This document is a NOTE made available by the W3 Consortium for discussion only. This indicates no endorsement of its content by the W3C, nor that the Consortium has allocated, is allocating, or will be allocating any resources to the issues addressed by the NOTE.
It replaces the previous version of the SOX language specification and represents the current, implemented version of the language.
This document is a submission to W3C from Commerce One, Inc.. Please see Acknowledged Submissions to W3C regarding its disposition.
Comments on this document should be sent to [email protected].
This document describes SOX 2.0, the second version of the Schema for Object-Oriented XML. SOX is a schema language (or metagrammar) for defining the syntactic structure and partial semantics of XML document types. As such, SOX is an alternative to XML DTDs and can be used to define the same class of document types (with the exception of external parsed entities). However, SOX extends the language of DTDs by supporting:
All of these features are supported with strong type-checking and validation. A SOX schema is also a valid XML instance according to the SOX DTD, enabling the application of XML content management tools to schema management.
SOX was initially developed to support the development of large-scale, distributed electronic commerce applications but is applicable across the whole range of applications of markup. As compared to XML DTDs, SOX dramatically decreases the complexity of supporting interoperation among heterogenous applications by facilitating software mapping of XML data structures, expressing domain abstractions and common relationships directly and explicitly, enabling reuse at the document design and the application programming levels, and supporting the generation of common application components.
Although SOX 2.0 retains many of the features of SOX 1.0, it represents an additional year of actual implementation experience. Commerce One has a working implementation of this language and will be releasing products based on it. Our goals in releasing this second version are two-fold:
From the markup world, the SOX proposal is informed by the XML 1.0 [XML] specification as well as the XML-Data submission [XML-Data] and the Document Content Description submission[DCD]. However many of SOX' requirements come from the distributed computing world and SOX features have been heavily influenced by the Java[JAVA] programming language.
As a SOX document is not defined by a DTD, the native XML 1.0 Doctype declaration is not appropriate for a document instance to declare SOX Schema information (note that this may not continue to be true once the W3C Schema WG defines an official mechanism). Therefore we have created the soxtype declaration , which mimics the XML doctype declaration, but using a processing instruction. The soxtype declaration must be the first statement in a document following the optional XML declaration and declares that the default namespace [XML-Namespaces] of this document is that of the given SOX Schema, just as a doctype declaration declares that an XML 1.0 document conforms to the given Document Type Declaration (DTD). The soxtype declaration is not currently intended to coexist in the same document with a doctype declaration, but there is no particular restriction prohibiting this.
The soxtype declaration includes two parts:
An example declaration would look like:
<?soxtype urn:x-commerceone:document:com:commerceone:schema1.sox$1.0?>
where the schema definition is located through resolving the urn urn:x-commerceone:document:com:commerceone:schema1:schema1.sox$1.0. The part following the "$" gives the version number.
A small sample document would look like:
<?soxtype urn:x-commerceone:document:com:commerceone:schema1.sox$1.0?> <Root> <Body/> </Root>
The root element of the instance is not required to belong to the default namespace. This will be further elaborated in the discussion of the import PI below.
With the arrival of namespaces and polymorphism, it is no longer necessarily possible to indicate a single schema to which the entire instance conforms. Nevertheless, we require a mechanism to indicate a set of schemata containing definitions for all the element and datatypes appearing in an instance. This function is handled by the import processing instruction. The import PI contains one argument, an absolute URI indicating a schema.
An example import PI would look like:
<?import urn:x-commerceone:document:com:commerceone:schema1.sox$1.0?>
where the schema definition is located through resolving the
URI:
urn:x-commerceone:document:com:commerceone:schema1.sox$1.0.
The part following the "$" gives the version number.
A small sample document would look like:
<?soxtype urn:x-commerceone:document:com:commerceone:schema1.sox$1.0?> <?import urn:x-commerceone:document:com:commerceone:schema2.sox$1.0?> <Root <s2:Body xmlns:s2="urn:x-commerceone:document:com:commerceone:schema2.sox$1.0"/> </Root>
In the following, the namespace attributes on Root overrides the default namespace declared in the soxtype PI, and Body is assumed to also be in schema2.sox.
<?soxtype urn:x-commerceone:document:com:commerceone:schema1.sox$1.0?> <?import urn:x-commerceone:document:com:commerceone:schema2.sox$1.0?> <Root xmlns="urn:x-commerceone:document:com:commerceone:schema2.sox$.0"> <Body/> </Root>
It is not necessary that for every element type appearing in the instance there be a corresponding import or soxtype processing instruction. Nevertheless, all schema information must be available before processing of the root element begins. We can define a transitive closure property over schemata to accomplish this.
We require each schema declared in a soxtype or import PI to be processed. Furthermore, if any definition in a schema being processed refers to a definition ( elementtype or datatype ) in another schema, that other schema must be processed. The set of imports must be sufficient so that starting with the soxtype and the imports , and processing all schemata transitively referenced by them, all elementtypes and datatypes found in the instance will have been processed.
The definition of an XML schema is performed with the schema element. The corresponding DTD fragment is:
<!ELEMENT schema (intro?,
(datatype | elementtype |join
|comment | namespace)*)>
<!ATTLIST schema
prefix NMTOKEN #IMPLIED
uri CDATA #REQUIRED
soxlang-version NMTOKEN #FIXED "V2.0">
The following is a minimal valid instance of a schema:
<schema uri="urn:x-commerceone:sampleSchema"/>
However it is unlikely there will be many schemata containing no definitions at all.
A schema consists of any number of definitions of datatypes or element types (both of which will be explained subsequently). In other words, a schema is a set of definitions, not the set of files, database entries, etc., we use to store a representation of these definitions: the physical storage mechanism we use may change frequently, without the set of definitions being affected at all. A schema may start out as a single file, then be split among several files (which would currently be linked using the join mechanism described below), before being stored in a database. But despite being stored in three different ways, the set of definitions remains the same, and it is that set which comprises the schema.
Each schema has a unique name to identify it. This name is a URI, given in the uri attribute of the schema element in a fragment. In essence, it proclaims the set of definitions in this schema element to be a subset of the definitions forming the schema identified by the uri attribute.
The join construct allows definitions from externally defined fragments belonging to the same schema to be pulled in.
There are three main classes of symbols created during the construction of a Schema: elementtype names, datatype names, and namespace prefixes. In the current version of the language, both elementtype names and datatype names are maintained in a single set; it is illegal to use the same name to represent both an element type and a datatype. Prefixes for namespaces, however, are kept separately. It is entirely legal, although potentially confusing, to use the same name for a namespace prefix as for an elementtype or datatype .
NOTE: It is our intention to separate the datatype and elementtype namespaces in a future version of SOX. When we do so, this will be entirely backwards compatible. The expressive power of SOX, however, is unchanged whether the namespaces are separated or coallesced.
The prefix attribute specifies a prefix available for referencing names created in this schema. As the current schema is the default namespace for dereferencing names, this is not strictly necessary, however it can be useful when definitions from several namespaces are mixed in close proximity.
The uri attribute provides a URI establishing the namespace of this schema fragment. This attribute has become required to prevent accidental name capture through the join mechanism described below. This must be an absolute URI.
The namespace element is considered a declaration, not a definition. The namespace being declared is defined elsewhere (probably in a Schema file)
N.B. SOX does not deal with a number of linking issues related to the organization of schemata into multiple files.
The intro element is available to provide an introduction to the schema as a whole. It consists of a number of HTML elements. The exact allowable contents of intro is available in the htmltext.ent file reproduced at the end of this document.
The following are the constraints placed on Schema files, aside from strict structural conformance to the DTD.
A Schema is about types. Types are both defined and referenced. Referencing is done using names. Once a type (or namespace) is given a name (or prefix) in a fragment, that name that name is bound in that fragment. Names that are only referenced are free in that fragment.
A Schema fragment is processed in an environment. That environment extends beyond the fragment to include other schemata and the SOX definition itself. This document defines how the environment of a fragment is defined, but does not discuss how it is physically constructed.
A Schema fragment cannot be successfully processed unless all the names that are free in it are bound somewhere in its environment. It is an error for a name to be bound twice in a fragment's environment.
A fragment specifies its environment by providing bindings for all the free names it contains. There are four pieces to its environment that a fragment must specify:
Names are either qualified or unqualified. Unqualified names must be defined either in the current fragment, in the rest of the schema, or in the SOX definition. It can only be defined in one of the three. Qualified names must be defined in the schema declared for that name with a namespace declaration. For each qualified name in a fragment there must be a corresponding namespace declaration in the same fragment. The schema element itself is considered a namespace declaration for the current namespace, so qualified names using the value in the prefix attribute of the schema element must resolve to the current schema (although not necessarily in the same fragment). In this version of the language it is an error to create a definition whose local name is either one of the intrinsic datatypes.
<!ELEMENT namespace (explain?)>
<!ATTLIST namespace
prefix NMTOKEN #REQUIRED
namespace CDATA #REQUIRED >
All of the names defined in a single Schema belong to the same namespace, and can be used without qualifier. Schemata, however, frequently need to refer to definitions in other namespaces. A namespace declaration allows access to definitions in the referenced schema when appearing with an appropriate prefix attribute. A namespace declaration is scoped to the current schema fragment only. Namespace declarations made in one schema fragment are not visible in other fragments belonging to the same schema, even when referenced through a join .
NOTE: We will use a prefix attribute for the prefix, instead of using colonized names, in accordance with the XSDL spec [XSDL]. The prefix attribute shows up on various elements, including attdef, scalar , and extends, all of which include a reference to a definition.
XML requires the use of qualified names to make such references in document instances. A qualified name in an instance consists of two parts separated by a colon:
SOX implements qualified names through the use of two attributes, one for the name, and one for the prefix. Qualified names can be used wherever a reference to a definition is allowed - a schema cannot define a name in another schema, but it can extend an elementtype from another schema, or use a datatype from another Schema.
For example, the following fragment declares the urn:foo namespace and associates it with prefix bar:
<namespace prefix="bar" namespace="urn:foo"/>
If we later need to include a foobar element from the urn:foo namespace in an element type that would be done using the following fragment:
<elementtype name="et">
<model>
<element prefix="bar" type="foobar" name="whatever"/>
<element prefix="bar" type="foobar"/>
</model>
</elementtype>
A valid instance of this would be:
<et><whatever><bar:foobar xmlns:bar="urn:foo"/></whatever>
<bar:foobar xmlns:bar="urn:foo"/></et>
Each prefix must be unique within a schema fragment. While it is not a fatal error to declare a non-existent namespace, it is a fatal error to reference an element in a non-existent namespace, or to reference a non-existent element in an existing namespace. Both of these are semantic errors. It is a fatal runtime error if the processor is unable to retrieve a definition during processing. A processor should distinguish among these cases. This specification does not address the issue of how to retrieve definitions.
The value of the namespace attribute of the namespace declaration must be the URI of a schema, as described above. In other words, it must be the same as the value appearing in the uri attribute of the schema element of files which define that schema.
The explain element exists inside several different SOX constructs. It provides a hook for including documentation within a schema and exploits commonly known HTML [HTML-4] constructs.
<!ELEMENT explain (title?, synopsis?, (%html.block;)+) >
The title and html.block elements are common HTML constructs whose exact definitions are specified in the htmltext.ent file included at the end of this document. The synopsis is used to give a purpose or synopsis to the thing being explain ed. It is a single paragraph of text.
Because it is HTML embedded in XML, all the HTML constructs must be used in a well-formed manner. It is common to take advantage of SGML tag minimization in writing HTML documents, but that would result in well-formedness errors in a SOX schema.
In XML Schema documents, element type definitions reproduce the expressiveness of XML element type declarations using explicit element and attribute markup. An element type may be defined by using the elementtype element with the required name attribute, and a subordinate model or extends element (both of which will be described subsequently).
The corresponding DTD fragment is:
<!ELEMENT elementtype (explain?, (extends | ((empty|model), (attdef)*)))>
<!ATTLIST elementtype
name NMTOKEN #REQUIRED >
The following example defines an element type of name inline :
<elementtype name="inline">
<explain>
<synopsis>This defines the <em>inline</em> element</synopsis>
</explain>
<model>
<string/>
</model>
</elementtype>
A mechanism for attaching attributes to an element type is described later.
A valid instance for this fragment would be:
<inline>This is a string</inline>
The name of an element type may be any valid unqualified XML element type name corresponding to the Name production in the XML 1.0 language definition. The name must be unique among the names of element types and datatypes defined in the current Schema, which includes the current document or other documents belonging to the same Schema processed through the join mechanism or other resolution mechanism.
An element type may be referenced by the element and extends elements.
It is a fatal error to re-assign an element name, or to reference an element type which is not defined.
The value of the name attribute must be unique across all elementtype names defined in this schema.
The content model of an element type defines the structure and composition of an element of that type in an XML instance. The definition of a content model in XML Schema documents extends the expressiveness of XML DTDs by providing greater specificity of the minimum and maximum number of times some content model atom may be repeated. This allows an XML Schema designer with more precise control than is offered by XML's *, ? and + occurrence indicators.
The DTD fragment corresponding to an content model definition is:
<!ELEMENT model (string|element|choice|sequence)>
An empty atom is used to indicate that an element may not contain any content, as in the case of the BR element below:
<elementtype name="BR"> <empty/> </elementtype>
In order to properly support extensibility (explained below) an empty content model is considered to be an empty sequence .
Valid instances are:
<BR/>
or:
<BR></BR>
The string atom indicates that a content model is simply string content and is an evolution of #PCDATA . It can be used as in the example above. In addition, the string value may be constrained to be of a particular datatype defined by the optional datatype attribute (and prefix , if the datatype is in another schema) as can be seen in the DTD fragment:
<!ELEMENT string EMPTY>
<!ATTLIST string
prefix NMTOKEN #IMPLIED
datatype NMTOKEN "string" >
Any element type with string in its content model is considered to be a choice group with an occurs value of "*". This is consistent with the XML 1.0 spec which requires that any content model containing #PCDATA be in a choice with a Kleene star. It is unclear if this restriction will be maintained in the Schema world.
In the example below, the size element type's content model is string content constrained to be an int :
<elementtype name="size">
<model>
<string datatype="int" />
</model>
</elementtype>
A valid example of this would be:
<size>12345</size>
However the following would not be valid:
<size>12r34</size>
A content model may also comprise zero or more repetitions of another element. The DTD fragment for this definition is:
<!ELEMENT element EMPTY>
<!ATTLIST element
prefix NMTOKEN #IMPLIED
type NMTOKEN #REQUIRED
name NMTOKEN #IMPLIED
occurs CDATA #IMPLIED >
The defined element is an instance of either a previously defined datatype or element type, which is refered to by the required type attribute. As before, it is a fatal error to reference a datatype or element type that is not defined.
For purposes of extensibility, a content model with just one element is considered a sequence of length one.
The name attribute may be used to assign a name to the defined element when it appears in an instance. As datatypes are not also element types, the name attribute must have a value when type references a datatype. When name is specified, this shows up as an additional element wrapping an element of the referenced type.
The following fragment demonstrates the use of name . The type int refers to the built-in integer datatype:
<elementtype name="paragraph">
<model>
<string/>
</model>
</elementtype>
<elementtype name="block">
<model>
<sequence>
<element name="p" type="paragraph"/>
<element name="position" type="int"/>
</sequence>
</model>
</elementtype>
A valid instance would look like this:
<block><p><paragraph>this is the paragraph</paragraph></p><position>12345</position></block>
The following would not be valid:
<block><paragraph>you must use the name</paragraph><int>1</int></block>
The occurs attribute indicates the number of repetitions of the instanced element . It can take on the values of:
We will call an occurs where N1 is not equal to N2 an indefinite occurs. The degenerate case of "0,0" is allowed and means exactly 0 repetitions, which is treated the same as if the declaration did not occur.
NOTE: Values of the form " N1,N2 " and " N1,* " are not currently supported. They will be treated as an occurs of " + " if N1 is greater than 0, or as a " * " otherwise.
In the following example, the definition of the content model for a list element type specifies that it contains a minimum of 2 and a maximum of 9 item elements.
<elementtype name="list">
<model>
<element type="item" occurs="2,9"/>
</model>
</elementtype>
A valid instance of the above would be:
<list><item/><item/></list>
The choice atom defines a content model to comprise one of a set of choices of element , choice or sequence content models. The relevant DTD fragment is:
<!ELEMENT choice ((element|choice|sequence),
(element|choice|sequence)+) >
<!ATTLIST choice
name NMTOKEN #IMPLIED
occurs CDATA #IMPLIED >
As with element , the occurs attribute specifies the number of repetitions, and it can take the same values as defined earlier.
In the following example, the dl element type's content model specifies that either a single dt or a single dd element is allowed.
<elementtype name="dl">
<model>
<choice>
<element type="dt"/>
<element type="dd"/>
</choice>
</model>
</elementtype>
A valid instance would be:
<dl><dt/></dl>
or
<dl><dd/></dl>
but not:
<dl><dd/><dt/></dl>
The sequence atom defines a content model to consist of the specified element , choice or sequence content models appended together in the order specified. The relevant DTD fragment is:
<!ELEMENT sequence ((element|choice|sequence),
(element|choice|sequence)+) >
<!ATTLIST sequence
name NMTOKEN #IMPLIED
occurs CDATA #IMPLIED >
As before the occurs attribute specifies the number of repetitions of the entire sequence , and it can take the same values as defined earlier.
In the following example, the dl element type's content model specifies that a single dt followed by a single dd is allowed.
<elementtype name="dl">
<model>
<sequence>
<element type="dt"/>
<element type="dd"/>
</sequence>
</model>
</elementtype>
A valid instance would be:
<dl><dt/><dd/></dl>
The various content model atoms defined above may be combined to allow the definition of complex content models.
For example, the dl element type's content model below specifies that a dh is followed by two or more dt or dd elements.
<elementtype name="dl">
<model>
<sequence>
<element type="dh"/>
<choice occurs="2,*">
<element type="dt"/>
<element type="dd"/>
</choice>
</sequence>
</model>
</elementtype>
A valid instance would be:
<dl><dh/><dt/><dd/><dd/><dt/></dl>
The occurs attribute may not occur on the outermost sequence or choice in an elementtype definition. This is the sequence or choice immediately contained within model . The outermost sequence or choice within the model must occur exactly once.
The value of the name attribute (if any) given to an element, choice, or sequence, must be unique within the innermost enclosing construct. For example, the following is legal:
<sequence> <element name="a" type="string"/> <element name="b" type="int"/> </sequence>
while the following is not:
<sequence> <element name="a" type="string"/> <element name="a" type="int"/> </sequence>
Likewise, the following is valid:
<sequence>
<element name="a" type="string"/>
<sequence name="c">
<element name="a" type="string"/>
<element name="c" type="int"/>
</sequence>
<element name="b" type="string"/>
</sequence>
As in object-oriented inheritance, an element may specialize (or subclass) from another element by inheriting its structure and then adding on to its content model. Inheritance is specified using the extends construct, and the relevant DTD fragment is:
<!ELEMENT extends (append?, attdef*)>
<!ATTLIST extends
prefix NMTOKEN #IMPLIED
type NMTOKEN #REQUIRED >
<!ELEMENT append (element|choice|sequence)+>
The type attribute refers to the base element type that is being extended, and the structure of the append atom has the same contents as that of model . The base type must be already defined. The contents of the append element are added to the end of the parent's content model (the outermost sequence ). Note that the append element has been made optional. This makes it possible to declare semantically distinct element types whose structures remain the same as that of some common parent.
In the following example, the element type datednote has the content model of the element type it extends ( note ) with an appended date (using the intrinsic date datatype). The multinote element type polymorphyically can use either.
<elementtype name="note">
<model><element type="p" occurs="+"></model>
</elementtype>
<elementtype name="datednote">
<extends type="note">
<append>
<element type="date" name="adate">
<element type="time" name="atime" occurs="?"/>
</append>
</extends>
</elementtype>
<elementtype name="multinote">
<element type="note" occurs="+"/>
</elementtype>
The following is a valid instance of multinote:
<multinote>
<note><p>This is a plain note</p></note>
<datednote>
<p>This is a dated note</p>
<adate>19981209</adate>
<atime>10:23:32</atime>
</datednote>
</multinote>
SOX permits an elementtype in one namespace to extend an elementtype in another Schema. In order to work correctly with the current draft of the W3C Namespace Recommendation, each element in an instance belongs to the namespace in which it was declared. In the above example, if note were declared in schema Foo.sox and both datednote and multinote in Bar.sox , then the following would be an appropriately prefixed version of the above example:
<bar:multinote xmlns:foo="Foo.sox" xmlns:bar="Bar.sox">
<foo:note><foo:p>This is a plain note</foo:p></foo:note>
<bar:datednote>
<foo:p>This is a dated note</foo:p>
<bar:adate>19981209</bar:adate>
<bar:atime>10:23:32</bar:atime>
</bar:datednote>
</bar:multinote>
A prefered solution would be to consider local names as being in the namespace of the elementtype they are declared in (or its subtypes) and therefore need not be prefixed. This would be similar to the treatment of attributes.
In order to support extensibility, each elementtype must be either a choice or a sequence. By default:
There are some constraints placed on the form of content models to avoid ambiguous models and interminable documents. These are an extension of similar restriction in XML 1.0.
For any possible path in the parse tree from an element to a descendant of itself there must be an intervening optional node (?, *, or indefinite occurs) or intevening choice node with at least two children. This assures that infinite documents are not required.
For any optional node in the parse tree (?, *, +, indefinite occurs, or choice), none of the descendants of elements in its first set may be in its follow set.
These conditions continue to hold when extending an elementtype.
Due to the desire to maintain substitutability of extended elementtypes with their base type, SOX 2.0 does not allow the extension of elementtypes whose content model is choice. In order to maintain substitutability, any element extending a choice would need to be a subtype of something already in the choice, so it would already be valid in the parent type.
When extending a sequence with a occurs of indefinite extent, the resulting content model must be checked for ambiguity with itself. In other words, if the parent model was x* , where x is some content model, then the content model xx ( x followed immediately by x ) must have been unambiguous. If the child model is now x+e (i.e., x plus some additional element), then the content model (x+e)(x+e) must still be unambiguous.
Within an element, the value of the name attribute is scoped to the surrounding elementtype. It neither creates nor prevents the creation of a top level definition using the same name. However, in order to maintain backwards compatibility with XML 1.0, the binding between name and type must be global. In other words, once a name has been used with a particular type, it cannot be used with another type. We expect this restriction to be relaxed in a future version.
Along the lines of being able to define element types, XML Schema provides the means to define and refer to datatypes. XML Schema defines a set of intrinsic datatypes, listed below, from which user-defined datatypes may be derived. This list is derived from existing ISO standards [ISO-31, ISO-8601, DATETIME] and common programming language practice. Some of these have been refered to elsewhere in this document. It also (currently) provides the datatype element for actually defining new datatypes. The appropriate DTD fragment is:
<!ELEMENT datatype (explain?, (enumeration|scalar|varchar)) >
<!ATTLIST datatype
name NMTOKEN #REQUIRED >
The value of the name attribute specifies the name of the new datatype. This must be unique across all the datatypes defined for this schema. The three operators, enumeration , scalar , and varchar (all further described below), each derive a new datatype from an existing datatype. The existing datatype can be either one of the intrisic types, or some other user-defined datatype, whether in this schema or another. Both scalar and varchar have some restrictions on which datatypes they can extend, as described below.
The intrinsic datatypes define the domains of the atomic data units in XML Schema documents. The list below includes those datatypes defined intrinsic to this version of the Schema language.
XML Schema documents provide a mechanism for defining enumerations to constrain attribute or element string content. An enumeration datatype is a finite set of values enumerated by the option elements inside the enumeration element.
<!ELEMENT enumeration (explain?, option)+ >
<!ATTLIST enumeration
prefix NMTOKEN #IMPLIED
datatype NMTOKEN #REQUIRED >
<!ELEMENT option (#PCDATA)* >
The datatype attribute of the enumeration specifies the instrinsic (see list above) or user-defined datatype (i.e., other enumeration , varchar , or scalar ) being refined. If the datatype is not defined in this Schema, then the prefix of the appropriate schema must be specified.
Each option has a value representing a valid value for the datatype being extended.
The following example demonstrates the definition and use of an enumeration datatype:
<datatype name="colortype">
<enumeration datatype="NMTOKEN">
<option>Red</option>
<option>Blue</option>
<option>Green</option>
</enumeration>
</datatype>
<elementtype name="car">
<empty/>
<attdef name="color"datatype="colortype"><required/></attdef>
</elementtype>
<elementtype name="bus">
<model>
<element name="color" type="colortype">
</model>
</elementtype>
The following are valid instances:
<car color="Red"/> <bus><color>Blue</color></bus>
Scalar datatypes are used for creating subtypes of number . The relevant DTD fragment is:
<!ELEMENT scalar EMPTY >
<!ATTLIST scalar
prefix NMTOKEN #IMPLIED
datatype NMTOKEN "number"
digits CDATA #IMPLIED
decimals CDATA #IMPLIED
minvalue CDATA #IMPLIED
maxvalue CDATA #IMPLIED
minexclusive (true | false) "false"
maxexclusive (true | false) "false" >
Digits specifies the maximum number of digits for the integral part of the number, decimals specifies the maximum number of digits for the fraction part of the number. The following constraints must hold:
An example scalar would be:
<scalar digits="4" decimals="3" minvalue="-9999" maxvalue="8888" minexclusive="true"/>
The following are valid: -9998.999, 8887.999, 0.0, 8888.
The following are invalid: -9999, 8888.001.
The number datatype covers all rational numbers which can be finitely specified as a decimal (i.e., it does not cover infinitely repeating decimals, such as 1/9, or irrational reals, such as p). The other intrinsic classes are designed to be easily mappable to existing datatypes in common programming languages.
Varchar , adapted from SQL[SQL], is for specifying string types with fixed maximum length.
<!ELEMENT varchar EMPTY>
<!ATTLIST varchar
prefix NMTOKEN #IMPLIED
datatype NMTOKEN "string"
maxlength CDATA #REQUIRED>
The value of maxlength must be a non-negative integer.
An example varchar use is:
<datatype name="var">
<varchar maxlength="4"/>
</datatype>
<elementtype name="wrap">
<model>
<string datatype ="var"/>
</model>
</elementtype>
A valid instance would be:
<wrap>abc</wrap>
as would:
<wrap>abcd</wrap>
An invalid instance would be:
<wrap>abcde</wrap>
Float , double , int , long , and byte are all predefined scalar types. This means they can be used as base types for any new definition of a scalar and can be referenced as datatypes. The value of maxlength in varchar must be greater than or equal to 0. Note that a length of 0 implies an empty string. The datatype of a varchar must be either string or another varchar or one of string, NMTOKEN, NMTOKENS, ID, IDREF, or IDREFS.
The names of the intrinsic types are reserved by SOX. It is an error to define a datatype or elementtype using the name of one of the intrinsic types.
Attribute definitions in XML Schema documents may be defined as part of the element type definition. An attribute definition has a name and a type, and must include a presence element. The relevant fragment of the DTD is:
<!ELEMENT attdef (explain?,(enumeration | scalar | varchar)?,
(required|implied|default|fixed)?)>
<!ATTLIST attdef
name NMTOKEN #REQUIRED
prefix NMTOKEN #IMPLIED
datatype NMTOKEN #IMPLIED>
The name of the attribute is defined by the value of name , and it may be of a certain datatype . It must be unique among the attributes for this elementtype. If a value is not given for the datatype attribute and the attdef does not contain an enumeration , scalar , or varchar , then a datatype of string is assumed.
An attribute value's presence in an instance may be specified as default , fixed , required or implied, as in the XML specification. If no value is given for the presence, then it defaults to implied . In all cases, the value of an attribute must be valid for its datatype. The DTD fragment for the presence elements is:
<!ELEMENT default (#PCDATA) > <!ELEMENT fixed (#PCDATA) > <!ELEMENT required EMPTY > <!ELEMENT implied EMPTY >
A default presence indicates that the value of the attribute is automatically set to the default value if none is specified. In an instance, if the attribute is defined to have another value, the default is ignored.
A fixed presence indicates that the value of the attribute is assumed to be the fixed value if no value is specified. The attribute may also be explicitly assigned exactly this fixed value. In an instance, if the attribute is defined to have a different value, this signals a fatal error.
An empty value can be specified for default or fixed.
A required presence indicates that in an instance, whenever the parent element appears, this attribute must be assigned a value.
And finally, an implied presence indicates that the application is given the responsibility of filling in a default value if no attribute value is defined in an instance.
In XML Schema documents, unlike XML DTDs, enumerations may be specified for any attribute type. This information will be lost if an XML DTD is generated from an XML Schema document, except for attributes of type NMTOKEN , indicating a name token. If there is an enumeration specified for an attribute and also a fixed or default value, then that fixed or default value must be a member of the enumeration .
Thus, example attribute definitions (within the definition of an elementtype , of course) might be:
<elementtype name="car">
<empty/>
<attdef name="owner" datatype="string"/>
<attdef name="color">
<enumeration datatype="NMTOKEN">
<option>Red</option>
<option>Blue</option>
<option>Green</option>
</enumeration>
<required/>
</attdef>
</elementtype>
An instance corresponding to this definition would look like:
<car color="Blue" owner="John Smith"/>
It is a fatal error for the datatype attribute of the attdef element to have a value if there is an enclosed enumeration , scalar , or varchar.
An externally-defined schema file whose definitions belong to the same namespace may be pulled in and parsed with the current schema definition. This is accomplished using the join construct. The relevant fragment of the DTD is:
<!ELEMENT join (explain?)>
<!ATTLIST join
datatype NMTOKEN #FIXED "schema"
public CDATA #IMPLIED
system CDATA #REQUIRED>
The datatype attribute is fixed as only schemas may be included for now. The public attribute is the public identifier of the file as defined in the XML 1.0 specification. The system attribute is the URI of the file containing the schema definition to be lexically included. The entity manager must resolve this URI.
A join ed file is read only once by the parser. The parser determines identity between files by comparison of the values of the system attribute. It is not illegal for two join elements to reference the same URI, but one will be ignored. It is a user error to use two different URIs which ultimately map to the same file. Note that reading a fragment twice will cause an error as the schema processor will create all its definitions twice.
The joined file must belong to the same namespace as the joining one, as identified by the URI attribute of the root elements.
In the current implementation, both namespace and join elements use URIs to point to external files, and processing a namespace involves retrieving the physical file referenced by the namespace element. It is not an error for a join element to reference this file again, however it is a runtime error for that file to actually be retrieved a second time. In other words, no fragment for a schema is to be processed more than once.
<!-- ************************************************************* -->
<!-- XML Schema DTD -->
<!-- PUBLIC "-//Commerce One Inc.//DTD XML Schema 2.0//EN" -->
<!-- SYSTEM "schema.dtd" -->
<!-- Copyright: Commerce One Inc., 1997, 1998, 1999
Date created: 17 Dec 1997
Date revised: 03 June 1999
Version: 2.0
-->
<!-- ************************************************************* -->
<!-- ************************************************************* -->
<!-- XML Schema ************************************************** -->
<!-- ************************************************************* -->
<!ENTITY % htmltext SYSTEM "htmltext.ent">
%htmltext;
<!ELEMENT schema (intro?, (datatype | elementtype | join | comment | namespace)*) >
<!ATTLIST schema
prefix NMTOKEN #IMPLIED
uri CDATA #REQUIRED
soxlang-version NMTOKEN #FIXED "V0.2.2">
<!-- ************************************************************* -->
<!-- ELEMENTS **************************************************** -->
<!-- ************************************************************* -->
<!-- An Element Type definition requires a name.
It is defined to extend a named element,
as an instance of a named element,
as an EMPTY or ANY element with optional attribute definitions,
or with a content model with optional attribute definitions. -->
<!ELEMENT elementtype (explain?, (extends | ((empty|model), (attdef)*)))>
<!ATTLIST elementtype
name NMTOKEN #REQUIRED >
<!ELEMENT empty EMPTY >
<!-- ************************************************************* -->
<!-- MODEL ******************************************************* -->
<!-- ************************************************************* -->
<!ELEMENT model (string|element|choice|sequence)>
<!ELEMENT extends (append?, attdef*)>
<!ATTLIST extends
prefix NMTOKEN #IMPLIED
type NMTOKEN #REQUIRED >
<!ELEMENT append (element|choice|sequence)+>
<!ELEMENT element EMPTY >
<!ATTLIST element
prefix NMTOKEN #IMPLIED
type NMTOKEN #REQUIRED
name NMTOKEN #IMPLIED
occurs CDATA #IMPLIED >
<!ELEMENT string EMPTY >
<!ATTLIST string
prefix NMTOKEN #IMPLIED
datatype NMTOKEN "string" >
<!ELEMENT choice ((element|choice|sequence),
(element|choice|sequence)+) >
<!ATTLIST choice
name NMTOKEN #IMPLIED
occurs CDATA #IMPLIED >
<!ELEMENT sequence ((element|choice|sequence),
(element|choice|sequence)+) >
<!ATTLIST sequence
name NMTOKEN #IMPLIED
occurs CDATA #IMPLIED >
<!-- replacement for "include" -->
<!ELEMENT join (explain?)>
<!ATTLIST join
datatype NMTOKEN #FIXED "schema"
public CDATA #IMPLIED
system CDATA #REQUIRED>
<!-- ************************************************************* -->
<!-- ATTRIBUTES ************************************************** -->
<!-- ************************************************************* -->
<!-- An attribute definition has a name and datatype, and must have a
presence element "required|implied|default|fixed" included.
It may have a namespace associated with it, or inherit
enumeration is supposed to define the domain of acceptable value -->
<!ELEMENT attdef (explain?, (enumeration | scalar | varchar)?,
(required|implied|default|fixed)?)>
<!ATTLIST attdef
name NMTOKEN #REQUIRED
prefix NMTOKEN #IMPLIED
datatype NMTOKEN #IMPLIED >
<!ELEMENT default (#PCDATA) >
<!ELEMENT fixed (#PCDATA) >
<!ELEMENT required EMPTY >
<!ELEMENT implied EMPTY >
<!-- ************************************************************* -->
<!-- DATATYPE **************************************************** -->
<!-- ************************************************************* -->
<!ELEMENT datatype (explain?, (enumeration|scalar|varchar)) >
<!ATTLIST datatype
name NMTOKEN #REQUIRED >
<!ELEMENT enumeration (explain?, option)+ >
<!ATTLIST enumeration
prefix NMTOKEN #IMPLIED
datatype NMTOKEN #REQUIRED >
<!ELEMENT option (#PCDATA)* >
<!ELEMENT scalar EMPTY >
<!ATTLIST scalar
prefix NMTOKEN #IMPLIED
datatype NMTOKEN "number"
digits CDATA #IMPLIED
decimals CDATA #IMPLIED
minvalue CDATA #IMPLIED
maxvalue CDATA #IMPLIED
minexclusive (true | false) "false"
maxexclusive (true | false) "false" >
<!ELEMENT varchar EMPTY>
<!ATTLIST varchar
prefix NMTOKEN #IMPLIED
datatype NMTOKEN "string"
maxlength CDATA #REQUIRED>
<!-- ************************************************************* -->
<!-- COMMENT ***************************************************** -->
<!-- ************************************************************* -->
<!ELEMENT comment (#PCDATA)>
<!-- Namespaces -->
<!ELEMENT namespace (explain?) >
<!ATTLIST namespace
prefix NMTOKEN #REQUIRED
namespace CDATA #REQUIRED >
<!-- ************************************************************* -->
<!-- HTML Text: SOX uses HTML element types for convenience. -->
<!-- ************************************************************* -->
<!-- Copyright: Commerce One Systems Inc., 1997, 1998
Date created: 17 Dec 1997
Date revised: 01 Mar 1999
Version: 1.0 -->
<!-- ************************************************************* -->
<!ENTITY % html.nonheading " table | p | bq | pre | ol | ul | dl" >
<!ENTITY % html.text "#PCDATA| a | abbr | b | big | br
| cite | code | em | i | img
| q | small | span | strike | strong
| sub | sup | tt | u " >
<!ENTITY % html.heading.text
"#PCDATA| a | abbr | b | big | br
| cite | code | em | i | img
| q | small | span | strike | strong
| sub | sup | tt | u " >
<!ENTITY % html.block "h1|h2|h3|h4|h5|DT|%html.nonheading;" >
<!-- ************************************************************* -->
<!-- The intro element type is used to introduce a schema. It contains a
general description of the purpose and use of the schema's document
type or components. -->
<!ELEMENT intro ((%html.block;)+) >
<!-- The explain element is use to document a component within a schema. -->
<!ELEMENT explain (title?, synopsis?, (%html.block;)+) >
<!-- The title is used to give a human readable title to some type name. -->
<!ELEMENT title (%html.text;)* >
<!-- The synopsis is used to give a purpose or synopsis to the thing being
explained. It is a single paragraph of text. -->
<!ELEMENT synopsis (%html.text;)* >
<!-- ************************************************************* -->
<!ELEMENT h1 (%html.heading.text;)* >
<!ELEMENT h2 (%html.heading.text;)* >
<!ELEMENT h3 (%html.heading.text;)* >
<!ELEMENT h4 (%html.heading.text;)* >
<!ELEMENT h5 (%html.heading.text;)* >
<!ELEMENT DT (%html.heading.text;)* >
<!-- ************************************************************* -->
<!ELEMENT b (#PCDATA)* >
<!ELEMENT br EMPTY >
<!ELEMENT big (#PCDATA)* >
<!ELEMENT i (#PCDATA)* >
<!ELEMENT small (#PCDATA)* >
<!ELEMENT sub (#PCDATA)* >
<!ELEMENT sup (#PCDATA)* >
<!ELEMENT strike (#PCDATA)* >
<!ELEMENT tt (#PCDATA)* >
<!ELEMENT u (#PCDATA)* >
<!ELEMENT abbr (#PCDATA)* >
<!ELEMENT cite (#PCDATA)* >
<!ELEMENT code (#PCDATA)* >
<!ELEMENT em (#PCDATA)* >
<!ELEMENT q (#PCDATA)* >
<!ELEMENT span (#PCDATA)* >
<!ELEMENT strong (#PCDATA)* >
<!-- ************************************************************* -->
<!ELEMENT a (%html.text;)* >
<!ATTLIST a
name CDATA #IMPLIED
href CDATA #IMPLIED
title CDATA #IMPLIED >
<!-- ************************************************************* -->
<!ELEMENT img (explain?) >
<!ATTLIST img
src CDATA #REQUIRED
alt CDATA #REQUIRED
longdesc CDATA #IMPLIED
usemap CDATA #IMPLIED >
<!-- ************************************************************* -->
<!ELEMENT pre (%html.text;)* >
<!ATTLIST pre
xml:space (preserve) #REQUIRED >
<!-- ************************************************************* -->
<!ELEMENT p (%html.text;)* >
<!ELEMENT bq (%html.text;)* >
<!ELEMENT ol (lh?, li+) >
<!ELEMENT ul (lh?, li+) >
<!ELEMENT lh (%html.heading.text;)* >
<!ELEMENT li (%html.text;|%html.block;)* >
<!ELEMENT dl (dh?,(dt,dd)+) >
<!ELEMENT dh (%html.heading.text;)* >
<!ELEMENT dt (%html.text;|%html.block;)* >
<!ELEMENT dd (%html.text;|%html.block;)* >
<!-- ************************************************************* -->
<!ELEMENT table
(thead?, tbody) >
<!ATTLIST table
cols CDATA #IMPLIED
width CDATA #IMPLIED
height CDATA #IMPLIED
align (left|center|right|justify) #IMPLIED
valign (top | middle | bottom | baseline) #IMPLIED
vspace CDATA #IMPLIED
hspace CDATA #IMPLIED
cellpadding CDATA #IMPLIED
cellspacing CDATA #IMPLIED
border CDATA #IMPLIED
frame (box|void|above|below|hsides|vsides|lhs|rhs) #IMPLIED
rules (none|groups|rows|cols|all) #IMPLIED >
<!ELEMENT thead
(tr)+ >
<!ATTLIST thead
align (left|center|right|justify) #IMPLIED
valign (top|middle|bottom|baseline) #IMPLIED >
<!ELEMENT tbody
(tr)+ >
<!ATTLIST tbody
align (left|center|right|justify) #IMPLIED
valign (top|middle|bottom|baseline) #IMPLIED >
<!ELEMENT tr (th | td)+ >
<!ATTLIST tr
align (left|center|right|justify) #IMPLIED
valign (top | middle | bottom | baseline) #IMPLIED >
<!ELEMENT th (%html.text;|%html.block;)* >
<!ATTLIST th
colspan CDATA #IMPLIED
rowspan CDATA #IMPLIED
width CDATA #IMPLIED
height CDATA #IMPLIED
align (left|center|right|justify) #IMPLIED
valign (top | middle | bottom | baseline) #IMPLIED >
<!ELEMENT td (%html.text;|%html.block;)* >
<!ATTLIST td
colspan CDATA #IMPLIED
rowspan CDATA #IMPLIED
width CDATA #IMPLIED
height CDATA #IMPLIED
align (left|center|right|justify) #IMPLIED
valign (top | middle | bottom | baseline) #IMPLIED >
<!-- ************************************************************* -->