Re: FYI: I-D ACTION:draft-dusseault-http-patch-02.txt

On Jul 9, 2004, at 12:38 PM, Roy T.Fielding wrote:

>
> URI is served).  Most servers actually define the media type according
> to the binding used to retrieve it, rather than an aspect of the file
> itself, ...

This may not now be relevant to the PATCH discussion, but I wanted to
check my understanding here.  Here's the statement I would have made:

   "Most servers have nearly complete independence between the
   media type of a resource and the binding used to retrieve it. 
Interactions
   between the two are mostly limited to resource creation time when if 
the
   client erroneously omits the Content-Type header, or if the content is
   importedthrough a mechanism other than HTTP, the server may have to
   guess a default MIME type to assign and may use the binding 
(particularly
   the filename extension) to make a good guess."

Obviously this is nearly the opposite of what you said Roy.  When you 
said
"according to the binding" what part of the binding did you mean? The
namespace or the file extension, or other?  If by namespace, do you have
an example of this?

I've been involved in several HTTP/WebDAV server implementations and
we always implemented the MIME type and binding(s) as independent bits
of resource data which both needed to be stored.  Xythos' 
interoperability
experience supported that choice.  When we worked with clients uploading
files without providing a MIME type (or providing a generic MIME type 
as the
Mac OS X client sometimes does) we found interoperability problems as
other clients relied on the MIME type, not the file extension.  And the 
server
couldn't always make a good guess based on the extension, and certainly
had no help based on any other part of the binding as this was in a file
sharing or document management environment where the namespace was
almost all arbitrary.

Lisa

Received on Monday, 12 July 2004 15:05:37 UTC