RE: [ubl] Global vs. Local -- Gunther's Recommendation

["CRAWFORD, Mark" <MCRAWFORD@lmi.org>]

Concur with Eve's remarks.  After a weeks worth of discussion in San Diego with ATG2, I still remain firmly convinced that the Garden of Eden approach is the best one for a standards based organization.  Although ATG2 is leaning towards venetian blind, my sense is the arguments are mostly the same that we had when we first started discussing this issue.  My hope is that after due deliberations, they will understand why we have made the choice we have.  
 
Eve's rebuttal to Gunther's paper gives more than sufficient responses to the points that Gunther made.  Although there appear to be some minor technical issues that still need to be addressed, my sense is Gunther has failed to provide convincing arguments for changing our decision.
 
Mark

	-----Original Message----- 
	From: Eve L. Maler [mailto:eve.maler@sun.com] 
	Sent: Fri 3/14/2003 8:40 PM 
	To: Stuhec, Gunther 
	Cc: 'Jim Wilson'; ubl@lists.oasis-open.org; 'VANDAMME Frank'; 'Garret Minakawa'; 'Dave Carlson' 
	Subject: Re: [ubl] Global vs. Local -- Gunther's Recommendation
	
	

	I don't doubt that local elements are more OO-like in that they scope
	elements to a type (class).  But...I don't understand the reasons you've
	given for saying that local elements are therefore preferred.  (My lack
	of understanding may be entirely due to my not being an OO developer, so
	please forgive any clueless questions below.)
	
	What is a pseudo-class, and why must one be created for every global
	element?  Multiple global elements could be bound to a single
	type/class, just as multiple local elements with the same name could be
	bound to a single type: no explosion of types.  Are you saying that with
	OO processing you can't distinguish between multiple elements that are
	bound to the same type because you only ever look at type information?
	In that case, isn't the problem the same whether you have local or
	global elements?
	
	And what is the packaging/maintenance barrier to OO class-based
	processing of complex types when global elements are used?
	
	        Eve
	
	Stuhec, Gunther wrote:
	> Hello all,
	>
	> I have spoken about the topic "Global versus Global" with Garret Minakawa (Oracle), Frank Vandamme (SWIFT) and Dave Carlson. All three colleagues said to me that for implementation reasons will be the "Venetian Blind" design much more better as the "Garden of Eden" design. Because the "Venetian Blind" design aligns to the OO (UML) model and "Garden of Eden" does not.
	>
	> That means, every type in "Venetian Blind" represents an object class and this can be used in UML class diagrams as well as in object oriented-programming language in the same manner. In object-oriented programming places a great deal of emphasis on packaging classes according to their services. The types in the venetian blind design represents the package structure and this organizes the code and facilitates modularity and maintenance. Therefore, the components of venetian blind are highly cohesive and have a high coupling. We do see how this is very OO-like, and how an OO (UML) model might map directly to this design. This kind of principle will be used many XML-APIs, like JAXB.
	>
	> In "Garden of Eden" must be created a pseudo-class for every global declared element. And this means an additional maintenance and redundant modularity of the UML models as well as OO-programming. And this might be not directly the OO modelling effort. Although is some additional work necessary by developing with APIs.
	>
	> Kind regards,
	>
	>       Gunther
	--
	Eve Maler                                        +1 781 442 3190
	Sun Microsystems                            cell +1 781 354 9441
	Web Technologies and Standards               eve.maler @ sun.com
	
	
	----------------------------------------------------------------
	To subscribe or unsubscribe from this elist use the subscription
	manager: <http://lists.oasis-open.org/ob/adm.pl>
	


----------------------------------------------------------------
To subscribe or unsubscribe from this elist use the subscription
manager: <http://lists.oasis-open.org/ob/adm.pl>