The Internet (The I in ISP...) is far more than than the web. Mere HTTP/s access is suffocating, and we should not normalize this as a customer expectation.
In contrast to requiring a particular standard to represent metadata, the better way is to require that the exported data be machine-readable and its contents documented. This gives providers the flexibility to innovate, adding their own identifiers or whatever useful data they want, but avoids legislators forcing everyone to use a common standard that, while useful today, can quickly go out of date. Competitors can then write converters or importers based on this documentation, as could motivated FOSS users.
I agree on the merits of "pure", but I wouldn't call "auto" bloat. As a quality-of-life improvement, especially for writing generic code, it's pretty useful.