I forget if this question has come up yet.
There seems to be a glitch in RFC-1521 having to do with the definition of
subtype. The BNF for the individual MIME types require that the subtypes be
either one of the 1521-defined types or an extension-token. However, the
overall BNF permits a token for subtype.
This results in an ambiguity as to whether or not it is permitted to use an
unregistered MIME subtype that does not begin with ``X-''. Although we feel
that it is a terrible idea to use such things, personal feelings are not
adequate ammunition in filing a bug report.
So which is it? Either way, the BNF needs to be fixed. I don't think that it
is possible to have subtype as a generic BNF value; you have to deal with each
of the types individually to avoid ambiguity.