First, my apologies if there is some mistake in this report, I'm not very knowledgeable in MPEG audio version. What I understood is that mpegaudioparse will parse mpeg audio 1, layer 1 / 3. Additionally, it parses the 3 generations of MP3. Those a commonly named as 1, 2 and 2.5, which we map into mpegaudioversion as 1, 2 and 3.

First issue I found, is that mpegaudioversion 3 is rejected by the audioringbuffer. Now, on wikipedia this is considered non-standard, but it's widely used to transport human voice, as it reduces the bitrate. I think we enable support for this in the audioringbuffer, sinks can still control their template caps.

Second issue, while the code in mpegaudioparse can produce 1, 2 and 3, the template advertise only 1 and 3.

Third issue, lame and mpegaudioparse disagree. An example:

gst-launch-1.0 -v audiotestsrc ! audio/x-raw,rate=8000 ! lamemp3enc  ! mpegaudioparse ! fakesink

8kHz is only supported for generation 2.5, lame report mpgeaudioversion=2, mpegaudioparse will report 3 correctly. (maybe this should be filed seperatly ?)