Despite attempts in this thread to over-simplify this issue, there is a lot of subtlety here, and this part of the product design is deeply interconnected with others. I don’t think anyone is suggesting that we roll back “editing support” or “prefer file data”, right? Because that’s what would be required if we were to simply reverse this decision–refining the approach to album artists was a direct prerequisite to implementing those features.
Also, lets not try not to confuse “technical” concerns with “product” concerns. it was product-level feedback from this community that drove us to change how this works–the technology is equally capable of implementing whatever we come up with.
Fundamentally, we made a mistake in v1.0: the mistake was to eschew the (almost universally adopted) concept of “album artist” and instead use a list of links to main performers to represent the album’s “artist”. This has several serious product consequences:
- It fails to represent collaborative albums cleanly or predictably. See this thread for background.
- It forces all albums released by the same artist to spell that person’s name in the same way. This creates many misleading, confusing, and inaccurate situations.
- Data structured in this way can’t be cleanly unified with album artist data stored in file tags.
- It’s deeply unfriendly to grooming
A side-effect of that mistake was that in 1.0 we were sorting albums based on the name of the first artist in their (arbitrarily ordered) main performers list. Last-first sorting is an artist level concept, not an album level concept, but since we were using artists to drive album-level sorting, that feature came “for free” in the album browser. It was an accidental (but happy) consequence of a poor design decision.
In 1.1, we’ve corrected that mistake, and created a first-class concept of “album artist” via the “performed by” field, and that has enabled many improvements in the product. It was not a decision taken lightly, or without considering the consequences.
There is no good source for last-name-first sortkey data for the “performed by” field. Last-first sorting is no longer a natural part of the system–it’s something that needs to be built from scratch, in a different way, in the 1.1 model.
From a practical standpoint, sorting albums by an artist’s last name is clearly useful, but the task at hand is not “putting the old feature back”, it’s “designing a version of the old feature that works in the new model”. This is tricky, because it means doing something automatic to generate sortkeys for album artist data.
We’ve spent the past few hours trying to work out a scheme for re-building this functionality in the new model. Obviously, many common cases are simple to handle gracefully, but there are quite a few nasty edge cases here (some of which 1.0 also handled poorly).
We’re going to take a stab at re-building this feature for the next release. Hang tight, guys.