Re: [apps-discuss] Getting 3023bis, a.k.a. draft-ietf-appsawg-xml-mediatypes, moving

[Ned Freed <[e-mail]>]

> Ned,

> Leaving registration considerations aside for a moment, if I want to
> create a media type which is more specific than application/xml
> for my application, is it better to use

> application/xml;profile="urn:my.app"
> or
> application/vnd.my.app+xml ?

The latter, unquestionably.

> Because this library https://code.google.com/p/mimeparse/, available in
> 8 programming languages, supports parsing the former as XML, but not the
> latter.

Sorry, one library demonstates diddley-sqaut. And while it's nice to support
default handling of entities as XML, it's hardly a requirement.

What is a requirement is the ability to have different handling for different
types of entities. And MIME dispatch capabilities are invariably capable of
operating on the conent type. This is the case in hundreds if not thousands of
different pieces of code. So having a proper type label gives you the ability
to process different sorts of entities differently.

It's also true that a some of them support dispatching on parameter values.
But definitely not all. In fact I'd go so far as to say that most do not.

And yes, most don't support default handling of media type suffixes either. But
that's hardly surprising given they were properly standardized less than a year
ago.

>  I haven't really looked at other libraries, but I think I may
> have landed on that one because the implementation of MIME in java did
> not manage negotiation concerns (I don't remember).

> Also, browsers (that I have used anyway) will parse the former as XML,
> but not the latter.  Same for html, which is potentially quite useful, IMHO.

Again, the ability to be able to default to parsing XML is a nice to have
feature. Nothing more. I'll also note that the rules for content type
definitiions are very clear that the type and subtype are what you register as
specifying the sort of entity you'r dealing with. As such, absent a willingness
to reopen the recently revised rules for how type registration work, which I
very much doubt is on the table, your proposal here is a total nonstarter.

				Ned