This library aims to take care of 90% of the boilerplate code required to write a Wayland back-end as a stand-alone compositor.
A Wayland back-end is a small, lightweight compositor that is only capable of displaying one full-screen client at a time. This client may be a simple full-screen client such as a simple animation or it may be another compositor.
The reason why I'm calling them back-ends comes from the idea of using the Wayland protocol itself as an input/output abstraction. Right now, the standard answer to "I want to write a back-end for platform Y" is to write it as a Weston plug-in. The problem is that the Weston back-end plug-in architecture is unstable and is not universal across compositors. My vision with Wayland back-ends is to provide a standard way for compositors and back-ends to talk. This would allow someone to write a stand-alone RDP or VNC server, for instance, and then use it with any any compositor that supports running nested inside another compositor with wl_fullscreen_shell.
Mozilla did that for web browser plugins. However, plug-ins are notoriously unstable. In order to solve this problem, every web browser has developed their own mechanism for running plug-ins out-of-process. I'm avoiding that whole mess by running out-of-process from the start. Also, the Wayland protocol already provides us with all the pieces we need for this.
In theory, running out-of-process is slower because you have to communicate over sockets and context switch between processes. However, I don't intend this to be used in extremely high-performance situations. In the case of something like a VNC or RDP server, the performance loss from the context switch pales in comparison to the cost of network communication.
This library is designed to make it easier to write these back-ends. There is a certain amount of work that has to be done in order to implement surfaces commit/attach semantics and to properly handle input. libwlb takes care of those nitty-gritty details for you so that the user can focus on the back-end.
libwlb is not designed to be a full compositor abstraction. Using libwlb comes with a few restrictions:
- Only one client may be connected at a time.
- Only one surface may be visible on a given output at a time.
libwlb does not have a full scene graph so you cannot implement a full desktop compositor with it. On the other hand, these assumptions keep the API simple to make it easier to use.
The libwlb project is currently a work-in-progress. There are a lot of pieces still missing and I'm sure there are lots of bugs. The provided X backend runs but everything should be considered pre-alpha.
The primary purpose of putting it on Github is to get feedback from the rest of the Wayland community and possibly a little help fixing those bugs.