> On 2017/08/10 17:35, ed mcmurray <emcmur...@aevee.com> wrote: > > From within the MPD application, I would like to access a Linux Named Pipe > > FIFO that has been created in a second application. The second application > > would place a STOP command into the Name Pipe, when appropriate. I would > > like to insert code in MPD to then monitor this Named Pipe for the STOP > > command and act on the command as though it was sent from an mpd > > client. > > But ... why? > > You can already do that with sockets, which is the core idea of MPD. > Why pipes? Pipes have serious disadvantages (such as being > unidirectional), but you did not describe any advantages of using them > instead of sockets.
The overhead of unix sockets during setup (you need socket()+bind() system calls, against open() in case of a FIFO) is little compared to the overhead of using them, as in order to successfully write() to a fifo, the read() must complete. Unix sockets, as any other sockets, will buffer in kernel. I agree that unix sockets are superior to fifo's in this matter. Kind regards, Kurt _______________________________________________ mpd-devel mailing list mpd-devel@musicpd.org http://mailman.blarg.de/listinfo/mpd-devel