Sadly, the author failed to read on the history of kdbus and AF_BUS.
Just using UNIX domain socks is not the solution. This is more or less what dbus currently does and has major issues with memory accounting and trusts.
Just using UNIX domain socks is not the solution. This is more or less what dbus currently does and has major issues with memory accounting and trusts.
It does seem to me that a "pipe with fanout", or similar new type of socket, might be the best way of doing this.
Or if you really want to annoy people, copy the Windows PostMessage() system.
There are various other properties of Bus1 that make it better than all other alternatives, but obviously relying on it makes your program non-portable. IPC is a mess on POSIX systems.
On the contrary, I've been following their development from the beginning, with some interest.
> This is more or less what dbus currently does
Actually D-Bus multiplexes all messages, so they all pass through the message bus daemon. What I'm proposing is clearly different.
> and has major issues with memory accounting
This is a valid technical point, though I'm not sure it's ever been a problem in practice and I'm not convinced it needed a whole new IPC mechanism to solve.