fon* variants

[Frank Ellermann]

Michael Everson wrote:
 
> Generic variants like "western" are truly generic and are bad.
> "fonipa" is well-defined.

My definition for "generic variant" is "it doesn't come with a
list of prefixes where it makes sense".

Based on that all speculations about the meaning of a subtag for
16th century French in combination with Old English or Esperanto
are futile, like some nedis and rozaj cases discussed in BCP 47.

The fonipa definition is for humans.  It doesn't help to create
tools for the identification of garbage like "tt-Cyrl-fonipa":
Because fonipa has no prefix in theory anything goes.

With "1694" or similar it's clearer, it will be bound to some
languages including "fr" (not necessarily the same definition
for all prefixes).  And as long as Old English and Esperanto
aren't listed as prefixes for "1694" such combinations can be
sorted out as nonsense by a tool.

Frank