Re: [idn] IDNRA comments

["Maynard Kang" <maynard@pobox.org.sg>]

> I don't see where RFC 2047 encoded words came into the discussion,
> sorry.  QP and base64 are transfer encodings which work with any data,
> binary included.  They are used external to RFC 2047.  Dave didn't
> mention RFC 2047 format.

The only reason why I used RFC 2047 format to describe my examples was that
it is a convenient (and common) way of stating the charset together with the
string. It was just to show how the same abstract characters may result in
different strings given different CCS and not intended to be put forth as a
proposal (certainly domain names with "=?" is a very bad idea).

> It is up to the proposal to specify the CCS or
> even CES which is to be transfer encoded using QP, base64, or another
> transfer encoding scheme.  It is only because MIME never restricted the
> charset that multiple charsets are allowed.  If the proposal restricts
> the CCS, then the ambiguity of the CCS is not an issue.

I agree with you here. The CCS must be restricted for IDN. There was no need
for this restriction in MIME as headers do not have to be unique. Domain
names, on the other hand, should have this requirement.

maynard