Part of #11.

There's a slight delay based on minclient's polling frequency, but this
is a lot cleaner (and it...

The naive implementation of the client cover traffic side of Loopix just
needs to well, send cov...

Might as well do this too, though the only upstream proxy provider that
will use this is likely ...

The supported protocols are:
* ``, `none` - Direct connection without a proxy.
* `socks5` - ...

The old algorithm was a straight forward stop-and-wait ARQ. The new
algorithm sends the entire ...

I held off on this, but there's no reason why this shouldn't happen now.

This is better and more obvious when using the package as a library.

Part of #11.

Disabled by default, the timeout restarts after a given message has made
forward progress, and t...

`Debug.PollingInterval` is now a thing. Tuning this to a sensible
default is left as a matter o...

This can be improved still:

* If the message is making forward progress still, then the bounce...

If the SMTP server fails to enqueue messages for one or more (but not
all) recipients, a multipa...

* Add a multipart/report generator to the imf package.
* Wire it in to the receive side error ...

`X-Katzenpost-Unreliable: true` will enable this behavior. Message
fragmentation still happens....

The old header was too long, and would get wrapped to be under the
suggested (SHOULD) maximum li...

Set the non-standard Noise max message size to something rediculously
large. The current fragme...

Since the minclient behavior is to dispatch another `ReceiveMessage`
command after the callback ...

This stores message plaintext on disk in a encrypted envelope using
Noise_N encrypted with a per...

In favor of something that is explicitly programatic, because no one
that is not using mailproxy...

I need to think about what the best behavior is more before I'm
comfortable using skewed time.

This makes the SMTP server actually do something. In theory everything
but bounce message gener...

Provider identifiers are now normalized following the IDNA Domain Lookup
Protocol conventions.

...

This is somewhat cleaner, and more stuff will happen in here.

Part of #5.

This doesn't actually do anything with the mail it accepts (though it
does validate `MAIL FROM`,...

The POP3 code uses textproto's DotWriter so there isn't a limit, and
attempting to enforce one j...

Regardless of what the RFC says, people will expect this to work, so we
should support it.

This provides an e-mail address -> ECDH public key store, and associated
thwack commands to set/...

The only command registered so far are `SHUTDOWN` and `QUIT`, but this
appears to work fine.

Pa...

Accounts can be case sensitive, provider identifiers as a general rule
can not.

This reverts fca35dfc7b2e495beb1bbd12c650a04b26f38b75 in favor of
enforcing a generous 64 KiB li...

The IMF spec mandates US-ASCII encoding for compliant messages, with the
solution for anything t...

There may come a time when someone adds a non-POP3 interface, and this
operation is something th...

There is no guarantee that the body of messages will adhere to the RFC
mandated line limits (998...

The de-duplication cache is now periodically pruned of entries that are
older than 5 days, at mo...