note possibility of implementation-dependent file:

[masinter]

( "file:" might need a special case pre-processor for
the web, with the result translated/normalized.)

File may end up being implementation defined. Perhaps that should be
listed as a possible outcome. See:

https://www.w3.org/Bugs/Public/show_bug.cgi?id=27518

https://mailarchive.ietf.org/arch/msg/apps-discuss/KyxFLw2FCwv7CSpiWAmLr8Ha7vc

There might be a lot of ways of slicing the domains of
interoperability, and perhaps something else other than "implementation"
(platform x file-system-type) might serve the community better.