mox/README.md
Mechiel Lukkien 0f8bf2f220
replace http basic auth for web interfaces with session cookie & csrf-based auth
the http basic auth we had was very simple to reason about, and to implement.
but it has a major downside:

there is no way to logout, browsers keep sending credentials. ideally, browsers
themselves would show a button to stop sending credentials.

a related downside: the http auth mechanism doesn't indicate for which server
paths the credentials are.

another downside: the original password is sent to the server with each
request. though sending original passwords to web servers seems to be
considered normal.

our new approach uses session cookies, along with csrf values when we can. the
sessions are server-side managed, automatically extended on each use. this
makes it easy to invalidate sessions and keeps the frontend simpler (than with
long- vs short-term sessions and refreshing). the cookies are httponly,
samesite=strict, scoped to the path of the web interface. cookies are set
"secure" when set over https. the cookie is set by a successful call to Login.
a call to Logout invalidates a session. changing a password invalidates all
sessions for a user, but keeps the session with which the password was changed
alive. the csrf value is also random, and associated with the session cookie.
the csrf must be sent as header for api calls, or as parameter for direct form
posts (where we cannot set a custom header). rest-like calls made directly by
the browser, e.g. for images, don't have a csrf protection. the csrf value is
returned by the Login api call and stored in localstorage.

api calls without credentials return code "user:noAuth", and with bad
credentials return "user:badAuth". the api client recognizes this and triggers
a login. after a login, all auth-failed api calls are automatically retried.
only for "user:badAuth" is an error message displayed in the login form (e.g.
session expired).

in an ideal world, browsers would take care of most session management. a
server would indicate authentication is needed (like http basic auth), and the
browsers uses trusted ui to request credentials for the server & path. the
browser could use safer mechanism than sending original passwords to the
server, such as scram, along with a standard way to create sessions.  for now,
web developers have to do authentication themselves: from showing the login
prompt, ensuring the right session/csrf cookies/localstorage/headers/etc are
sent with each request.

webauthn is a newer way to do authentication, perhaps we'll implement it in the
future. though hardware tokens aren't an attractive option for many users, and
it may be overkill as long as we still do old-fashioned authentication in smtp
& imap where passwords can be sent to the server.

for issue #58
2024-01-05 10:48:42 +01:00

21 KiB

Mox is a modern full-featured open source secure mail server for low-maintenance self-hosted email.

See Quickstart below to get started.

Features

  • Quick and easy to start/maintain mail server, for your own domain(s).
  • SMTP (with extensions) for receiving, submitting and delivering email.
  • IMAP4 (with extensions) for giving email clients access to email.
  • Webmail for reading/sending email from the browser.
  • SPF/DKIM/DMARC for authenticating messages/delivery, also DMARC aggregate reports.
  • Reputation tracking, learning (per user) host-, domain- and sender address-based reputation from (Non-)Junk email classification.
  • Bayesian spam filtering that learns (per user) from (Non-)Junk email.
  • Slowing down senders with no/low reputation or questionable email content (similar to greylisting). Rejected emails are stored in a mailbox called Rejects for a short period, helping with misclassified legitimate synchronous signup/login/transactional emails.
  • Internationalized email, with unicode in email address usernames ("localparts"), and in domain names (IDNA).
  • Automatic TLS with ACME, for use with Let's Encrypt and other CA's.
  • DANE and MTA-STS for inbound and outbound delivery over SMTP with STARTTLS, including REQUIRETLS and with incoming/outgoing TLSRPT reporting.
  • Web admin interface that helps you set up your domains and accounts (instructions to create DNS records, configure SPF/DKIM/DMARC/TLSRPT/MTA-STS), for status information, managing accounts/domains, and modifying the configuration file.
  • Account autodiscovery (with SRV records, Microsoft-style, Thunderbird-style, and Apple device management profiles) for easy account setup (though client support is limited).
  • Webserver with serving static files and forwarding requests (reverse proxy), so port 443 can also be used to serve websites.
  • Prometheus metrics and structured logging for operational insight.
  • "mox localserve" subcommand for running mox locally for email-related testing/developing, including pedantic mode.
  • Most non-server Go packages mox consists of are written to be reusable.

Mox is available under the MIT-license and was created by Mechiel Lukkien, mechiel@ueber.net. Mox includes BSD-3-claused code from the Go Authors, and the Public Suffix List by Mozilla under Mozilla Public License, v2.0.

Mox has automated tests, including for interoperability with Postfix for SMTP. Mox is manually tested with email clients: Mozilla Thunderbird, mutt, iOS Mail, macOS Mail, Android Mail, Microsoft Outlook. Mox is also manually tested to interoperate with popular cloud providers: gmail.com, outlook.com, yahoo.com, proton.me.

The code is heavily cross-referenced with the RFCs for readability/maintainability.

Quickstart

The easiest way to get started with serving email for your domain is to get a (virtual) machine dedicated to serving email, name it [host].[domain] (e.g. mail.example.com), login as root, and run:

# Create mox user and homedir (or pick another name or homedir):
useradd -m -d /home/mox mox

cd /home/mox
... compile or download mox to this directory, see below ...

# Generate config files for your address/domain:
./mox quickstart you@example.com

The quickstart creates configuration files for the domain and account, generates an admin and account password, prints the DNS records you need to add and prints commands to start mox and optionally install mox as a service.

A machine that doesn't already run a webserver is highly recommended because modern email requires HTTPS, and mox currently needs it for automatic TLS. You could combine mox with an existing webserver, but it requires a lot more configuration. If you want to serve websites on the same machine, consider using the webserver built into mox. It's pretty good! If you want to run an existing webserver on port 443/80, see "mox help quickstart".

After starting, you can access the admin web interface on internal IPs.

Download

You can easily (cross) compile mox if you have a recent Go toolchain installed (see "go version", it must be >= 1.20; otherwise, see https://go.dev/dl/ or https://go.dev/doc/manage-install and $HOME/go/bin):

GOBIN=$PWD CGO_ENABLED=0 go install github.com/mjl-/mox@latest

Or you can download a binary built with the latest Go toolchain from https://beta.gobuilds.org/github.com/mjl-/mox@latest/linux-amd64-latest/, and symlink or rename it to "mox".

Verify you have a working mox binary:

./mox version

Mox only compiles for and fully works on unix systems. Mox also compiles for Windows, but "mox serve" does not yet work, though "mox localserve" (for a local test instance) and most other subcommands do. Mox does not compile for Plan 9.

You can also run mox with docker image r.xmox.nl/mox, with tags like v0.0.1 and v0.0.1-go1.20.1-alpine3.17.2, see https://r.xmox.nl/r/mox/. Though new docker images aren't (automatically) generated for new Go runtime/compile releases. See docker-compose.yml in this repository for instructions on starting. It is important to run with docker host networking, so mox can use the public IPs and has correct remote IP information for incoming connections (important for junk filtering and rate-limiting). Given these caveats, it's recommended to run mox without docker.

Future/development

Mox will receive funding for essentially full-time continued work from August 2023 to August 2024 through NLnet/EU's NGI0 Entrust, see https://nlnet.nl/project/Mox/.

Roadmap

  • Improve documentation
  • Improve SMTP delivery from queue
  • Webmail improvements
  • HTTP-based API for sending messages and receiving delivery feedback
  • Calendaring with CalDAV/iCal
  • More IMAP extensions (PREVIEW, WITHIN, IMPORTANT, COMPRESS=DEFLATE, CREATE-SPECIAL-USE, SAVEDATE, UNAUTHENTICATE, REPLACE, QUOTA, NOTIFY, MULTIAPPEND, OBJECTID, MULTISEARCH)
  • ARC, with forwarded email from trusted source
  • Forwarding (to an external address)
  • Add special IMAP mailbox ("Queue?") that contains queued but undelivered messages, updated with IMAP flags/keywords/tags and message headers.
  • Sieve for filtering (for now see Rulesets in the account config)
  • Expose threading through IMAP extension
  • Autoresponder (out of office/vacation)
  • OAUTH2 support, for single sign on
  • Privilege separation, isolating parts of the application to more restricted sandbox (e.g. new unauthenticated connections)
  • Using mox as backup MX
  • JMAP
  • Milter support, for integration with external tools
  • IMAP extensions for "online"/non-syncing/webmail clients (SORT (including DISPLAYFROM, DISPLAYTO), THREAD, PARTIAL, CONTEXT=SEARCH CONTEXT=SORT ESORT, FILTERS)
  • IMAP Sieve extension, to run Sieve scripts after message changes (not only new deliveries)
  • Improve support for mobile clients with extensions: IMAP URLAUTH, SMTP CHUNKING and BINARYMIME, IMAP CATENATE

There are many smaller improvements to make as well, search for "todo" in the code.

Not supported/planned

But perhaps in the future...

  • Mailing list manager
  • Functioning as SMTP relay
  • POP3
  • Delivery to (unix) OS system users
  • Support for pluggable delivery mechanisms
  • iOS Mail push notifications (with XAPPLEPUSHSERVICE undocumented imap extension and hard to get APNS certificate)

FAQ - Frequently Asked Questions

Why a new mail server implementation?

Mox aims to make "running a mail server" easy and nearly effortless. Excellent quality (open source) mail server software exists, but getting a working setup typically requires you configure half a dozen services (SMTP, IMAP, SPF/DKIM/DMARC, spam filtering), which are often written in C (where small bugs often have large consequences). That seems to lead to people no longer running their own mail servers, instead switching to one of the few centralized email providers. Email with SMTP is a long-time decentralized messaging protocol. To keep it decentralized, people need to run their own mail server. Mox aims to make that easy.

Where is the documentation?

See all commands and help output at https://pkg.go.dev/github.com/mjl-/mox/.

See the commented example config files at https://pkg.go.dev/github.com/mjl-/mox/config/. They often contain enough documentation about a feature and how to configure it.

You can get the same information by running "mox" without arguments to list its subcommands and usage, and "mox help [subcommand]" for more details.

The example config files are printed by "mox config describe-static" and "mox config describe-dynamic".

Mox is still in early stages, and documentation is still limited. Please create an issue describing what is unclear or confusing, and we'll try to improve the documentation.

Is Mox affected by SMTP smuggling?

Mox itself is not affected: it only treats "\r\n.\r\n" as SMTP end-of-message. But read on for caveats.

SMTP smuggling exploits differences in handling by SMTP servers of: carriage returns (CR, or "\r"), newlines (line feeds, LF, "\n") in the context of "dot stuffing". SMTP is a text-based protocol. An SMTP transaction to send a message is finalized with a "\r\n.\r\n" sequence. This sequence could occur in the message being transferred, so any verbatim "." at the start of a line in a message is "escaped" with another dot ("dot stuffing"), to not trigger the SMTP end-of-message. SMTP smuggling takes advantage of bugs in some mail servers that interpret other sequences than "\r\n.\r\n" as SMTP end-of-message. For example "\n.\n" or even "\r.\r", and perhaps even other magic character combinations.

Before v0.0.9, mox accepted SMTP transactions with bare carriage returns (without newline) for compatibility with real-world email messages, considering them meaningless and therefore innocuous.

Since v0.0.9, SMTP transactions with bare carriage returns are rejected. Sending messages with bare carriage returns to buggy mail servers can cause those mail servers to materialize non-existent messages. Now that mox rejects messages with bare carriage returns, sending a message through mox can no longer be used to trigger those bugs.

Mox can still handle bare carriage returns in email messages, e.g. those imported from mbox files or Maildirs, or from messages added over IMAP. Mox still fixes up messages with bare newlines by adding the missing carriage returns.

Before v0.0.9, an SMTP transaction for a message containing "\n.\n" would result in a non-specific error message, and "\r\n.\n" would result in the dot being dropped. Since v0.0.9, these sequences are rejected with a message mentioning SMTP smuggling.

How do I import/export email?

Use the import functionality on the accounts web page to import a zip/tgz with maildirs/mbox files, or use the "mox import maildir" or "mox import mbox" subcommands. You could also use your IMAP email client, add your mox account, and copy or move messages from one account to the other.

Similarly, see the export functionality on the accounts web page and the "mox export maildir" and "mox export mbox" subcommands to export email.

Importing large mailboxes may require a lot of memory (a limitation of the current database). Splitting up mailboxes in smaller parts (e.g. 100k messages) would help.

How can I help?

Mox needs users and testing in real-life setups! So just give it a try, send and receive emails through it with your favourite email clients, and file an issue if you encounter a problem or would like to see a feature/functionality implemented.

Instead of switching email for your domain over to mox, you could simply configure mox for a subdomain, e.g. [you]@moxtest.[yourdomain].

If you have experience with how the email protocols are used in the wild, e.g. compatibility issues, limitations, anti-spam measures, specification violations, that would be interesting to hear about.

Pull requests for bug fixes and new code are welcome too. If the changes are large, it helps to start a discussion (create an "issue") before doing all the work. In practice, starting with a small contribution and growing from there has the highest chance of success.

By contributing (e.g. code), you agree your contributions are licensed under the MIT license (like mox), and have the rights to do so.

Where can I discuss mox?

Join #mox on irc.oftc.net, or #mox on the "Gopher slack".

For bug reports, please file an issue at https://github.com/mjl-/mox/issues/new.

How do I change my password?

Regular users (doing IMAP/SMTP with authentication) can change their password at the account page, e.g. http://localhost/. Or you can set a password with "mox setaccountpassword".

The admin can change the password of any account through the admin page, at http://localhost/admin/ by default (leave username empty when logging in).

The account and admin pages are served on localhost for configs created with the quickstart. To access these from your browser, run ssh -L 8080:localhost:80 you@yourmachine locally and open http://localhost:8080/[...].

The admin password can be changed with "mox setadminpassword".

How do I configure a second mox instance as a backup MX?

Unfortunately, mox does not yet provide an option for that. Mox does spam filtering based on reputation of received messages. It will take a good amount of work to share that information with a backup MX. Without that information, spammers could use a backup MX to get their spam accepted.

Until mox has a proper solution, you can simply run a single SMTP server. The author has run a single mail server for over a decade without issues. Machines and network connectivity are stable nowadays, and email delivery will be retried for many hours during temporary errors (e.g. when rebooting a machine after updates).

How do I stay up to date?

Please set "CheckUpdates: true" in mox.conf. Mox will check for a new version through a DNS TXT request for _updates.xmox.nl once per 24h. Only if a new version is published will the changelog be fetched and delivered to the postmaster mailbox.

The changelog, including latest update instructions, is at https://updates.xmox.nl/changelog.

You can also monitor newly added releases on this repository with the github "watch" feature, or use the github RSS feed for tags (https://github.com/mjl-/mox/tags.atom) or releases (https://github.com/mjl-/mox/releases.atom), or monitor the docker images.

Keep in mind you have a responsibility to keep the internet-connected software you run up to date and secure.

How do I upgrade my mox installation?

We try to make upgrades effortless and you can typically just put a new binary in place and restart. If manual actions are required, the release notes mention them. Check the release notes of all version between your current installation and the release you're upgrading to.

Before upgrading, make a backup of the data directory with mox backup <destdir>. This writes consistent snapshots of the database files, and duplicates message files from the outgoing queue and accounts. Using the new mox binary, run mox verifydata <backupdir> (do NOT use the "live" data directory!) for a dry run. If this fails, an upgrade will probably fail too. Important: verifydata with the new mox binary can modify the database files (due to automatic schema upgrades). So make a fresh backup again before the actual upgrade. See the help output of the "backup" and "verifydata" commands for more details.

During backup, message files are hardlinked if possible, and copied otherwise. Using a destination directory like data/tmp/backup increases the odds hardlinking succeeds: the default mox systemd service file mounts the data directory separately, so hardlinks to outside the data directory are cross-device and will fail.

If an upgrade fails and you have to restore (parts) of the data directory, you should run mox verifydata <datadir> (with the original binary) on the restored directory before starting mox again. If problematic files are found, for example queue or account message files that are not in the database, run mox verifydata -fix <datadir> to move away those files. After a restore, you may also want to run mox bumpuidvalidity <account> for each account for which messages in a mailbox changed, to force IMAP clients to synchronize mailbox state.

How secure is mox?

Security is high on the priority list for mox. Mox is young, so don't expect no bugs at all. Mox does have automated tests for some security aspects, e.g. for login, and uses fuzzing. Mox is written in Go, so some classes of bugs such as buffer mishandling do not typically result in privilege escalation. Of course logic bugs will still exist. If you find any security issues, please email them to mechiel@ueber.net.

I'm now running an email server, but how does email work?

Congrats and welcome to the club! Running an email server on the internet comes with some responsibilities so you should understand how it works. See https://explained-from-first-principles.com/email/ for a thorough explanation.

What are the minimum requirements to run mox?

Mox does not need much. Nowadays most machines are larger than mox needs. You can start with a machine with 512MB RAM, any CPU will do. For storage you should account for the size of the email messages (no compression currently), an additional 15% overhead for the meta data, and add some more headroom. Expand as necessary.

Can I see some screenshots?

Yes, see https://www.xmox.nl/screenshots/.

Mox has a webmail for reading/writing messages.

Mox also has an "account" web interface where users can view their account and manage their address configuration, such as rules for automatically delivering certain incoming messages to a specific mailbox.

And mox has an "admin" web interface where the administrator can make changes, e.g. add/remove/modify domains/accounts/addresses.

Won't the big email providers block my email?

It is a common misconception that it is impossible to run your own email server nowadays. The claim is that the handful big email providers will simply block your email. However, you can run your own email server just fine, and your email will be accepted, provided you are doing it right.

If your email is rejected, it is often because your IP address has a bad email sending reputation. Email servers often use IP blocklists to reject email networks with a bad email sending reputation. These blocklists often work at the level of whole network ranges. So if you try to run an email server from a hosting provider with a bad reputation (which happens if they don't monitor their network or don't act on abuse/spam reports), your IP too will have a bad reputation and other mail servers (both large and small) may reject messages coming from you. During the quickstart, mox checks if your IPs are on a few often-used blocklists. It's typically not a good idea to host an email server on the cheapest or largest cloud providers: They often don't spend the resources necessary for a good reputation, or they simply block all outgoing SMTP traffic. It's better to look for a technically-focused local provider. They too may initially block outgoing SMTP connections on new machines to prevent spam from their networks. But they will either automatically open up outgoing SMTP traffic after a cool down period (e.g. 24 hours), or after you've contacted their support.

After you get past the IP blocklist checks, email servers use many more signals to determine if your email message could be spam and should be rejected. Mox helps you set up a system that doesn't trigger most of the technical signals (e.g. with SPF/DKIM/DMARC). But there are more signals, for example: Sending to a mail server or address for the first time. Sending from a newly registered domain. Sending messages with content that resembles known spam messages.

Should your email be rejected, you will typically get an error message during the SMTP transaction that explains why. In the case of big email providers the error message often has instructions on how to prove to them you are a legimate sender.

Can I use existing TLS certificates/keys?

Yes. The quickstart command creates a config that uses ACME with Let's Encrypt, but you can change the config file to use existing certificate and key files.

You'll see "ACME: letsencrypt" in the "TLS" section of the "public" Listener. Remove or comment out the ACME-line, and add a "KeyCerts" section like in the example config file in https://pkg.go.dev/github.com/mjl-/mox/config#hdr-mox_conf. You can have multiple certificates and keys: The line with the "-" (dash) is the start of a list item. Duplicate that line up to and including the line with KeyFile for each certificate/key you have. Mox makes a TLS config that holds all specified certificates/keys, and uses it for all services for that Listener (including a webserver), choosing the correct certificate for incoming requests.

Keep in mind that for each email domain you host, you will need a certificate for mta-sts.<domain> and autoconfig.<domain>, unless you disable MTA-STS and autoconfig for that domain.

Mox opens the key and certificate files during initial startup, as root (and passes file descriptors to the unprivileged process). No special permissions are needed on the key and certificate files.