RE: Protocol independence

A more interesting discussion would be, in my opinion, the analysis of properties and features "carried" by the underlying transfer protocol. I am talking about the fact that, for example, SOAP over HTTP might require the use of a headers-supported reliability mechanism, but the same SOAP message carried over JMS might not, assuming the underlying JMS provider supports reliable messaging. 

Things can get more complicated when the same message travels across intermediaries, where each hop might use different transfer protocols. How does the intermediate node figure out how features and properties are to be translated between hops? In cases like reliable messaging, it might be decided that, even though the JMS hop provides reliability, the presence of other hops whose transfer protocol does not support reliability might still dictate the adoption of end-to-end headers-supported reliable messaging.

My hope is that these issues will be addressed by the WSDL Properties & Features Task Force, or by one of the Policy consortiums. Otherwise, this looks to me like a good subject for the WSA group.

Ugo

> -----Original Message-----
> From: Champion, Mike [mailto:Mike.Champion@softwareag-usa.com]
> Sent: Tuesday, April 08, 2003 3:39 AM
> To: www-ws-arch@w3.org
> Subject: RE: Protocol independence
> 
> 
> 
>  
> 
> > 
> > Does it mean the same thing to transfer a file over HTTP PUT, 
> > as it does to transfer it over HTTP POST, or NFS WRITE?  No.
> > 
> > Does it mean the same thing to send 'getStockQuote("sunw")' 
> > over BEEP, TCP, IIOP, or MOM?  Yes, it does.
> > 
> > The difference is that NFS and HTTP are application 
> > protocols, and BEEP, TCP, IIOP, and MOM are not.
> 
> Does any current member of the WG agree with Mark that the 
> "meaning" of a
> SOAP message is affected in any way by the mechanism by which it is
> transferred between nodes?  I'd like to put this to rest if we can, or
> figure out how to come to an acceptable resolution if there 
> is room for
> discussion.
> 
> As for the "application protocol" issue, I think we disposed 
> of that in our
> response to Mark's last issues. Of course we can discuss the 
> implications on
> www-ws, but I think that this trout has been filleted, 
> sauteed, and consumed
> as far as WSA is concerned.
> 
>

Received on Tuesday, 8 April 2003 13:58:12 UTC