This is a subtle issue that very few document creators will
understand
True, but then the argument against text/html is also rooted in
such subtlety?
The subtlety doesn't lie in the intent, but in the difficulty of achieving it.
You can construct HTML that can be read without an HTML processor of any sort.
But few if any agents that emit text/html actually do this. Instead they emit
junk that is quite difficult to read even if you're familiar with HTML, and
nearly impossible if you aren't.
It's much better to pick one MIME type for each media type then to try to
pass on the decision to folks who won't understand or won't care about the
subtlety. That one type, IMHO, should almost always be application/foo+xml
not text/foo+xml.
In that case it would seem much better to define application/foo rather
than application/foo+xml
No. There's a specific purpose to the +xml suffix and it should be used
when it is appropriate.
XML is a complex object type (as is text in general), the question is where
that complexity best resides, and where one desires interoperability. I
would argue that text/xml (as in my example) is useful, but not sufficient
because it doesn't capture everything needed to process the content. This
is one of the problems that XML brings to the fore because it can represent
so many things, and be interpreted in so many ways (MIME is built around
the assumption that the media type canonically defines how it should be
processed).
This isn't necessarily true in light of the definition of the +xml suffix. In
addition, there is more to it than this. Top level types in particular are
intended to tell you generally how something is or can be processed. This can
assist an agent in handling material even when a specific definition isn't
available.
To get back to ground zero, one must ask why one is exchanging XML in the
first place. If is as a structured peice of text, then text/* is fine.
No it isn't, because that's not what the top-level text type is for. The intent
of the text type is to label material that can sensible be presented to people
without any processing.
You seem to think that the top-level text type label is intended to label
structured text, but that's just not what it is for.
More
often than not, this will *not* be the case, and so there is a good argument
for application/*. In cases where XML content is being served with additional
defined processing, the message type will probably be a multipart message,
with one part being labelled as text/xml (another might be application/xsl
or text/css).
Bottom line: XML in well-defined cases should use XML, otherwise it will
probably use a multipart message.
The +xml suffix is intended to make XML detectable even when the specific
type isn't known. This eliminates the need to use application/xml to
insure such processing is possible.
Ned