+1. Client and server should also be completely arbitrary to make the system
loosely coupled. Otherwise, we'll just end up with just a rewrite of the 
old system.

H.Ozawa

Robin wrote:
> Services operations in a real world will quite always return an
> element that is described in XML schema notation using a complex type.
> If the developer uses a tool that is marshalling/unmarshalling the XML
> element into an object, this marshalling/unmarshalling should not
> tightly couple the Object model(=value object model) to the XML
> schema. If you do this, it means any change in any XML element will
> potentially break your client.
> <http://blogs.ittoolbox.com/eai/applications/archives/the-web-services-wise-man-9598>
>
> I would strongly advise not to use this kind of
> marshalling/unmarshalling tool unless it is change-tolerant.
> For example the mapping between the XML and the Object model can be
> configured without changing the code.
> Robin
>
> --- In [email protected], Keith
> Harrison-Broninski <[EMAIL PROTECTED]> wrote:
>   
>> You're missing Viju's point, Steve.  Of course it's easy to build 
>> services that utilize complex objects.  Viju's question - a deep and 
>> important one, imho - is about limitations on the client, not on the 
>> server.  I hope Viju won't mind if I paraphrase it:
>>
>> /If a service returns a "complex object" - i.e., an object whose 
>> properties are not all primitive - then what are the limitations on the 
>> client?/
>>
>> Many services are produced on the assumption that the consumer uses a 
>> code base related to that of the producer - including, for instance, 
>> interfaces that can be used as proxies for the objects returned by the 
>> service.  If this is true, it makes a mockery of a key part of the SOA 
>> concept, namely the supposed independence of producer and consumer.
>>     
>  But 
>   
>> what client technologies allow you to call services from anywhere, 
>> without access to the producer's code base, and utilize
>>     
> programmatically 
>   
>> the complex objects that are returned?
>>
>> I have asked exactly this question before to this group 
>>
>>     
> <http://tech.groups.yahoo.com/group/service-orientated-architecture/message/2400>,
>
>   
>> since I have been looking for solutions since 2001.  And the _only_ 
>> practical suggestion I got was from Aleksander Slominski, whose XSUL 
>> project <http://www.extreme.indiana.edu/xgws/xsul/> includes a "Super 
>> Dynamic Invoker" (based on WSIF, which I also used in an earlier
>>     
> attempt 
>   
>> to deal with the problem).
>>
>> As you say, Steve, "there should be no limitations" - but there 
>> certainly appear to be at the moment.
>>
>> -- 
>>
>> All the best
>> Keith
>>
>> http://keith.harrison-broninski.info
>>
>> Steve Jones wrote:
>>
>>     
>>> 1) Service is a business concept
>>>  
>>> So I'm assuming you are talking about WS-*.  If you can't return 
>>> complex XML schemas from your WS-* implementation then quite frankly 
>>> its rubbish.  There should be no limitations.
>>>  
>>> Steve
>>>
>>>
>>>  
>>> On 06/12/06, *darlin' Viju* <[EMAIL PROTECTED] 
>>> <mailto:[EMAIL PROTECTED]>> wrote:
>>>
>>>     All,
>>>      
>>>     Is there any limitations for the return type of a service ?
>>>     My understanding is the service should return only primitive types
>>>     and objects which has primitive types as variables. This is to
>>>     ensure that the service can be used across languages.
>>>     Is this correct or is it like there is no limitation at all?
>>>     Is it posible to return an xml from a service?
>>>      
>>>     Thanks,
>>>     - d Vj
>>>
>>>    
>>>       
> ------------------------------------------------------------------------
>   
>>>     Check out the all-new Yahoo! Mail beta
>>>    
>>>       
> <http://us.rd.yahoo.com/evt=43257/*http://advision.webevents.yahoo.com/mailbeta>
>   
>>>     - Fire up a more powerful email and get things done faster.
>>>
>>>
>>>
>>>       
>
>
>
>   

Reply via email to