Re: [relax-ng] Encoding declaration, MIME type

[EB2M-MRT@asahi-net.or.jp (Murata Makoto)]

>Would it be text/something or 
>application/something? Murata-san, any thoughts on this?

There are no good solutions.  Sigh.

However, I do have one suggestion.  We should strongly refrain from 
inventing another hack, since there are so many already.

I incidentally wrote a memo about the current situation.  Although it is 
still sketchy and may contain some mistakes, it is probably helpful.  This 
memo is linked from the second bullet in the following document.  

    http://www.asahi-net.or.jp/~eb2m-mrt/jump_makoto.html

Here are some observations about media types.

- text/*

MIME people believe that text/* should be used only when the message is 
readable for casual users (in other words,  only when fallback to text/plain 
makes sense).  Thus, text/* is inappropriate for RELAX NG schemas.

If we use text/*, the charset parameter is authoritative and its default 
is (accornding to the HTTP 1.1 RFC) ISO-8859-1.  Implementors frequently 
ignore the default and even the charset parameter, and has invented lots 
of hueristics.  

- application/* or other top-level media types

We can do whatever we would like to do.  However, I strongly oppose to 
another encoding declaration mechanism.



By the way, do we allow schemas in the non-xml syntax to reference to 
schemas in the XML syntax and vice versa?  If this is the case, do we rely 
on HTTP headers and use content negotiation?

Cheers,

Makoto