A radio broadcast is continuous signal and can therefore simply be translated into sound in real time.
An internet dataflow of a radio station is not real time and may be routed through multiple paths of varying delay to get to you.
Any non local radio station data source, if processed into sound without a buffer, would be susceptible to the variability of the flow of digital information to the Brennan unit. You would be able to hear this.
Therefore the Brennan starts by receiving the data stream and storing this in a small section of memory (a buffer) before actually starting to play this.
Variability in timing of how the music is coming to you therefore gets hidden from you by the fact that there is a "perfect" buffered section of music that is available to play.
The more variable the data flow packet timing, the the larger this buffer needs to be.
Therefore 98.2% buffer full is telling you how much of the allocated buffer memory space is in use, a filled buffer is an ideal condition. One would not want to see large fluctuations in this % measure, as this would indicate problems with the digital dataflow to the unit.
Fred.