Earlier quoted context omitted.
Nothing should be under .git/ except things owned by git (or at least allowed for in an explicit way like the hook scripts). You would have to either add the features to git itself, or at least add to git the knowledge of and allowance for extra features like that. But not just toss non-git data in .git/ simply because the dir happens to exist.
> You would have to either add the features to git itself, That is exactly what I'm suggesting.
There are probably an infinite and always growing and changing number of repo-adjacent things similar to the handful of added features that github currently tacks on to git.
I think it doesn't make sense for git, already complicated enough, to try to do all of that other related stuff, however related.
Maybe in fact git is already doing too much meaningful metadata work directly itself, and should instead try to switch it's current metadata into some kind of generic interface that other software could hook in to better?
So rather than git managing, say, issues, git just manages data. Not merely a db, it would facilitate associating high level feature data like issues and ci and conversations with low level commit data. An issue tracker would just be one of many clients using the interface.
git itself should probably not know very much about the data other than how it is associated to some commit. IE, the actual meaning of the data is either only in the consumer, or perhaps is some agreed standard that multiple consumers adhere to and understand, or perhaps comes with it's own schema definition like dtd or protobuf. Or all of the above, a paid product could have data only it understands, and other software could use standardized data, all stored in git at the same time. Multiple different consumers could use the same generic interface to their own data.
Because these add-on related functions are probably not universal or done being invented. Tomorrow there will be 5 new things you want to attach to a repo besides what github provides today, and so without even knowing what they are, that's how I know it's wrong for git to provide the features itself.
But managing the data without caring what the data is, that could totally be a new built in git function, where you give git itself one new function and interface, and an open-ended number of new feature-providing consumers all just use the same underlying git feature in whatever way they want.
Other than that the only other new thing I think that might be right to add to git itself is a proper way to manage large binary assets.