>Peter Laufenberg <open...@laufenberg.ch> wrote: > >> >That looks like the most difficult part, because offhand I have no >> >idea how to interface those input devices with a tty. >> >> The point is not to need tty or network client/server messaging. I query >> a USB device directly, never leaves the Geode. > >Well, on the box you still need to talk somehow to the MP3 playing >program.
Sure and the Geode has 2 USB ports (and a header for 2 more I think), one for the audio interface which may take a MIDI interface on input directly or if not a 2nd USB interface (gamepad, numpad, etc.) with a daemon. I could get reeeal fancy and use only the Alix's GPIO button with port-knocking-style sequence but it'd be silly. The radio stations are hardcoded and I hardly do any zapping; if both suck I turn streaming off and switch to my offline library which uses a completely different infrastructure. >> >mpg123 is probably the fastest one, although all of them will be >> >fast enough. >> >> Ok I saw mpg321 picked up that project and depends on madlib like a ton >> of other players. > >That's not how it happened. > >mpg123 is a floating point decoder. That code has also been used >in XMMS and MPlayer. libmad is a completely independent code base >and uses fixed point arithmetic. > >Years ago, when mpg123 was stuck at 0.59r, unmaintained and crufty, >and with a problematic license, somebody decided to throw a rough >clone together as a programming exercise by combining libmad and >libao with some glue and under the GPL--thus was born mpg321. >Eventually mpg321 was abandoned as well, but in the meantime mpg123 >had been picked up again and the project is alive and well today, >including a cleaned-up license. Lately mpg321 has also been revived, >sort of. > >Historically, there was also a bit of jostling over the patent >situation, where the mpg123 author said that MP3 patents probably >applied and the libmad author thought they might not because of the >fixed-point code, but by now they are probably expired anyway. I hear the Frauenhofer Institut is still swinging baseball bat about its mp3 patent; nothing like AAC but still... >Speaking as the guy who is the OpenBSD maintainer for both ports, >I very much prefer mpg123, and if you are concerned about CPU usage, >it should be the fastest MP3 decoder around (unless you are on an >architecture without floating point, i.e., ARM). Thanks a lot for the extensive information. -- p