Lan Mouse Share

Goal of this project is to be an open-source replacement for proprietary tools like Synergy, Share Mouse.

Focus lies on performance and a clean, manageable implementation that can easily be expanded to support additional backends like e.g. Android, iOS, ... .

Of course blazingly fast™ and stable, because it's written in rust.

For an alternative (with slightly different goals) you may check out Input Leap.

OS Support

Backend	Event Receiving	Event Emitting
Wayland (wlroots)	✔️	✔️
Wayland (KDE)	WIP	✔️
Wayland (Gnome)	TODO (libei support)	TODO (wlr-layer-shell not supported)
X11	WIP	TODO
Windows	WIP	TODO
MacOS	TODO (I dont own a Mac)	TODO (I dont own a Mac)

Wayland compositor support

Wayland support for consuming and producing input-events currently relies on unstable wayland protocols:

• zwlr_virtual_pointer_manager_v1 is required to display surfaces on screen edges -> not supported by Gnome
• virtual-keyboard-unstable-v1 and wlr-virtual-pointer-unstable-v1 are used to emulate input on wlroots compositors
• kde-fake-input is used to emulate input in KDE (WIP)

Required Protocols (Event Emitting)	Sway	Kwin	Gnome
pointer-constraints-unstable-v1	✔️	✔️	✔️
relative-pointer-unstable-v1	✔️	✔️	✔️
keyboard-shortcuts-inhibit-unstable-v1	✔️	✔️	✔️
wlr-layer-shell-unstable-v1	✔️	✔️	❌

Required Protocols (Event Receiving)	Sway	Kwin	Gnome
wlr-virtual-pointer-unstable-v1	✔️	❌	❌
virtual-keyboard-unstable-v1	✔️	❌	❌
fake-input	❌	✔️	❌

The wlr_layer_shell protocol will likely never be implemented in Gnome, so a Gnome-Shell extension is probably the way to go here.

In order for layershell surfaces to be able to lock the pointer using the pointer_constraints protocol this patch needs to be applied to sway. (this works natively on sway versions >= 1.8)

For the receiving end, Gnome uses libei for input emulation, which might be the better approach in general moving forward (TODO).

Build and run

First configure the client / server in config.toml. (A misconfiguration currently does not produce a very informative error message)

Build only

cargo build

Run

cargo run

TODO

• Capture the actual mouse events on the server side via a wayland client and send them to the client
• Mouse grabbing
• Window with absolute position -> wlr_layer_shell
• DNS resolving
• Keyboard support
• Scrollwheel support
• Button support
• Latency measurement + logging
• Bandwidth usage approximation + logging
• Multiple IP addresses -> check which one is reachable
• Merge server and client -> Both client and server can send and receive events depending on what mouse is used where
• Liveness tracking (automatically ungrab mouse when client unreachable)
• Clipboard support
• Graphical frontend (gtk?)
• Encrytion
• Gnome Shell Extension (layer shell is not supported)

Protocol considerations

Currently all mouse and keyboard events are sent via UDP for performance reasons. Each event is sent as one single datagram, currently without any acknowledgement to guarantee 0% packet loss. This means, any packet that is lost results in a discarded mouse / key event, which is ignored for now.

UDP also has the additional benefit that no reconnection logic is required. Any client can just go offline and it will simply start working again as soon as it comes back online.

Additionally a tcp server is hosted for data that needs to be sent reliably (e.g. the keymap from the server or clipboard contents in the future) can be requested via a tcp connection.

Bandwidth considerations

The most bandwidth is taken up by mouse events. A typical office mouse has a polling rate of 125Hz while gaming mice typically have a much higher polling rate of 1000Hz. A mouse Event consists of 21 Bytes:

• 1 Byte for the event type enum,
• 4 Bytes (u32) for the timestamp,
• 8 Bytes (f64) for dx,
• 8 Bytes (f64) for dy.

Additionally the IP header with 20 Bytes and the udp header with 8 Bytes take up another 28 Byte. So in total there is 49 * 1000 Bytes/s for a 1000Hz gaming mouse. This makes for a bandwidth requirement of 392 kbit/s in total even for a high end gaming mouse. So bandwidth is a non-issue.

Larger data chunks, like the keymap are offered by the server via tcp listening on the same port. This way we dont need to implement any congestion control and leave this up to tcp. In the future this can be used for e.g. clipboard contents as well.

Security

Sending key and mouse event data over the local network might not be the biggest security concern but in any public network or business environment it's QUITE a problem to basically broadcast your keystrokes.

• There should probably be an encryption layer below the application to enable a secure link
• The encryption keys could be generated by the graphical frontend