W3C home > Mailing lists > Public > www-ws-desc@w3.org > October 2006

Re: Synchronous v/s Asynchronous, a WSA question, and few suggestions

From: Anish Karmarkar <Anish.Karmarkar@oracle.com>
Date: Wed, 04 Oct 2006 22:49:32 -0700
Message-ID: <45249CEC.2090108@oracle.com>
To: Ramkumar Menon <ramkumar.menon@gmail.com>
CC: www-ws-desc@w3.org

Ramkumar Menon wrote:
> Gurus,
>  
> I am troubled by a few questions since a few days. Appreciate your 
> comments in this matter.
>  
> a) What defines "synchronous" or "asynchronous" - Are this 
> terms normatively defined in any related specification ?  Are these 
> terms "adjectives" for message exchange patterns? 

That is hard to define in the abstract.
I agree with Tony's definition wrt to a synchronous protocol such as HTTP.

> b) Does "synchronous" interaction require the caller program to 
> "block" on the response from the service ? [I know it really depends on 
> what (a)'s answer is]

Never, IMHO.
Blocking/non-blocking is quite independent of synchronous/asynchronous.
One can block on a synchronous interaction or not; one can block on an 
asynchronous interaction or not. This is assuming that the response is 
going to go to the 'same' place.

> c) Can't a Request/Response transmission primitive [that wsdl 1.1 
> defines]  be asynchronous [thru wsdl extensions] ? The specification 
> does not talk about the relationship between synchronicity and the type 
> of the transmission primitive. So I assume this does not violate the 
> conformance requirements.

Yes. In the SOAP/HTTP case, one just has to specify the appropriate 
value for wsa:ReplyTo MAP.

> d) Is it correct if I state that synchronicity and asynchronicity of an 
> interaction cannot be defined at the abstract level of a service 
> definition and depends purely on the transport bindings being used ? [or 
> if the user employs extensions for the operation in the abstract portion 
> of the WSDL] 

The async task force spend quite a bit of time on this and IIRC there 
was never any agreement on the definition of what synchronous or 
asynchronous means. It all depends on which layer you are at.

> e) What is the relationship between the transport specified in the WSDL 
> 1.1 Bindings and the ReplyTo/FaultTo requirements that are imposed on 
> the service ? - confusing question, aint it ? :-)
> Let me explain with an example.
>   I define a WSDL 1.1 service with a portType with a 
> request/response/fault operation. I define WS-A headers for each of the 
> input, output and fault in the binding section.  I wish to return faults 
> from this operation to a FaultTo endpoint that is different from the 
> ReplyTo endpoint. Shouldnt it be possible to send messages to the 
> FaultTo endpoint on a different transport ? [ i.e. Lets say I wish to 
> send faults to an Email Address]. Question is as follows - If so, would 
> this require separate bindings for the operation to be defined within 
> the WSDL ?". If this answer to this question is "yes", then, since 
> transports are specified at the operation level, this would require two 
> bindings, one for http and one over smtp.  And in the second binding, 
> what does it mean to specify the information for the input and output - 
> they are unused. :-)  I am gonna get killed for making this assumption , 
> but I am really confused on this last point - I maybe wrong here. Or 
> maybe WSDL 1.1 is tough to gel in with WSA requirements without relying 
> on some extension mechanism.
>  

WSDL 1.1 (and WSDL 2.0) does support specifying bindings at the message 
level. Specifying promiscuous transport bindings is hard. AFAIK, the 
only way to do this is to specify a binding (i.e., create your own that 
says input is over HTTP and output over SMTP) or a binding extension 
that specifies the transports or says that the transport can be SMTP or 
HTTP.

HTH.

-Anish
--

> Few other points:-
> a) I noticed that Figure 2-1 [xml infoset] in Section 2.2.1 in WSDL 2.0 
> primer states that an interface should have 1-* number of operations. 
> This should be changed to 0-*.[since there could be interfaces with zero 
> operations, for instance, an interface that just defines faults]
> b) Section 2.9.1 in the Core Language Spec states that "A Binding 
> component that defines bindings for an Interface component MUST define 
> bindings for all the operations of that Interface component".  Shouldnt 
> a similar assertion be made regarding the Faults declared in the 
> interface as well? i.e.  "A Binding component that defines bindings for 
> an Interface component MUST define bindings for all the faults of that 
> Interface component"
> c) An interesting thought [on wsdl 2.0] - Why cant faults be global to a 
> description - I have a scenario where the wsdl defines two interfaces - 
> one for reserving flight tickets for the travel, and one for making 
> hotel reservations for the travel.Each of these interfaces are served by 
> two separate endpoints [lets say, outsourced to two different organizations]
> Both of them throw a fault namely "CreditCardAuthorizationFault" and a 
> "InsufficientFundsFault". Why cant this fault be declared globally, and 
> referenced within each of the interfaces ? [I'm being too impractical, 
> aint I ? :-) ] - But would definitely appreciate an explanation to this 
> point.
>  
> rgds,
> Ram
Received on Thursday, 5 October 2006 05:51:01 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:58:42 GMT