If we go with the third option, I'd be slightly pressing the suggestion to parsers to allow for an option to restrict any of the `h-` values to be ones that can represent known values, namely those with defined specifications

This could be an evolving list, which is already done with Micropub extensions and new specifications of objects themselves. This also could break backcompat with those parsers that fallback to being a h-entry or the like, but I see that as very unlikely.