Re: [EAI] Question about draft-ietf-eai-rfc5335bis-12.txt

[Frank Ellermann <[e-mail]>]

On 5 October 2011 21:45, Julien ÉLIE wrote:

Hi, I pick your easiest question:

> * Shouldn't a way to distinguish currently suggested UTF-8 from
>   a possible upcoming suggested encoding (UTF-32, UTF-128 or
>   anything that could be standardized in a few years) be provided?

All UTFs unsuited for MIME including those with NUL-octets like
UTF-32 will never be used for Internet messages.  All new Internet
protocols already MUST support at least UTF-8, that is an IETF
policy since 1998 stated in RFC 2277 (= BCP 18).

In theory UTF-7 and BOCU-1 are suited for MIME, but if you MUST
support UTF-8, and the protocols or document formats (= here the
message header fields) offer no way to negotiate something else,
it will be UTF-8.   It can be limited to subsets such as NFC; see
RFC 5198 for the details affecting all protocols remotely related
to Telnet (incl. WHOIS, SMTP, FTP, HTTP, NNTP).  Actually HTTP is
a special case, HTTP embraced Latin-1 before UTF-8 and RFC 2277
existed:  The HTTPbis WG has "fun" with this issue, but they are
certainly also trying to adopt UTF-8, not something else.

In other words, as far as the Internet is concerned we will get
UTF-8 everywhere, anything else would be science fiction.  In a
bold statement RFC 2277 predicts that this transition will take
at least 50 years.

> Is "message/global" the way to know UTF-8 is expected?  Isn't
> it necessary to mention UTF-8 somewhere (in a "charset" value)?

The message/global business is about the message header fields;
for message/rfc822 it is ASCII.  The content (message body) is
still default ASCII unless there is a MIME header field saying
something else:  Something else can be anything (not limited to
UTF-8) encoded for 7bit (MIME B or QP) or 8bit (for UTF8SMTP,
does not require an encoding for MIME-compatible charsets such
as UTF-8, BOCU-1, Windows-1252, etc.)  UTF8SMTP does not allow
"unencoded binary" or charsets such as UTF-16, UTF-32, or SCSU
in the message body.

> * Can an internationalized message contain a multipart
>   content-type using an UTF-8 "boundary" value?

Good question, without cheating (= looking into the EAI drafts
if they cover MIME Content-* header fields somewhere) I don't
know the answer.

-Frank