Sanjiva Weerawarana wrote:

>On Tue, 2005-09-13 at 09:55 +0100, Keith Harrison-Broninski wrote:
>  
>
>>In fact, when I was trying to get RADRunner to do dynamic invocation
>>of Web services with complex parameters, the only way I could find was
>>to use a piece of technology from IBM's "Alphaworks" - their testbed
>>for potential new technologies, that are open for anyone to try out.
>>This technology, known as "JROM", claimed to be able to make the
>>necessary programmatic objects from arbitrary XML documents.  Used in
>>conjunction with another piece of technology from IBM that they gave
>>to Apache (WSIF), it seemed as though JROM provided the missing piece
>>of the picture:
>>
>>http://ws.apache.org/wsif/faq.html#faq-N100E5 (Sample DynamicInvoker
>>and Complex Type handling in WSIF)
>>
>>So, I started badgering people in IBM about a commercial license to
>>JROM.  I even met with a key player in the development team.  Several
>>years later, JROM has not moved on since the last release in 2002 (I
>>looked again yesterday) - and IBM have neither offered me a commercial
>>license, built JROM into a commercial product of their own nor donated
>>it to Apache (the latter approach being the natural outcome that the
>>people I spoke to seemed to expect).
>>    
>>
>
>Paul Fremantle and I originally created WSIF and I'm the one who started
>JROM stuff with Rania Khalaf. 
>
>The reason JROM never made it past alphaWorks is because of
>techno-politics. The techno part is that JROM is for XML only and what
>IBM wanted was a unified data model for *any* kind of data, not just
>XML. That's what SDOs are (http://www.jcp.org/en/jsr/detail?id=235) ..
>but the reality is that trying to unify the world results in much
>complexity and one loses the simplicity of JROM rapidly.
>
>  
>
>>Why?  I think IBM are aware that JROM does not work.  Further, I think
>>there are technical reasons why it cannot work.  There is just not
>>enough information in an XML document to tell you how to convert it
>>into an object.
>>    
>>
>
>Ah therein lies the fallacy. There's no inherent reason whatsoever that
>you need to take every bit of XML and map it to an object. In fact, XML
>Schema has all kinds of stuff which doesn't map naturally to normal
>language type systems. 
>
if you just need to invoke WS with compelxTypes inside WSDL you can send 
XML message directly (really not much work!) - the only difficulty is if 
you do want to have some kind of UI such as, say, creating dynamic XHTML 
Form to represent input parameters - if you desire to do so then take a 
look on Xydra which is doing exactly this and is under BSD-like license 
(research prototype!):
http://www.extreme.indiana.edu/xgws/xydra
Xydra has internal XML infoset model, simple XML Schema parsing library 
and way to generate XML instance documents based on XML schema and 
round-trip XHTML form name-value pairs to XML and back.

>>What this means, is that a key vision underpinning the whole WS effort
>>- of a world of freewheeling enterprise applications that dynamically
>>pick up and use Web services discovered in a registry - may never be
>>achievable.  The only way to do it is to use a specific technology as
>>your client (e.g., a Java framework such as WSIF) and prime it at
>>development time (not runtime) with the information necessary to call
>>specific services (the mappings from their complex type parameters
>>into programmatic objects).
>>    
>>
>
>I don't agree! In Axis2 (http://ws.apache.org/axis2) we have built a new
>data model called AXIOM which is used internally for all XML stuff. I'm
>hoping to re-do WSIF one of these days using that as the internals ..
>think of it as a much improved JROM. (JROM's difference was that it was
>schema aware natively- AXIOM isn't but it has sufficient flexibility to
>compensate for that plus we do have a data binding part too to handle
>that fully). In any case, we haven't ruled out adding native simple type
>support into AXIOM.
>
>Once we have that done, with Axis2 modules etc., you will be able to
>invoke a service fully dynamically, engage security/RM etc. dynamically
>(even negotiate those policies dynamically). If all goes well we'll have
>all this working this year.
>
>There's nothing wrong with the WS vision IMO. The problem is with the
>middleware that's been developed for it so far - its all been band-aid
>over some current stuff. To make WS real you need a new kind of
>middleware ... which is what is being built in Apache now!
>
>Indigo is also native WS middleware. That's the only thing that comes
>close to Axis2's approach now (IMO :)).
>  
>
XSUL2 <www.extreme.indiana.edu/xgws/xsul> [1] has all of this stuff 
(including improved WSIF APIs, more dynamic AXIOM-like API, handlers for 
security) and more - but it is also research OSS code (under BSD-like 
license) so it has no industry heavyweight to support it but just few 
graduate students ...

so yes i agree - the part to blame is middleware - nothing wrong with 
basic WS-* (XML, HTTP, SOAP, WSDL) - they work great as long as not too 
much is hidden in name of "XML is ugly" and when code generators/tooling 
do not go wild with complexity generation/hiding ...

best,

alek

[1] http://www.extreme.indiana.edu/xgws/xsul/guide

-- 
The best way to predict the future is to invent it - Alan Kay






------------------------ Yahoo! Groups Sponsor --------------------~--> 
Most low income households are not online. Help bridge the digital divide today!
http://us.click.yahoo.com/cd_AJB/QnQLAA/TtwFAA/NhFolB/TM
--------------------------------------------------------------------~-> 

 
Yahoo! Groups Links

<*> To visit your group on the web, go to:
    http://groups.yahoo.com/group/service-orientated-architecture/

<*> To unsubscribe from this group, send an email to:
    [EMAIL PROTECTED]

<*> Your use of Yahoo! Groups is subject to:
    http://docs.yahoo.com/info/terms/
 


Reply via email to