RE: [ubl-cmsc] restriction and extension are both subclassing

[Matthew Gertner <matthew.gertner@schemantix.com>]

OK, at least I understand what you are getting at now. I guess this is a
valid (although to me counterintuitive) way of looking at this. I still
don't see the applicability to our issues. While extending a complex type
and restricting a simple type could both been seen as defining a subset of
the value set of the base type, in the first case we are extending this
value set with new values (which necessarily become part of the now broader
set), whereas in the second case we are not changing the value set, simply
drawing a new wall around it and giving it a name. How does this prove that
creating restricted types should be seen as a type of derivation?

Matt

> -----Original Message-----
> From: Burcham, Bill [mailto:Bill_Burcham@stercomm.com]
> Sent: Thursday, May 02, 2002 4:29 PM
> To: 'Matthew Gertner'
> Cc: 'ubl-cmsc@lists.oasis-open.org'
> Subject: RE: [ubl-cmsc] restriction and extension are both subclassing
> 
> 
> Ah, that's the other side of the question.  Sure, here's a 
> value in the set
> of instances of the derived type -- a manager:
> 
> <?xml version="1.0" encoding="UTF-8"?>
> <Manager xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance";
> xsi:noNamespaceSchemaLocation="emp.xsd">
> 	<id>2</id>
> 	<numberOfReports>15</numberOfReports>
> </Manager>
> 
> Here's that same manager used where an Employee is expected:
> 
> <?xml version="1.0" encoding="UTF-8"?>
> <Employee xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance";
> xsi:noNamespaceSchemaLocation="emp.xsd" xsi:type="Manager">
> 	<id>2</id>
> 	<numberOfReports>15</numberOfReports>
> </Employee>
> 
> Both these instances documents are valid.  Your question was:
> 
> > Can you provide at least
> > one example of a value in D that is also in B?
> 
> The value (described by the tuple: <id>2</id>
> <numberOfReports>15</numberOfReports>) is in D (Manager) and 
> is also in B
> (Employee).
> 
> -----Original Message-----
> From: Matthew Gertner [mailto:matthew.gertner@schemantix.com]
> Sent: Thursday, May 02, 2002 9:19 AM
> To: 'Burcham, Bill'; Matthew Gertner
> Cc: 'ubl-cmsc@lists.oasis-open.org'
> Subject: RE: [ubl-cmsc] restriction and extension are both subclassing
> 
> 
> Maybe we have different definitions of "subset". Can you 
> provide at least
> one example of a value in D that is also in B?
> 
> > -----Original Message-----
> > From: Burcham, Bill [mailto:Bill_Burcham@stercomm.com]
> > Sent: Thursday, May 02, 2002 4:16 PM
> > To: 'Matthew Gertner'
> > Cc: 'ubl-cmsc@lists.oasis-open.org'
> > Subject: RE: [ubl-cmsc] restriction and extension are both 
> subclassing
> > 
> > 
> > Sure.  (forgive me) base: Employee, specialization: Manager.  
> > Employee has
> > an employee identifier (id).  A Manager is also an Employee, 
> > but in addition
> > to an identifier, a manager has a number of direct reports
> > (numberOfReports).
> > 
> > Here is the schema (with elements declared of each type as well):
> > 
> > <?xml version="1.0" encoding="UTF-8"?>
> > <xs:schema xmlns:xs="http://www.w3.org/2001/XMLSchema";
> > elementFormDefault="qualified" attributeFormDefault="unqualified">
> > 	<xs:complexType name="Employee">
> > 		<xs:all>
> > 			<xs:element name="id" type="xs:decimal"/>
> > 		</xs:all>
> > 	</xs:complexType>
> > 	<xs:complexType name="Manager">
> > 		<xs:complexContent>
> > 			<xs:extension base="Employee">
> > 				<xs:all>
> > 					<xs:element 
> > name="numberOfReports"
> > type="xs:decimal"/>
> > 				</xs:all>
> > 			</xs:extension>
> > 		</xs:complexContent>
> > 	</xs:complexType>
> > 	<xs:element name="Employee" type="Employee"/>
> > 	<xs:element name="Manager" type="Manager"/>
> > </xs:schema>
> > 
> > Matt's question was:
> > 
> > >  Can you can provide an example of values valid in the base 
> > type but not
> > in the >  extended type?
> > 
> > Here's a value valid in the base type (Employee element is of type
> > Employee):
> > 
> > <?xml version="1.0" encoding="UTF-8"?>
> > <Employee xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance";
> > xsi:noNamespaceSchemaLocation="emp.xsd">
> > 	<id>1</id>
> > </Employee>
> > 
> > Here's that same value (same content), but given as the 
> content for an
> > element who's type is Manager (Manager element is of type 
> > Manager).  This
> > instance fails validation:
> > 
> > <?xml version="1.0" encoding="UTF-8"?>
> > <Manager xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance";
> > xsi:noNamespaceSchemaLocation="emp.xsd">
> > 	<id>1</id>
> > </Manager>
> > 
> > Take any example of derivation of complex types in XSD and 
> > they'll work the
> > same.  If you take the "set of all possible content for the 
> > base type" as B
> > and the "set of all possible content for any derived type" as 
> > D then D will
> > always be a subset of B.
> > 
> > -----Original Message-----
> > From: Matthew Gertner [mailto:matthew.gertner@schemantix.com]
> > Sent: Thursday, May 02, 2002 4:01 AM
> > To: 'Burcham, Bill'; 'Eduardo Gutentag'
> > Cc: 'ubl-cmsc@lists.oasis-open.org'
> > Subject: RE: [ubl-cmsc] restriction and extension are both 
> subclassing
> > 
> > 
> > I agree that extension of a base type can be described as 
> > specialization. I
> > heartily disagree that the resulting type has a value set 
> > that is a subset
> > of that of the base type. Can you can provide an example of 
> > values valid in
> > the base type but not in the extended type?
> > 
> > Matt
> > -----Original Message-----
> > From: Burcham, Bill [mailto:Bill_Burcham@stercomm.com]
> > Sent: Thursday, May 02, 2002 1:49 AM
> > To: 'Eduardo Gutentag'
> > Cc: 'ubl-cmsc@lists.oasis-open.org'
> > Subject: RE: [ubl-cmsc] restriction and extension are both 
> subclassing
> > 
> > 
> > Maybe these two references will help:
> > (From the category theory perspective) Start at section 2.3 
> > of this paper.
> > There are two examples in that section of specialization by adding
> > attributes (properties) to the subtypes -- and the 
> > explanation is from the
> > set-theoretic perspective.
> > (From Entity-Relationship Modeling perspective) Start at 
> > slide 56 in this
> > presentation.
> > To your point about the value of this "pure theory analysis" 
> > -- Matt G.
> > presented an alternative Specialization Architecture in the 
> > CMSC today --
> > the one described in his recent paper (Schema Adjunct Framework +
> > Schematron).  As a motivator for the value of that 
> > architecture, Matt said
> > essentially "restriction and extension are two very different 
> > things and the
> > way XSD tries to treat them both as derivation is just 
> wrong -- so we
> > shouldn't use XSD".  Sorry I didn't include that context with 
> > my post, but
> > as a rebuttal argument I'd say it isn't just "pure theory 
> > analysis" -- I'd
> > say instead that it is rebuttal.
> > 
> > -----Original Message-----
> > From: Eduardo Gutentag [mailto:eduardo.gutentag@sun.com]
> > Sent: Wednesday, May 01, 2002 5:47 PM
> > To: Burcham, Bill
> > Cc: 'ubl-cmsc@lists.oasis-open.org'
> > Subject: Re: [ubl-cmsc] restriction and extension are both 
> subclassing
> > 
> > 
> > > "Burcham, Bill" wrote:
> > >
> > > Since I couldn't get this thought out in the conference 
> > call today, I
> > thought I'd get in into the record this way:
> > >
> > > From the set-theoretic standpoint, XSD restriction of a 
> > simple type and
> > extension of a complex type are the same -- they both define 
> > a subset of a
> > base set. To use type terminology, they are both kinds of 
> > specialization in
> > the generalization-specialization paradigm.
> > >
> > > A simple type is single-valued -- think scalar. When we 
> > restrict a scalar
> > type we take away values from its domain. This means that the 
> > restricted
> > type has fewer possible values than the original (or base) 
> > type. We never
> > add values to the domain. An example of this kind of 
> > specialization is: base
> > type Integer, specialized type Whole Number. These are 
> > scalars and the set
> > of Whole Numbers is a proper subset of the set of Integers.
> > >
> > > A complex type is not single-valued. Instead it is 
> > comprised of one or
> > more (simple type or complex type). A complex type can be 
> > specialized in two
> > ways: 1) a constituent simple type can be restricted or 2) a 
> > property of
> > simple or complex type can be added. In both cases, the new 
> > type has fewer
> > possible (compound) values than the original type. Note: that 
> > transitivity
> > in the definition of complex type means that there is kind of 
> > a third way to
> > specialize a complex type: (3) a constituent complex type can
> > > be specialized.
> > >
> > >
> > 
> > I totally fail to see how a complex type that has been 
> > "specialized" by
> > having
> > a property of either type added has fewer possible values. I 
> > just don't see
> > it. No. Nope. What am I missing?
> > 
> > (and I have to confess that, even if the assertion were true, 
> > I fail to see
> > what is the value added of this kind of pure theory analysis 
> > - undoubtedly
> > the result of a mind that wears glazed eyes at all times and 
> > carries an
> > extra
> > pair just in case...)
> > --
> > Eduardo Gutentag               |         e-mail: 
> > eduardo.gutentag@Sun.COM
> > XML Technology Center          |         Phone:  (510) 986-3651
> > Sun Microsystems Inc.          |
> > 
>