On Tue, Mar 25, 2008 at 03:45:54PM -0700, Steven Dake wrote: > Placing the callback in track would work, however the track function > would have to return a new handle type "dataconf_track_handle_t" so that > later that same handle's tracking could be stopped.
First, you could cancel by xml_request, or by function,data tuple, or whatever. But you are right,and this even leads to my other email, proposing you return the handle on the track() call rather than init(&handle); track(handle, request); At its core, I didn't realize you were doing it this way. I was assuming a global handle you'd open once. > >From an implementation perspective doing this would be highly painful, > but could possibly make the api user's life easier. But then if we were > to do that we may need no dataconf_handle_t type at all since there is > only one dispatch function. I think we're circling around the same thing. But it begs another question. I was assuming a global handle because having a handle-per-track leads to 100s of file descriptors. And now we're actually approaching places where efficiency is a problem. 100s or 1000s of file descriptors take kernel memory, especially if they have IPC buffers associated with them. Joel -- Life's Little Instruction Book #232 "Keep your promises." Joel Becker Principal Software Developer Oracle E-mail: [EMAIL PROTECTED] Phone: (650) 506-8127 _______________________________________________ Openais mailing list Openais@lists.linux-foundation.org https://lists.linux-foundation.org/mailman/listinfo/openais