About performance of D-Bus

[Alberto Mardegan]

ext Jerome Philbert wrote:
> My D-Bus implementation is 16 times much slower than my actual socket 
> bus and 8 times much slower if D-Bus is configured in direct mode.

As it seems that communication via the bus daemon is 2 times slower than direct 
communication, has anyone considered having the system bus only as a register of 
running D-Bus applications/objects/interfaces, and have all the real 
communication done directly?

That is:

- application A registers object /A with the bus daemon
- application B calls AddMatch on the bus daemon
- the bus daemon calls AddDirectMatch (or something of that kind) an A, passing 
the "coordinates" of B
- AddDirectMatch returns
- the bus emits the reply of AddMatch, which also carry the "coordinates" of A
- when /A emits a signal, A emits it directly towards B (and all other listeners)
- when B wants to call a method on /A, it goes directly to A

Where by "coordinates" I mean all the information needed to have a 1-1 
communication between the processes.
I don't really know how feasible this would be, but it might be worth investigating.

Ciao,
   Alberto

-- 
http://www.mardy.it <-- Geek in un lingua international!