it seems linux machines with systemd-resolved don't always set up
/etc/resolv.conf correctly. there may be no "nameserver" entry, causing Go's
net resolver to fallback to 127.0.0.1 and ::1. Systemd-resolved is listening on
127.0.0.53, so users will likely get a "connection refused". So point users to
the systemd-resolved manual page.
for issue #38 by ArnoSen
with tls with acme (with pebble, a small acme server for testing), and with
pregenerated keys/certs.
the two mox instances are configured on their own domain. we launch a separate
test container that connects to the first, submits a message for delivery to
the second. we check if the message is delivered with an imap connection and
the idle command.
we were adding the missing date and/or message-id header, but didn't sign it.
and the default dkim signing config is to (over)sign those headers. so that was
causing errors with bad signatures.
found while setting up automated tests for quickstart, while sending a very
basic message between a fresh install.
i wondered why self-signed mtasts certs didn't result in delivery failure. it's
because it was a first-time request of the mtasts policy (clean test
container). and for that case it means mtasts should be ignored.
so external tools (like fail2ban) can monitor the logs and block ip's of bots.
for issue #30 by inigoserna, though i'm not sure i interpreted the suggestion correctly.
makes it easier to use tls keys/certs managed by other tools, with or without
acme. the root process has access to open such files. the child process reads
the key from the file descriptor, then closes the file.
for issue #30 by inigoserna, thanks!
someone asked at the the recent golang rotterdam meetup if this would be added.
i looked into it, and it requires implementing an imap extension
XAPPLEPUSHSERVICE (not documented, but apple published modified dovecot
software for macos server that implemented it). to send push notifications to
the ios mail app, you need a APNS certificate. the tutorials online explain you
have to purchase macos server (a deprecated product) and extract the APNS
certificate. the certificate is valid for one year. i'm not sure it still
works, and it feels like it could stop working at any moment. but implementing
it seems doable.
if we recognize that a request for a WebForward is trying to turn the
connection into a websocket, we forward it to the backend and check if the
backend understands the websocket request. if so, we pass back the upgrade
response and get out of the way, copying bytes between the two. we do log the
total amount of bytes read from the client and written to the client. if the
backend doesn't respond with a websocke response, or an invalid one, we respond
with a regular non-websocket response. and we log details about the failed
connection, should help with debugging and any bug reports.
we don't try to parse the websocket framing, that's between the client and the
backend. we could try to parse it, in part to protect the backend from bad
frames, but it would be a lot of work and could be brittle in the face of
extensions.
this doesn't yet handle websocket connections when a http proxy is configured.
we'll implement it when someone needs it. we do recognize it and fail the
connection.
for issue #25
the backup command will make consistent snapshots of all the database files. i
had been copying the db files before, and it usually works. but if the file is
modified during the backup, it is inconsistent and is likely to generate errors
when reading (can be at any moment in the future, when reading some db page).
"mox backup" opens the database file and writes out a copy in a transaction.
it also duplicates the message files.
before doing a restore, you could run "mox verifydata" on the to-be-restored
"data" directory. it check the database files, and compares the message files
with the database.
the new "gentestdata" subcommand generates a basic "data" directory, with a
queue and a few accounts. we will use it in the future along with "verifydata"
to test upgrades from old version to the latest version. both when going to the
next version, and when skipping several versions. the script test-upgrades.sh
executes these tests and doesn't do anything at the moment, because no releases
have this subcommand yet.
inspired by a failed upgrade attempt of a pre-release version.
such deliveries would fail because a canceled "context" was reused, so the dns
lookups would fail.
the tests didn't catch it before because they ignored their context parameters.
e.g. when outlook.com puts your IP on a blocklist, it will respond with 550 to
MAIL FROM, then closes the connection (without responding to the remaining
commands). we were reading the 550 response, not yet acting on it, then reading
the response to RCPT TO. that read failed, and we would return that error. now,
we will properly return the 550 (permanent error, instead of the temporary read
error) from the first MAIL FROM (but we do still always try to read the
response for RCPT TO and DATA).
the screenshots are not in the git repo, they may change quite a bit and are
larger binary blobs. i don't want to make the repo too big (the code with its
dependencies is already pretty big!).
the website with the screenshots is in github.com/mjl-/mox.
for github issue #26
dmarc verifiers will only accept a dkim signature if the domain the message From
header matches the domain of the signature (i.e. it is "aligned").
i hadn't run into this before and when testing because thunderbird sets the
"smtp mail from" to the same address as a custom "message from" header. but
other mail clients don't have to do that.
should fix issue #22
by specifying a "destination" in an account that is just "@" followed by the
domain, e.g. "@example.org". messages are only delivered to the catchall
address when no regular destination matches (taking the per-domain
catchall-separator and case-sensisitivity into account).
for issue #18
such configurations are certainly errors, but were silently accepted and highly
likely not doing what you may have hoped. i suspect no one has configured mox
this way.
by default 1000 messages per day, and to max 200 first-time receivers.
i don't think a person would reach those limits. a compromised account abused
by spammers could easily reach that limit. this prevents further damage.
the error message you will get is quite clear, pointing to the configuration
parameter that should be changed.
