RE: WSD Requirements: add a requirement about safe and idempotent characteristics

IMHO, and despite what the theorist say, HTTP/Get is only idempotent if
the implementation of that method doesn't have any side effects. 
Now I'm not saying whether it's good or bad to have side effects on a
get, but nobody can dictate what people want their applications to do.

I would be against the annotation in wsdl for a similar reason. It
really is a feature of the  implementation and not the interface. 
This will be especially true for the verticals building up generic type
libraries. I'm just visualizing all the arguments in all the bodies over
safe vs unsafe (and dirty) operations. Far better would be a way to
express the semantics in a meaningful and digestible way. 

Martin.




> -----Original Message-----
> From: www-ws-arch-request@w3.org 
> [mailto:www-ws-arch-request@w3.org] On Behalf Of Mark Baker
> Sent: Tuesday, November 26, 2002 11:13 AM
> To: www-ws-arch@w3.org
> Subject: Re: WSD Requirements: add a requirement about safe 
> and idempotent characteristics
> 
> 
> 
> Hi Hugo,
> 
> On Mon, Nov 25, 2002 at 04:52:35PM -0500, Hugo Haas wrote:
> > I would therefore like to submit the following requirements to the
> > WSDWG:
> > 
> > 	The description language MUST be able to indicate the
> > 	characteristics of an operation as a Web interaction. The
> > 	description of a service MUST indicate whether an operation is
> > 	safe. A service SHOULD be exposed as a Web method, in a
> > 	similar way as the Web Method Specification Feature[2].
> > 
> > Any comments?
> 
> I personally don't think this is necessary.  It's not enough 
> for an application to know that an operation is safe or 
> idempotent, it has to know what the operation *means*, and 
> safety and idempotency are just a small part of that meaning. 
>  For example, the HTTP OPTIONS operation/method is safe and 
> idempotent, as is HTTP GET, but they mean something very 
> different.  Without the full meaning of the operation 
> available to them, intermediaries don't have enough 
> information to know whether they can cache responses, or 
> transcode, or whatever their task might be.
> 
> I think that WSD issue #64[1] covers your concern.  If it 
> were resolved such that HTTP methods, including GET, were 
> given their rightful status as WSDL operations, then that's 
> all a developer has to know, and they can write software that 
> uses GET as it's defined in RFC 2616; any "safe" or 
> "idempotent" flag would be superfluous.
> 
> Thanks.
> 
>  [1] 
> http://dev.w3.org/cvsweb/~checkout~/2002/ws/desc/issues/wsd-is
sues.html#x64

MB
-- 
Mark Baker.  Ottawa, Ontario, CANADA.   http://www.markbaker.ca

   Will distribute objects for food

Received on Tuesday, 26 November 2002 18:20:42 UTC