wtxpwh / tty-share

Terminal sharing over the Internet.

tty-share

It is a very simple command line tool that gives remote access to a UNIX terminal session. It's using the PTY system, so it should work on any UNIX system (Linux, OSX). Because it's written in GO, the tool will be a single binary, with no dependencies, which will also work on your ARM Raspberry Pi.

The most important part about it is that it requires no setup on the remote end. All I need to give remote access to the terminal (a bash/shell session) is the binary tool, and the remote person only needs to open a secret URL in their browser.

The project consists of two command line utilities: tty-share and tty-server.

The tty-share is used on the machine that wants to share the terminal, and it connects to the server to generate a secret URL, over which the terminal can be viewed in the browser.

The server runs at tty-share.com, so you only need the tty-server binary if you want to host it yourself.

More documentation

The documentation is very poor now. More will follow. This describes briefly some thoughts behind the architecture of the project.

Running

Download the latest tty-share binary release, and run it:

bash$ tty-share
Web terminal: https://go.tty-share.com/s/J5U6FAwChWNP0I9VQ9XyPqVD6m6IpI8-sBLRiz98XMA=

bash$

Building tty-share locally

If you want to just build the tool that shares your terminal, and not the server, then simply do a

make out/tty-share

This way you don't have to bother about the server side, nor about building the frontend, and you will get only the tty-share cmd line tool, inside out folder.

Building and running everything

For an easy deployment, the tty-server is by bundling by default all frontend resources inside the final binary. So in the end, there will be only one file to be copied and deployed. However, the frontend resources can also be served from a local folder, with a command line flag.

Build all

cd frontend
nvm use
npm install
npm run build # builds the frontend
cd -
make all # builds both the sender and server

Run a development server

make runs

Will run the server on the localhost.

Run a development sender

make runc

Will run the sender and connect it to the server running on the local host (so the above command has to be ran first).

For more info, on how to run, see the Makefile, or the help of the two binaries (tty-share and tty_receiver)

The project didn't follow the typical way of building go applications, because everything is put in one single project and package, for the ease of development, and also because the bundle containing the frontend has to be built as well. Perhaps there's a better way, but this works for now.

TLS and HTTPS

At the moment the tty-share supports connecting over a TLS connection to the server, but the server doesn't have that implemented yet. However, the server can easily run behind a proxy which can take care of encrypting the connections from the senders and receivers (doing both TLS and HTTPS), without the server knowing about it.

The server at tty-share is using both TLS and https for both sides, relying on nginx reverse proxy.

However, the tty-server should maybe also have native support for being able to listen on TLS connections from the sender as well. This can easily be added in the future.

TODO

There are several improvements, and additions that can be done further:

• Update and write more tests.
• Add support for listening on sender connections over TLS.
• React on the tty-receiver window size as well. For now, the size of the terminal window is decided by the tty-share, but perhaps both the sender and receiver should have a say in this.
• Read only sessions, where the tty_receiver side can only watch, and cannot interact with the terminal session.
• Command line tty_receiver can be implemented as well, without the need of a browser.
• End-to-end encryption. Right now, the server can see what messages the sender and receiver are exchanging, but an end-to-end encryption layer can be built on top of this. It has been thought from the beginning, but it's just not implemented. The terminal IO messages are packed in protocol messages, and the payload can be easily encrypted with a shared key derived from a password that only the sender and receiver know.
• Notify the tty-share user when a tty-receiver got connected (when the remote person opened the URL in their browser).
• Many other

Similar solutions

VSC (Visual Studio Code) Live Share

I've tried Visual Studio Code sharing, and it seems to work relatively well. One big advantage is that both persons in the session can write code, and navigate independently of each other. It also supports terminal sharing.

However, the two disadvantages with this tool are the need of logging in with a Github (or Microsoft) account, and having to install Visual Studio Code on the remote side too. I don't want to force the remote person to install VSC just for them to get access to a terminal session. Visual Studio Code might be popular in the web development circles, but it is not popular in the other development corners.

tmate.io

This is a great tool, and I used it quite a few times before. At the time I started my project, tmate.io didn't have the option to join the session from the browser, and one had to use ssh. In most cases, ssh is not a problem at all - in fact it's even preferred, but there are cases when you just don't have easy access to an ssh client, e.g.: joining from a Windows machine, or from your smartphone. In the meantime, the project added some support for joining a terminal session in the browser too.

Perhaps one downside with tmate is that it comes with quite a few dependencies which can make your life complicated if you want to compile it for ARM, for example. Running it on my raspberry pi might not be as simple as you want it, unless you use Debian.

Credits

• xterm.js - used in the browser receiver.
• gotty - used for inspiration and motivation.
• tmate - inspiration and motivation.
• https://codepen.io/chiaren/pen/ALwnI - for the free 404 page.