RE: Registration of new charset "UTF-16"

[Larry Masinter <masinter@parc.xerox.com>]

> > What about end of line canonicalization?  Do we stick with CRLF, or should
> > we use the ISO 10646 Line Separator and Paragraph Separator characters?
> > Or do we give up on a canonical form and just state that widetext/etext
> > probably isn't suitable for use with digital signatures.

I don't think any of these alternatives are correct. A canonical form
is a choice among equivalent forms, where transformation among the equivalents
is allowed. Here are two choices that might be acceptable:

a)  treat widetext/etext as you treat application, for the purpose of
digital signatures: no transformations allowed.

b) specify that the 'canonical form' of "widetext/blah" is the transformation
into "text/blah;charset=utf-8" with CRLF end-of-line.

What do you think about a "utf-16" specific top level type, e.g.,

utf-16/html  == text/html;charset="utf-16"
utf-16/plain == text/plain;charset="utf-16"

no charset parameter allowed. It would simplify things; there aren't so
many 16-bit charsets.