I'm with Ian in the ticket discussion on image entities being undesirable. The thinking in this community is that CD-in-hand is the best reference, but accurate and direct scans are a close second. The issue with just about every other potential source is that most databases will fall back on a single standard image for every release in a group, or at the very least every release without a more specific cover, and doing so lose the small details we rely on; a while ago, me and another user were only able to verify we had different releases because one catalog number was truncated and one was oranger - all the rest of the printing we looked at was identical - and because of that one small difference I wound up adding a new release with new artwork.
If we decouple art from releases by moving it to its own entities, and ignoring everything that would affect on the CAA (which, though the reminder is likely unnecessary, is not technically part of MB), it by definition makes it so much easier for the same image to be linked to multiple releases. New users and tagging completionists - as opposed to those who prioritize data quality, and not that I hold it against either group - will abuse that ability to add art to the wrong release. Yeah, there's a small argument to be made about any images that are actually shared identically within a group, but it comes at the price of giving up one of the main things allowing us to be (ideally) more accurate than any of the other metadata sites.
We do still need to eventually have some in-schema way of associating images with particular media/tracks, but entities and relationships are not how I want to go about it. I myself am partial to something like Ian's second JSON example given in the third comment, with the addition of an optional
id field alongside the
type to contain the relevant MBID, though I could also see it going the other way and having the image ID linked from the track entity. Either way, it will need new UI to restrict the selection to only what's associated with the same release and I'm not sure if there's any lower-level way to ensure that. Would finally give the
track table some inarguable purpose, at least!
And, yes, I know it probably already groups recordings to their displayed titles; unless that's also where pseudo-release alternate tracklists get merged, though, I remain somewhat unconvinced that that couldn't be covered by bloating
medium a bit.