On Jun 8, 2007, at 6:53 PM, Todd Biske wrote:
> On Jun 8, 2007, at 11:30 AM, Anne Thomas Manes wrote:
>
> >
> > If you strictly follow the method semantics defined in the HTTP 1.1
> > protocol, then it is RESTful. But HTTP 1.1 doesn't enforce those
> > method semantics. That would be considered "abuse", though.
> >
>
> Is it true that a system that makes extensive use of the POST
> loophole, but only does so where "the origin server is accepting the
> entity enclosed in the request as a new subordinate of the resource
> identified by the Request-URI" (from spec) would still be RESTful
> then?

This sound like "if a system uses POST the way it's supposed to, is  
it RESTful"? If this was the question I'd answer that yes, it is (at  
least in this regard).

> Many WS-* implementations would certainly not qualify because
> things that should be GETs are being tunneled as POST, correct?
>

Yes; even clearer than the loss of GET's benefits (e.g. caching) is  
the fact that typical WS-* applications usually expose a small number  
of "endpoints" as URLs, whereas in a REST approach, they'd expose  
thousands or millions of them - e.g. one for each customer, each  
order, each process, each account, each transaction ... lots of  
information added to the Web directly instead of hidden behind a  
small channel available to the initiated only.

> The POST loophole concerns me. Clearly, you need POST because you
> need the ability to make a relative change to a resource, but it
> seems that as long as some form of update is happening, you can push
> about anything you want through POST and still be considered
> RESTful. It seems that this would be very tempting for someone to
> take a very un-resourceful approach. Am I correct on this?

I'm not entirely sure we have the same understanding about POST (if  
at all, I'd rather map it to CREATE than UPDATE), but I would state  
again that the key difference between REST and WS-* (or POX, for that  
matter) is the primary usage of URIs to identify resources.

Stefan
--
Stefan Tilkov, http://www.innoq.com/blog/st/

>
> -tb
>
> 

Reply via email to