMAIL_KEEP_VERSIONS images — never the one actually running.
./deploy.sh # origin/main, asks before shipping new commits
./deploy.sh --dry-run # run the guards and stop
./deploy.sh v2026.8.30 --yes # a named ref, no prompt (there is no tty over ssh)
--yes does not override a hold; clearing one means deleting its line.
Contributing
CONTRIBUTING.md · CODE_OF_CONDUCT.md · SECURITY.md — please report vulnerabilities privately.
License
Copyright (C) 2026 Coffey Labs — AGPL-3.0-or-later. See LICENSE.
ihasmail was relicensed from GPL-3.0 to AGPL-3.0 on 2026-08-25: webmail is nearly always run as a network service rather than handed to anyone as a binary, and the AGPL's section 13 closes that gap.
That offer has to point at your source, not this one. If you run a modified
ihasmail, set SOURCE_URL to your own repository — the sign-in page and
Settings › About both show it. See
Rebranding.