Re: PaceCategorize

["Roy T. Fielding" <fielding@xxxxxxxx>]

On Jun 30, 2005, at 12:51 PM, Sam Ruby wrote:

> Tim Bray wrote:
> > On Jun 30, 2005, at 9:59 AM, Joe Gregorio wrote:
> > > I like this train of thought, even allowing PUT and declaring that 
> > > any
> > > missing elements means don't bother updating that element.
> > Boy, that sounds too-good-to-be-true seductive.  -Tim
>
> I do believe it is too-good-to-be-true.  Example: How does one delete 
> the last category in an entry?

Atom cannot redefine the meaning of PUT.  Use PATCH instead.

> I also believe it violates the first SHOULD in section 9.6 of RFC 
> 2616.  http://www.w3.org/Protocols/rfc2616/rfc2616-sec9.html#sec9.6
>
> Perhaps Roy can correct me if I am wrong, or provide guidance on the 
> reasoning behind that SHOULD.

A server needs to know the intention of a client.  PUT is a very
generic operation that can be implemented by a very generic handler,
unless some joker decides to define application-specific variants
of the meaning of PUT in order to satisfy their own whims.  HTTP
says don't do that unless you have an extremely good reason.
I would say never do that, but that's just me.

PATCH is intended for this very purpose.  The method tells the
server the intention is to modify the representation and the
media type tells the server how to do so.  At least, that is how
it was originally defined in the early drafts of HTTP/1.1.
The "diff" media type could even be specific to atom (or XSLT).

....Roy