Re: MPEG asks for MIME review for the MPEG21 file format

[Martin Duerst <[e-mail]>]

At 14:54 07/05/18, Anne van Kesteren wrote:
>
>On Fri, 18 May 2007 03:12:09 +0200, Chris Lilley <[e-mail]> wrote:
>> The successor to RFC 3023 needs to indicate that binary XML which is  
>> presented as a new encoding (in the xml sense) can use +xml, while other  
>> binary forms cannot.
>
>You wouldn't be able to still parse the retrieved resource in that case  
>with a generic XML parser. Wasn't that the whole idea of +xml?

Well, yes, but the language lawyers argue as follows:

Except for UTF-8 and UTF-16, there is absolutely no guarantee
that an XML parser accepts any encoding whatsoever. There is
a lot of XML out there with e.g.
   <?xml version='1.0' encoding='Shift_JIS'?>
but no XML parser is required to grok that (although many do).
So you can view binary XML just as an extremely weird and special
character encoding. I personally wish it wouldn't be necessary,
but there are people who claim that it is, for whatever it's worth.

The successor of RFC 3032 shouldn't mention binary XML explicitly
(except by example), but just say that the boundary between what's
okay and what's not okay for a +xml prefix is whether the document
starts as an XML document as described in appendix F of the XML
Rec, and can correctly be parsed after the conversion from that
'character' encoding to XML.

What I also would like to see (in particular in my role as
one of the IETF charset reviewers) is how to make sure that
the labels for binary encodings won't in the future collide
with real registered charsets. Probably the best way to do
this would be to create some kind of 'provisional' registration,
or 'reserving' registration in the charset registry.
But that would need some work on a spec, at the moment,
probably the best way to do things would be to try and
get some kind of comment into the charset registry.

Regards,    Martin.



#-#-#  Martin J. Du"rst, Assoc. Professor, Aoyama Gakuin University
#-#-#  http://www.sw.it.aoyama.ac.jp       mailto:[e-mail]