Request for review of media sub-type registration request

[Tim Moses]

Thanks Mark.  I'll incorporate your suggestion in the next submission.
All the best.  Tim.

Tim Moses
+1 613 270 3183

-----Original Message-----
From: mark at coactus.com [mailto:mark at coactus.com] On Behalf Of Mark Baker
Sent: Thursday, August 14, 2008 5:07 PM
To: Tim Moses
Cc: Bjoern Hoehrmann; ietf-types at iana.org; David M'Raihi
Subject: Re: Request for review of media sub-type registration request

On 7/14/08, Tim Moses <tim.moses at entrust.com> wrote:
> Bjoern - Thanks for your suggestions.  I will incorporate them into
the next version.
>
>  In answer to your question regarding the file extension, the choice
of '.xml' facilitates generic XML processing.  RFC3023 talks about this
approach extensively.  In particular, an originator who does not have
access to a special-purpose application can produce reports and check
their schema-validity using only an XML authoring tool.  Likewise for a
recipient.

RFC 3023 offers mixed advice there.  It suggests that there could be
some generic processing triggered from the +xml suffix, but also warns
against assuming namespace dispatching on */xml.  It's the latter point
that's the concern here because the Apache mime.types file[1] (which is
reused by other Web servers) associates ".xml" files with the
application/xml media type, meaning that any file using that extension
will be served by that server with that type unless the default
configuration is changed.  As I doubt this is what your users would
want, I would recommend minting a specific file extension so that the
default behaviour is to use this new registered type: try filext.com.

 [1]
https://svn.apache.org/repos/asf/httpd/httpd/trunk/docs/conf/mime.types

>
>  In terms of 'magic numbers', I cannot find an example of a
registration request that contains magic numbers for xml documents.  I
would be happy to follow an example if you are able to direct me to one.

Some registrations pay lip service to it, but in fact there aren't any
magic numbers for XML content.  I'd just say "None".

Mark.