I see I’m in a minority here. I would merge all digitally silent recordings.
And in my opinion, there can be no artist for silence.
About 4′33″: I think recording of that work would not be silence, but background noise. If the recording was digitally silent, the listener would not be listening the recording, but the background noise. If that noise is not recorded, it is not a recording.
Well, I think digital silence never is a recording.
While I say that, I still think that silence can be (and is) very important. Digitally silent tracks can have titles and track artists, but in my opinion, as they would not really be recordings (as nothing was recorded), they could as well be merged under the same MB recording.
But in my opinion, it wouldn’t be a good idea to just merge recordings marked [silent], as they can as well not be digitally silent.
There seems to be no AcoustID entries for digitally silent tracks (unless incorrectly linked). I tried this myself just now. The [silence] on Turisas: The Varangian Way album provides no AcoustID, while all the other tracks on that album get AcoustIDs when I Analyze them with Picard. But it seems that all the fingerprints for digital silences end with a long spring of letters “A” (based on the lenght of the track), when the actual fingerprint is calculated with fpcalc.
- AQAAfkmUaEkSRZEGAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA (18s recording)
- AQAA3UmUaEkSRZEGAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA (30s recording)
- AQABzEmUaEkSRZEGAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA (59s recording)
The recordings don’t even have to be digitally silent for them to get no AcoustIDs. I changed one sample of the 18 second recording and it gets the exact same fingerprint, resulting in no AcoustID.