Certain (testing) scenarios may find it useful to disable the per-client
rate limiter.

This is 95% done, traffic is generated at the approriate times, is
dispatched, and the responses...

Oops, no wonder the #50 branch was having decoy traffic arrive early.

This eventually needs to be tunable, but it's important to avoid
starving the send side of the s...

This isn't massive and by no means makes the cases when the database
lookup needs to happen chea...

The external memory queue is backed by a BoltDB database running with
`NoSync` for performance r...

This is the first step to allowing more than one queue implementation to
be present in the code ...

* The inbound queue is now batch processed, and all avaliable enqueued
worker packets are sc...

This implements a basic leaky bucket rate limit on outgoing client
traffic based on the PKI Send...

This is the delay spent waiting for an unwrap to occur. Under no
normal realistic circumstance ...

I still don't think blown queue deadlines are going to happen all that
often, but this change sh...

Requests and responses are a simple (versioned) JSON structure.

Fixes #44.

This is a breaking change for the pgx backend that will require
recreating the database. Additi...

This only supports a trivial `loop` Kaetzchen, but it works, and this
lays down the framework re...

The auto-responder code needs to do something similar eventually, and
common code is common code.

This adds support for interfacing with an RDBMS, where "RDBMS" is
defined as Postgresql. The Sp...

The result of a `Bucket.Get()` are only guaranteed to be valid till
the end of a transaction, an...

The server code was kind of hard to maintain, and component boundaries
were not very clear. Thi...

I need to figure out how I want to let this override standard addresses,
but it works fine for t...

This changes the config file format, but will be less of a mess when
people add more backend imp...

This also clamps the enforced MaxDelay to the number of mix keys that
are published per descript...

Fixes #31.

This should make it harder (but not impossible) for an adversary to fill
up the mix queue. Than...

The current user message spool implementation renders this largely
pointless as the bulk of the ...

Infinite channels are scary, so do the drop head on rising delay thing
here as well. The tunabl...

Part of #10.

Part of #10.

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

...

The `domain` component of an address is case insensitive by spec, so
preserving the case here is...

* `ExternAuth` -> `externAuth`, access should use the `userdb.UserDB`
interface.
* Switch t...

Everything else will use `github.com/BurntSushi/toml` because it can
handle unmarshalling nested...

The Provider now has additional configuration options:

* BinaryRecipients - Treat all user nam...

Part of katzenpost/mixnet_uprising#15

Part of #26.

Part of #26.

Currently "too long" is a extremely generous 250 ms by default. I also
added some insturmentati...

Tangentially related to #14.

Instead of getting stuck trying to publish for the current epoch, give
up on said epoch if the a...

Most of the cases where panic was called were better handled as
non-fatal errors anyway.

Fixes ...

Since nodes should honor prior topologies when possible, it should fetch
documents for prior epo...

The PKI fetch logic will now skip fetching the Document for any given
epoch if:

* The fetch op...

Fix url parser in the provider configuration

Implement external http userdb

Not sure if saving the copy here is worth it, but it probably is.

When using "stale" documents, it's neccecary to ensure that more recent
documents haven't delist...

Bad things happen if nodes get their layer assignment changed since it
essentially is equivalent...

`pki.autheticateIncoming` and `pki.authenticateOutgoing` were basically
almost identical, so mer...

This shouldn't be a common occurence, but the ACK mechanism for reliable
delivery is technically...

This should always work, but it's slightly inefficient.

Tangentially related to #14.

Removed some redundant code.