rnsh
is a utility written in Python that facilitates shell
sessions over Reticulum networks.
It is based on the rnx
utility that ships with Reticulum and
aims to provide a similar experience to SSH.
The interface is starting to firm up, but some bug fixes at this point still may introduce breaking changes, especially in the protocol layers of the software.
- Fix invalid escape sequence handling for terminal escape sequences
- Fix an issue where disconnecting a session using ~. would result in further connections to the same initiator would appear to hang.
- Setting
-q
will suppress the pre-connect spinners
- Adaptive compression (RNS update) provides significant performance improvements (PR)
- Allowed identities file - put allowed identities in a file instead of on the command line for easier service invocations. (see PR for details)
- Escape Sequences, Session Termination & Line-Interactive Mode (see PR for details)
- Fix issue with intermittent data corruption
- First beta! Includes peformance improvements.
- Bug fixes
- Remove service name from RNS destination aspects. Service name now selects a suffix for the identity file and should only be supplied on the listener. The initiator only needs the destination hash of the listener to connect.
- Show a spinner during link establishment on tty sessions
- Attempt to catch and beautify exceptions on initiator
- Event loop bursting improves throughput and CPU utilization on both listener and initiator.
- Packet retries use RNS resend feature to prevent duplicate packets.
- Rate limit window change events to prevent saturation of transport
- Tweaked some loop timers to improve CPU utilization
- Switch to a new packet-based protocol
- Bug fixes and dependency updates
Tested (thus far) on Python 3.11 macOS 13.1 ARM64. Should run on Python 3.6+ on Linux or Unix. WSL probably works. Cygwin might work, too.
- Activate a virtualenv
pip3 install rnsh
- Or from a
whl
release,pip3 install /path/to/rnsh-0.0.1-py3-none-any.whl
- Or from a
- Configure Reticulum interfaces, check with
rnstatus
- Ready to run
rnsh
. The options are shown below.
Before running the listener or initiator, you'll need to get the listener destination hash and the initiator identity hash.
# On listener
rnsh -l -p
# On initiator
rnsh -p
Note: service name no longer is supplied on initiator. The destination hash encapsulates this information.
- Listening for default service name ("default").
- Using user's default Reticulum config dir (~/.reticulum).
- Using default identity ($RNSCONFIGDIR/storage/identities/rnsh).
- Allowing remote identity
6d47805065fa470852cf1b1ef417a1ac
to connect. - Launching
/bin/zsh
on authorized connect.
rnsh -l -a 6d47805065fa470852cf1b1ef417a1ac -- /bin/zsh
- Connecting to default service name ("default").
- Using user's default Reticulum config dir (~/.reticulum).
- Using default identity ($RNSCONFIGDIR/storage/identities/rnsh).
- Connecting to destination
a5f72aefc2cb3cdba648f73f77c4e887
rnsh a5f72aefc2cb3cdba648f73f77c4e887
Usage:
rnsh -l [-c <configdir>] [-i <identityfile> | -s <service_name>] [-v... | -q...] -p
rnsh -l [-c <configdir>] [-i <identityfile> | -s <service_name>] [-v... | -q...]
[-b <period>] (-n | -a <identity_hash> [-a <identity_hash>] ...) [-A | -C]
[[--] <program> [<arg> ...]]
rnsh [-c <configdir>] [-i <identityfile>] [-v... | -q...] -p
rnsh [-c <configdir>] [-i <identityfile>] [-v... | -q...] [-N] [-m] [-w <timeout>]
<destination_hash> [[--] <program> [<arg> ...]]
rnsh -h
rnsh --version
Options:
-c DIR --config DIR Alternate Reticulum config directory to use
-i FILE --identity FILE Specific identity file to use
-s NAME --service NAME Service name for identity file if not default
-p --print-identity Print identity information and exit
-l --listen Listen (server) mode. If supplied, <program> <arg>...will
be used as the command line when the initiator does not
provide one or when remote command is disabled. If
<program> is not supplied, the default shell of the
user rnsh is running under will be used.
-b --announce PERIOD Announce on startup and every PERIOD seconds
Specify 0 for PERIOD to announce on startup only.
-a HASH --allowed HASH Specify identities allowed to connect
-n --no-auth Disable authentication
-N --no-id Disable identify on connect
-A --remote-command-as-args Concatenate remote command to argument list of <program>/shell
-C --no-remote-command Disable executing command line from remote
-m --mirror Client returns with code of remote process
-w TIME --timeout TIME Specify client connect and request timeout in seconds
-q --quiet Increase quietness (move level up), multiple increases effect
DEFAULT LOGGING LEVEL
CRITICAL (silent)
Initiator -> ERROR
WARNING
Listener -> INFO
DEBUG (insane)
-v --verbose Increase verbosity (move level down), multiple increases effect
--version Show version
-h --help Show this help
Listener instances are the servers. Each listener is configured with an RNS identity, and a service name. Together, RNS makes these into a destination hash that can be used to connect to your listener.
Each listener must use a unique identity. The -s
parameter
can be used to specify a service name, which creates a unique
identity file.
Listeners can be configured with a command line to run on connect. Initiators can supply a command line as well. There are several different options for the way the command line is handled:
-C
no initiator command line is allowed; the connection will be terminated if one is supplied.-A
initiator-supplied command line is appended to listener- configured command line- With neither of these options, the listener will use the first
valid command line from this list:
- Initiator-supplied command line
- Listener command line argument
- Default shell of user listener is running under
If the -n
option is not set on the listener, the initiator
is required to identify before starting a command. The program
will be started with the initiator's identity hash string is set
in the environment variable RNS_REMOTE_IDENTITY
.
Listeners are set up using the -l
flag.
Initiators are the clients. Each initiator has an identity hash which is used as an authentication mechanism on Reticulum. You'll need this value to configure the listener to allow your connection. It is possible to run the server without authentication, but hopefully it's obvious that this is an advanced use case.
To get the identity hash, use the -p
flag.
With the initiator identity set up in the listener command
line, and with the listener identity copied (you'll need to
do -p
on the listener side, too), you can run the
initiator.
I recommend staying pretty vanilla to start with and
trying /bin/zsh
or whatever your favorite shell is these
days. The shell should start in login mode. Ideally it
works just like an ssh
shell session.
The protocol is build on top of the Reticulum Packet
API.
Application software sends and receives Message
objects,
which are encapsulated by Packet
objects. Messages are
(currently) sent one per packet, and only one packet is
sent at a time (per link). The next packet is not sent until
the receiver proves the outstanding packet.
A future update will work to allow a sliding window of outstanding packets to improve channel utilization.
-
Initiator establishes link. Listener session enters state
LSSTATE_WAIT_IDENT
, orLSSTATE_WAIT_VERS
if running with--no-auth
option. -
Initiator identifies on link if not using
--no-id
.- If using
--allowed-hash
, listener validates identity against configuration and if no match, sends a protocol error message and tears down link after prune timer.
- If using
-
Initiator transmits a
VersionInformationMessage
, which is evaluated by the server for compatibility. If incompatible, a protocol error is sent. -
Listener responds with a
VersionInfoMessage
and enters stateLSSTATE_WAIT_CMD
-
Initiator evaluates the listener's version information for compatibility and if incompatible, tears down link.
-
Initiator sends an
ExecuteCommandMessage
(which could be an empty command) and enters the session event loop. -
Listener evaluates the command message against the configured options such as
-A
or-C
and responds with a protocol error if not allowed. -
Listener starts the program. If success, the listener enters the session event loop. If failure, responds with a
CommandExitedMessage
.
Process messages received from initiator.
WindowSizeMessage
: set window size on child tty if appropriateStreamDataMessage
: binary data stream for child process; streams ids 0, 1, 2 = stdin, stdout, stderrNoopMessage
: no operation - listener replies withNoopMessage
- When link is torn down, child process is terminated if running and session destroyed
- If command terminates, a
CommandExitedMessage
is sent and session is pruned after an idle timeout.
Process messages received from listener.
ErrorMessage
: print error, terminate link, and exitStreamDataMessage
: binary stream information; streams ids 0, 1, 2 = stdin, stdout, stderrCommandExitedMessage
: remote command exited- If link is torn down unexpectedly, print message and exit
- Plan a better roadmap
- ?
- Keep my day job
-
Initial version -
Pip package with command-line utility support -
Publish to PyPI -
Improve signal handling -
Make it scriptable (currently requires a tty) -
Protocol improvements (throughput!) -
Documentation improvements -
Fix issues with runningrnsh
in a binary pipeline, i.e. piping the output oftar
overrsh
. -
Test on several platforms -
Fix issues that come up with testing -
v0.1.0 beta -
Test and fix more issues - More betas
- Enhancement Ideas
-
authorized_keys
mode similar to SSH to allow one listener process to serve multiple users - Git over
rnsh
(git remote helper) - Sliding window acknowledgements for improved throughput
-
- v1.0 someday probably maybe
By piping into/out of rnsh
, it is possible to transfer
files using the same method discussed in
this article.
It's not terribly fast currently, due to the round-trip rule
enforced by the protocol. Sliding window acknowledgements will
speed this up significantly.
Running tests: poetry run pytest tests