I still hope that subclassing native elements (i.e. `is=…`) will become
non-contentious and universally supported eventually, so it helps to
distinguish custom tags from type extensions (though this might be
outdated terminology)

this abstraction is only useful for the former, as we cannot dynamically
determine the base class (e.g. `HTMLFormElement` instead of
`HTMLElement`) without resorting to awkward mixins