On Sat, 13.06.09 17:07, Nix (nix at esperi.org.uk) wrote: > > On 27 May 2009, Lennart Poettering outgrape: > > > On Thu, 28.05.09 04:53, Patrick Shirkey (pshirkey at boosthardware.com) wrote: > > > >> I think it is useful that you have the internal api calls so dbus is not > >> a requirement for communicating with PA. > > > > I am sorry to inform you that eventually PA will use D-Bus for client > > communication too. > > What, no shared library? Did I say that? there'll always be a library for stuff like the channel mapping/sample spec/volume APIs. It makes no sense to make all those IPC interfaces. > Won't that break compatibility with, well, every single PA user? No. > Plus: anything can make function calls, but personally after one > experience too many with libdbus killing my entire application without > warning or notice because it disliked something I'd done, I'd rather > throw myself off a very high cliff than use libdbus in anything for > which crashing has bad consequences. Note that e.g. the X developers > came to the same conclusion regarding using dbus in the X server. If PA > moves to only allowing clients to communicate with it via dbus, then > applications that must not crash will have to stop talking to PA, it's > that simple. Ah, more D-Bus FUD. How I love it. Can't get enough of it. Lennart -- Lennart Poettering Red Hat, Inc. lennart [at] poettering [dot] net http://0pointer.net/lennart/ GnuPG 0x1A015CC4
