W3C

[Editorial Draft] Extending and Versioning Languages: XML Languages

Draft TAG Finding 04 July 2007

This version:
http://www.w3.org/2001/tag/doc/versioning-xml-20070704.html ( xml )
Latest version:
http://www.w3.org/2001/tag/doc/versioning-xml
Previous versions:
Unapproved Editors Drafts: http://www.w3.org/2001/tag/doc/versioning-xml-20070518.html, http://www.w3.org/2001/tag/doc/versioning-xml-20070326.html, http://www.w3.org/2001/tag/doc/versioning-xml-20070202.html, http://www.w3.org/2001/tag/doc/versioning-xml-20061212.html, http://www.w3.org/2001/tag/doc/versioning-part2-20041124.html
Editor:
David Orchard, BEA Systems, Inc. <David.Orchard@BEA.com>

Abstract

This document discusses the XML related aspects of versioning. It describes XML based terminology, technologies and versioning strategies. It provides XML Schema examples for each of the strategies and discussion about various schema design patterns. A number of XML languages, including XHTML and Atom, are used as case studies in different strategies.

Status of this Document

This document has been developed for discussion by the W3C Technical Architecture Group. It does not yet represent the consensus opinion of the TAG.

Publication of this finding does not imply endorsement by the W3C Membership. This is a draft document and may be updated, replaced or obsoleted by other documents at any time.

Additional TAG findings, both approved and in draft state, may also be available. The TAG expects to incorporate this and other findings into a Web Architecture Document that will be published according to the process of the W3C Recommendation Track.

Please send comments on this finding to the publicly archived TAG mailing list www-tag@w3.org (archive).

Table of Contents

1 Introduction
    1.1 XML Terminology
    1.2 Kinds of XML Languages
2 XML Language Requirements
3 Version Identification technologies
    3.1 Qualified Name: Namespace + Local name
    3.2 Type
    3.3 Version Numbers
4 Component version identification strategies
    4.1 Versioning Strategy #1: all components in new namespace(s) for each version
        4.1.1 Advantages and Disadvantages
    4.2 Versioning Strategy #2: all new components in new namespace(s) for each compatible version
        4.2.1 Advantages and Disadvantages
    4.3 Versioning Strategy #2.5: all new components in new or existing namespace(s) for each compatible version
        4.3.1 Advantages and Disadvantages
    4.4 Versioning Strategy #3: all new components in existing namespace(s) for each compatible version
        4.4.1 Advantages and Disadvantages
    4.5 Versioning Strategy #4: all new components in existing or new namespace(s) for each version and a version identifier
        4.5.1 Advantages and Disadvantages
    4.6 Versioning Strategy #5: all components in existing namespace(s) for each version and a version identifier
        4.6.1 Advantages and Disadvantages
5 Indicating compatibility of changes or extensions
    5.1 Compatible
    5.2 Incompatible
        5.2.1 Must Understand flag
6 XML Schema 1.0
7 Schemas for Version Identification Strategies
    7.1 #1: all components in new namespace(s) for each version
    7.2 #2: all new components in new namespace(s) for each compatible version
        7.2.1 Redefine
    7.3 #2.5: All new components in existing or new namespace(s) for each compatible version
    7.4 #3: All new components in existing namespace(s) for each compatible version
        7.4.1 Redefine
    7.5 #4: all new components in existing or new namespace(s) for each version and a version identifier
    7.6 #5: all components in existing namespace(s) for each version and a version identifier
8 Indicating Incompatible changes
    8.1 Type extension
    8.2 Substitution Groups
    8.3 Must Understand
9 Survey of Languages Versioning Strategies
10 Unique Particle Attribution
11 Other technologies
12 Conclusion
13 References
14 Acknowledgements

Appendix

A Change Log (Non-Normative)


1 Introduction

Extending and Versioning XML Languages Part 1 described extending and versioning languages. Part 2 focuses on XML and includes schema language specific aspects of extending and versioning XML. The choices, decisions, and strategies described in Part 1 are augmented with XML and XML Schema instances herein.

1.1 XML Terminology

We will describe some key refinements to our versioning terminology for XML. An XML language has a vocabulary that may use terms from one or more XML Namespaces (or none), each of which has a namespace name. [Definition: An XML language is an where all the Texts MUST be well-formed XML.]. As XML is a markup language, the most significant parts of XML Languages are the elements and attributes. [Definition: A component is an XML element or attribute.] The Name Language - consisting of name, given, family terms - has a namespace for the terms. We use the prefix "namens" to refer to that namespace. The Name Language could consist of terms from other vocabularies, such as Dublin Core or UBL. These terms each have their own namespaces, illustrating that a language can be comprised of terms from multiple namespaces. An XML Namespace is a convenient container for collecting terms that are intended to be used together within a language or across languages. It provides a mechanism for creating globally unique names.

We use the term instance when speaking of sequences of characters (aka text) in XML. [Definition: An instance is a Text of well-formed XML and the texts are usually constrained by a schema language.] The schema language may be machine processable such as DTDs, XML Schema, Relax NG, or the schema language may be human readable text. Documents are instances of a language. In XML, they must have a root element. A name text might have a name element as the root element. Alternatively, the name vocabulary may be used by another language such as purchase orders so the purchase order texts may contain texts that can be considered name texts. Thus instances of an XML language are always at least part of a text and may be the entire text. XML instances (and all other instances of markup languages) consist of markup and content. In the name example, the given and family elements including the end markers are the markup. The values between the start and end markers are the content. An instance has an information model. There are a variety of data models within and without the W3C, and the one standardized by the W3C is the XML infoset.

The XML related terms and their relationships are shown below

UML diagram of XML terms

ednote: Update to final terminology version..

Some examples of XML consumers and producers are: A stylesheet processor is a consumer of the XML text that it is processing (the producer isn't mentioned); in the Web services context the roles of producer and consumer alternate as messages are passed back and forth. Note that most Web service specifications provide definitions of inputs and outputs. By our definitions, a Web service that updates its output schema is considered a new producer. A Web service that updates its input schema is a new consumer.

1.2 Kinds of XML Languages

Ultimately, there are different kinds of XML languages. The versioning approaches and strategies that are appropriate for one kind of language may not be appropriate for another. Among the various kinds of vocabularies, we find:

This is by no means an exhaustive list. Nor are these categories completely clear cut. MathML can certainly be used standalone, for example, and languages like SVG are a combination of standalone, containers, and mixtures.

2 XML Language Requirements

The general language questions described in Part 1 Requirements (../versioning#requirements). These requirements are augmented in XML by:

3 Version Identification technologies

Version identification of elements and attributes is often an important part of correctly processing xml documents. There are a large variety of version identification technologies in XML. The fundamental technologies available for identification of versions in an xml document are:

  1. Qualified Names: Namespaces + Local Name

  2. Types

  3. Version Numbers

The decisions about which technologies to use are affected by the general language requirements and by the XML environments.

3.2 Type

Many systems use type information associated with the component as part of the version identification of the component. There are generally two strategies for determining the type of a component, which we will call "Top-typing" and "Bottom-typing". Top-typing is a style where the type of the component is determined by the type of the top element. Bottom-typing is a style where the type of the component is determined by the type or types of the descendents of the top element. Bottom-typing designs can be such that a single type is not possible, rather it is effectively the composite of all the descendent types. When top-types are extended, the type becomes more difficult to specify. It could be the top-type, or it could be the top-type plus all of the extension types. A nameType extended with a middleNameType could be considered a nameType or a nameType+middleNameType. As the number of extension types grows, specifying the actual type may become equivalent to some bottom-typing designs. The extreme of this is container languages, which have a single invariable top-type and typically intended to have the type determined by the content at the "bottom" of the container.

In either typing case, the type(s) used for type determination is determined by the type associated with the qualified name of the component during type assignment. In XML Schema, this is during validation. It is also possible to specify the type in the instance. XML schema provides an xsi:type attribute that specifies the type of the component. This overrides the assocation between the qualified name of the component and the type as specified in the schema, or it provides a type where the qualified name of the component might not be resolvable into a type.

XML Schema is designed to assign type information as part of validation. Other languages, notably RelaxNG, do not assign type information and have no notion of types. The decision to use types and re-use types across components is an important factor in component version identification because the component definition and the component's type may be versioned separately.

4 Component version identification strategies

The strategy for identifying the version of a component is perhaps the most important decision in designing an XML Language. The use of namespace names, component names, version numbers and type information are all critical in achieving the desired versioning characteristics. The strategies range from many namespaces per version of a language to only 1 namespace for all versions of a language. A few of the most common are listed below and described in more detail later.

Whatever the design chosen, the language designer must decide the component name, namespace name, and any version identifier for new and all existing components.

Elaborating on these designs is illustrative.

4.1 Versioning Strategy #1: all components in new namespace(s) for each version

The following names would be valid:

The 2nd and 3rdexamples shows all the components in the same new namespace, with the 3rd showing a new name as well.. The 4th and 5th example show an additional middle element in 2 different namespace names. The 4th example comes from a namespace name that is in the same domain as the name element’s new namespace name. One reason for 2 namespaces is to modularize the language. The 5th example shows a namespace name from a different domain for the middle.

4.2 Versioning Strategy #2: all new components in new namespace(s) for each compatible version

In this strategy, the following names would be valid:

The 2nd and 3rd example show an additional middle element in 2 different namespace names. The first middle, the 2nd example, comes from a namespace name that is in the same domain as the name element’s namespace name. The 3rd example shows a complete different namespace name for the middle. It is probable that the midns:middle was created by the name author, and the middiffdomain:middle was created by a 3rd party.

4.3 Versioning Strategy #2.5: all new components in new or existing namespace(s) for each compatible version

We have called this strategy "2.5" because it is a mixture of strategy #2 and strategy #3. In this strategy, the following names would be valid:

The 2nd example shows the use of the optional middle name in the name namespace. The 3rd and 4th example show an additional middle element in 2 different namespace names. The first middle, the 3rd example, comes from a namespace name that is in the same domain as the name element’s namespace name. The 4th example shows a complete different namespace name for the middle. It is probable that the midns:middle was created by the name author, and the middiffdomain:middle was created by a 3rd party.

4.4 Versioning Strategy #3: all new components in existing namespace(s) for each compatible version

In this strategy, the following names would be valid:

The 2nd example shows the use of the same namespace because the middle is optional. The 3rd example shows the use of the same namespace because the middle is optional and the middle embedded inside an "extension element". The 4th example shows the use of a new namespace for all the components, such as a mandatory middle name.

4.5 Versioning Strategy #4: all new components in existing or new namespace(s) for each version and a version identifier

Using a version identifier, the name instances would change to show the version of the name they use, such as:

In the last two example, the version number has been changed from 1.0 to 2.0. Incrementing the major part of a version number often indicates an incompatible change. In this case, perhaps it indicates that the middle name is now mandatory where it had previously been optional.

5 Indicating compatibility of changes or extensions

As a language designer will have chosen a component version identification strategy, they must also choose how compatible or incompatible changes will be indicated.

5.1 Compatible

As mentioned in the forwards compatibility section, forwards compatibility requires a substitution mechanism. Ignoring unknown content is a very popular model. It may be specified as the default for any extensions. It could also be specified in an instance where the default is for incompatible versioning. This could be a flag, such as ns:mayIgnore="true".

5.2 Incompatible

A version author can use new namespace names, local names, or version numbers to indicate an incompatible change. An extension author may not have these mechanisms available for indicating an incompatible extension. A language designer that wants to allow extension authors to indicate that an extension is incompatible must provide a mechanism for indicating that consumers must understand the extension, and the consumer must generate an error if it does not understand the extension. If only specific consumers must understand the extension, then the language designer must also provide a mechanism for indicating which consumers. If the language designer has allowed for forwards compatibility, then the forwards compatibility rule must be over-ridden.

Languages with forwards compatibility support MAY provide an override for indicating incompatible extensions but should only do so IF the incompatible extensions can be clearly targeted or scoped.

5.2.1 Must Understand flag

Arguably the simplest and most flexible over-ride of the Must Ignore Unknowns technique is a Must Understand flag that indicates whether the item must be understood. The SOAP, WSDL, and WS-Policy attributes and values for specifying understand are respectively: soap:mustUnderstand=”1”, wsdl:required=”1”, wsp:Usage=”wsp:Required”. SOAP is probably the most common case of a container that provides a Must Understand model. The default value is 0, which is effectively the Must Ignore rule.

A language designer can re-use an existing Must Understand model by constraining their language to an existing Must Understand model. A number of Web services specifications have done this by specifying that the components are SOAP header blocks, which explicitly brings in the SOAP Must Understand model.

A language designer can design a Must Understand model into their language. A Must Understand flag allows the producer to insert extensions into the container and use the Must Understand attribute to over-ride the must Ignore rule. This allows producers to extend instances without changing the extension element’s parent’s namespace, retaining backwards compatibility. Obviously the consumer must be extended to handle new extensions, but there is now a loose coupling between the language’s processing model and the extension’s processing model. A Must Understand flag is provided below:

An example of an instance of a 3rd party indicating that a middle component is an incompatible change:

Specification of a Must Understand flag must be treated carefully as it can be computationally expensive. Typically a processor will either: perform a scan for Must Understand components to ensure it can process the entire text, or incrementally process the instance and is prepared to rollback or undo any processing if an not understood Must Understand is found.

There are other refinements related to Must Understand. One example is providing an element that indicates which extension namespaces must be understood, which avoids the scan of the instance for Must Understand flags.

It is also possible to re-use the SOAP processing model with it's mustUnderstand. Use of a SOAP header for an extension may be because the body was not designed to be extensible, or because the extension is considered semantically separate from the body and will typically be processed differently than the body.

6 XML Schema 1.0

XML Schema provides a variety of mechanisms for extensibility and versioning: wildcards, type extension, type restriction, redefine, substitution groups, and xsi:type attributes. The wildcard construct enables authors to create schemas that are both forwards and backwards-compatible. Generally, a new schema using wildcards is backwards compatible because it will validate old and new instances. The exception is instances that have content that is legal in the wildcard but not in the new content. An example might be a middle name that has structure or digits. However, that scenario means that an author created a middle name instance in the middle name namespace according to one schema AND an author defined a new middle name in the same namespace according to a different schema. Arguably there is an authority over the namespace that will prevent such clashes and so in practice this exception won't happen. Alternatively, we can make a slightly different compatibility guarantee, which is the new schema is backwards compatible with validate old and new instance where new instances do not have any extensions in the defined namespaces. The old schema is forwards compatible because it will validate old and new instances - of course it sees these as current and future instances.

When an author creates a new version, a new schema can created by the replacement of wildcard(s) in the original, with an optional-element, optional-wildcard sequence, in the later schema. The new schema explicitly states the entire new content model, including everything from the original schema as well as the new explicit declaration for middle, and for that reason we call it a "Complete Respecification" of the type.

A new type declared using wildcards could be declared as an explicit <xs:restriction/> of the original type, because every document accepted by the new type is also accepted by the old. XML Schema's type <xs:restriction/> allows alteration of wildcards anywhere in the content model, like Complete Respecification, but allows the original type to be preserved. Alternatively, XML Schema's type extension mechanism <xs:extension/> @@provide ref to Recommendation@@ provides a different way of specifiying a modified type, in which the original content is not restated, but only the new elements are explicitly referenced. The differences are: (1) xs:extension allows new content only at the end of the model and (2) using wildcards as shown above, the original type will accept not only documents in the original language, but also documents containing the middle name, something that's not true in typical uses of xs:extension. Thus the schema author of new version of a type has 3 options outlined above: 1) Complete Respecification without explicit use of xs:restriction; 2) Complete Respecification with explicit use of xs:restriction; 3) xs:extension.

These mechanisms can be combined together. For example, a schema that supports new components in existing or new namespaces and supports multiple schema versions (described in @@) uses wildcards, type extension, and use of Extension elements in instances.

Given an extensibility point that allows different namespaces, the language designer and 3rd parties can now use different namespaces for their versions. In general, an extension can be defined by a new specification that makes a normative reference to the earlier specification and then defines the new content. No permission should be needed from the authors of the specification to make such an extension. In fact, the major design point of XML namespaces is to allow decentralized extensions. The corollary is that permission is required for extensions in the same namespace. A namespace has an owner; non-owners changing the meaning of something can be harmful.

Attribute extensions can be in any namespace because in XML schema, attributes do not have non-determinism (aka Unique Particle Attribution) constraints that elements do. In XML Schema, the attributes are always unordered and the model group for attributes uses a different mechanism for associating attributes with schema types than the model group for elements. We will discuss this important issue later in the finding.

7 Schemas for Version Identification Strategies

7.1 #1: all components in new namespace(s) for each version

Using XML Schema 1.0, the name owner might like to write a schema such as:

The next version of the schema, with middle name added, might look like

This schema is not perhaps quite what is desired because there are now 2 wildcards in the content model, the original wildcard then the new middle and the new wildcard. Type extension does not replace any existing wildcard trailing wildcard with the additive content. An alternative is to not have the wildcard in the first version but that removes forwards compatible extensibility as both sides must have the new schema to understand the type. Because of the type extension problem, the language designer cannot re-use the existing name definition and force a single wildcard at the end. They must create a new schema without any re-use of the previous schema's type information by respecifying the type. They can simply respecify the type or they can use xsd:restriction. Using xsd:restriction has some extra value in that a Schema processor can guarantee that the content model is a true restriction, but in general, respecification with or without xsd:restriction are equivalent.

The new namespace for all components does not allow compatible evolution by the language designer, unless they choose to put new components in a new namespace, which is strategy #2. Additionally, the version 2 schema cannot re-use the existing type definition.

7.2 #2: all new components in new namespace(s) for each compatible version

We previously saw how re-use by importing and extending schemas with wildcards is not possible. In this strategy, the schema designer attempts to insert the new extension in the existing schema definition, like:

The Unique Particle Attribution(UPA) constraint of XML Schema, described in more detail in Unique Particule Attribution, prevents this from working. The problem arises in a version when an optional element is followed by a wildcard. In this example, this occurs when an optional element is added and extensibility is still desired. This is an ungentle introduction to the difference between extensibility and versioning. An optional middle name added into a subsequent version is a good example. Consumers should be able to continue processing if they don’t understand an additional optional middle name, and we want to keep the extensibility point in the new version. We can't write a schema that contains the optional middle name and a wildcard for extensibility. The previous schema schema is roughly what is desired using wildcards, but it is illegal because of the UPA constraint.

The author has 5 options for the v2 schema for name and middle, listed below and detailed subsequently:

  1. the new middle is defined, extensibility is retained, and the new name type does not refer to the new middle;

  2. the new middle is defined, extensibility is lost, and the new name type refers to the new middle as optional;

  3. the new middle is defined, extensibility is retained, and the new name type refers to the new middle as required - the result is that compatibility is lost (essentially strategy #1);

  4. the new middle is defined, extensibility is retained, and there is no new name type

  5. no update to the Schema

If they leave the middle as optional and retain the extensibility point, the best schema that they can write is:

This is not a very helpful XML Schema change. The problem is that they cannot insert the reference to the optional midns:middle element in the name schema and retain the extensibility point because of the aforementioned Unique Particle Attribution Constraint.

The core of the problem is that there is no mechanism for constraining the content of a wildcard. For example, imagine that ns1 contains foo and bar. It is not possible to take the SOAP schema—an example of a schema with a wildcard - and require that ns1:foo element must be a child of the header element and ns1:bar must not be a child of the header element using just W3C XML Schema constructs.   Indeed, the need for this functionality spawned some of the WSDL functionality.

They could decide to lose the extensibility point (option #2), such as

This does lose the possibility for forwards-compatible evolution.

Option #3 is adding a required middle. They must indicate the change is incompatible. A new namespace name for the name element can be created. This is essentially strategy #1, new namespace for all components.

The downsides of the 3 options for new components in new namespace name(s) design have been described. Additionally, the design can result in specifications and namespaces that are inappropriately factored, as related constructs will be in separate namespaces.

7.3 #2.5: All new components in existing or new namespace(s) for each compatible version

It is possible to create Schemas with additional optional components. This requires re-using the namespace name for optional components where possible, and use a new namespace where re-using the namespace is not possible. The re-using namespace rule is:

Good Practice

Re-use namespace names Rule: If a backwards compatible change can be made to a specification, then the old namespace name SHOULD be used in conjunction with XML’s extensibility model.

Strategy #1 uses a new namespace for all existing components and any additions, strategy #2 uses a new namespace for all additions (compatible and incompatible). strategy #3 re-uses namespaces for compatible extensions and uses a new namespace for all incompatible additions. Said slightly differently, strategies #1 and #2 use a new namespace name for any extension and strategy # 3 uses a new namespace only for incompatible change is made.

Earlier examples showed that it is not possible to have a wildcard with ##any (or even ##targetnamespace) following optional elements in the targetnamespace. This strategy is a "middle-ground" strategy, where the ##any is used wherever possible and ##other is used where ##any cannot be used. ##any can be used after mandatory elements or for attributes.

The addition of an optional middle can be done in the same namespace, but the wildcard must change to ##other.

After the 2nd version has been deployed, the versioning strategy is now Versioning Strategy #2: all new components in new namespace(s).

7.4 #3: All new components in existing namespace(s) for each compatible version

It is possible to create Schemas with additional optional components. This requires re-using the namespace name for optional components and special schema design techniques. The re-using namespace rule is:

Good Practice

Re-use namespace names Rule: If a backwards compatible change can be made to a specification, then the old namespace name SHOULD be used in conjunction with XML’s extensibility model.

Strategy #1 uses a new namespace for all existing components and any additions, strategy #2 uses a new namespace for all additions (compatible and incompatible). strategy #3 re-uses namespaces for compatible extensions and uses a new namespace for all incompatible additions. Said slightly differently, strategies #1 and #2 use a new namespace name for any extension and strategy # 3 uses a new namespace only for incompatible change is made.

Good Practice

New namespaces to break Rule: A new namespace name is used when backwards compatibility is not permitted, that is software SHOULD reject texts if it does not understand the new language components.

Earlier examples showed that it is not possible to have a wildcard with ##any (or even ##targetnamespace) following optional elements in the targetnamespace. The solution to this problem is to introduce an element in the schema that will always appear if the extension appears. The content model of the extensibility point is the element + the extension. There are two styles for this. The first, which we will call Extension element style, was published in an earlier version of this Finding in December 2003. It uses an Extensibility element with the extensions nested inside. The second, which we weill call Sentry style, was published in July 2004, then updated on MSDN. It uses a Sentry or Marker element with extensions following it.

A name type with extension elements is

Because each extension in the targetnamespace is inside an Extension element, each subsequent target namespace extensions will increase nesting by another layer. While this layer of nesting per extension is not desirable, it is what can be accomplished today when applying strict XML Schema validation. It seems to at least this author that potentially having multiple nested elements is worthwhile if multiple compatible revisions can be made to a language. This technique allows validation of extensions in the targetnamespace and retaining validation of the targetnamespace itself.

The previous schema allows the following sample namens:

The namespace author can create a schema for this type

The advantage of this design technique is that a forwards and backwards compatible Schema V2 can be written. The V2 schema can validate documents with or without the middle, and the V1 schema can validate documents with or without the middle. This is the only schema design that enables all versions of the language to have complete schemas.

Further, the re-use of the same namespace has better tooling support. Many applications use a single schema to create the equivalent programming constructs. These tools often work best with single namespace support for the “generated” constructs. The re-use of the namespace name allows at least the namespace author to make changes to the namespace and perform validation of the extensions.

An obvious downside of this approach is the complexity of the schema design. Another downside is that changes are linear, so 2 potentially parallel extensions must be nested rather than parallel.

7.4.1 Redefine

The author could use redefine to add the middle in the same namespace. However, the first version does not allow extensions in the same namespace, so this is an incompatible change.

In the previous example, the author of the redefined schema replaced the type with an update. If the author of the nameType wanted to make the change, they could presumably just change the type without using redefine. In cases where the author of the extension is not the author of the base type, then redefine allows them to change the type. Some people may consider this an illegal redefinition of the nameType because they believe that only the namespace owner of the nameType should make changes to the type. Redefine also allows extension and restriction, subject to the limitations of them. Redefine does not help the nameType owner or an extension author create a revised type that refers to any new construct.

7.5 #4: all new components in existing or new namespace(s) for each version and a version identifier

Using a version identifier, the name instances would change to show the version of the name they use, such as:

The last example shows a middle that is a mandatory part of the name, which is indicated by the use of a new major version number. As with Design #2, the schema for the optional middle cannot fully express the content model. A schema for the mandatory middle is

A significant downside with using version identifiers is that software that supports both versions of the name must perform special processing on top of XML and namespaces. For example, many components “bind” XML types into particular programming language types. Custom software must process the version attribute before using any of the “binding” software. In Web services, toolkits often take SOAP body content, parse it into types and invoke methods on the types. There are rarely “hooks” for the custom code to intercept processing between the “SOAP” processing and the “name” processing. Further, if version attributes are used by any 3rd party extensions—say midns:middle has a version—then the schema cannot refer to the correct middle.

7.6 #5: all components in existing namespace(s) for each version and a version identifier

Using a version identifier, the name instances would change to show the version of the name they use, such as:

This is not a very helpful XML Schema change. The problem is that they cannot insert the reference to the optional midns:middle element in the name schema and retain the extensibility point because of the aforementioned Unique Particle Attribution Constraint.

The last example shows that the middle is now a mandatory part of the name. As with Design #2, the schema for the optional middle cannot fully express the content model. A schema for the mandatory middle is

This design has the significant drawback that XML Schema cannot be used for many of the changes. Because the same namespace is used for all versions of the language, then the wildcard namespace attribute must contain ##any. This means that any changes that are compatible, such as the addition of an optional middle in the 2nd example, cannot be completely modeled in XML Schema.

8 Indicating Incompatible changes

A new qualified name can be created by specifying standalone content, respecifying existing content or by some kind of relationship with existing content. A variety of compatible extension mechanisms have been shown. There are more mechanism for incompatible changes in Schema 1.0

8.2 Substitution Groups

Another mechanism for extending a type in XML Schema is substitution groups. Substitution groups enable an element to be declared as substitutable for another. This can only be used for incompatible extensions as the consumer must understand the new element and the schema that contains the substitution type. Substitution groups require that elements are available for substitution, so the name designer must have provided a name element in addition to the name type.

A schema for a substitution group is provided below:

Substitution groups do allow a single extension author to indicate that their changes are mandatory. The limitations are that the extension author has now taken over the type’s extensibility. A visual way of imagining this is that the type tree has now been moved from the language designer over to the extensions author. And the language designer probably does not want their type to be “hijacked”.

However, this is not substantially different than an extension being marked with a “Must Understand”. In either case—with the extensions higher up in the tree (sometimes called top-typing) or lower in the tree (bottom-typing)—a new type is effectively created.

The difference is that there can only be 1 element at the top of an element hierarchy. If multiple mandatory extensions are added, then the only way to compose them together is at the bottom of the type because that is where the extensibility is.

Substitution groups do not allow a language designer and an extension author to incompatibly change the language as they end up conflicting over what to call the name element. Thus substitution groups are a poor mechanism for allowing an extension author to indicate that their changes are incompatible. A Must Understand flag is a superior method because it allows multiple extension authors to mix their mandatory extensions with a language designer’s versioning strategy. Hence language designers should prevent substitution groups and provide a Must Understand flag or other model when they wish to allow 3rd parties to make incompatible changes.

In some cases, a language does not provide a Must Understand mechanism. In the absence of a Must Understand model, the only way to force consumers to reject a message if they don’t understand the extension namespace is to change the namespace name of the root element, but this is rarely desirable.

8.3 Must Understand

Each of the various component identification schemes can support a mustUnderstand flag. Two schema for a Must Understand flag are provided below:

An example of an instance of a 3rd party indicating that a middle component is an incompatible change:

9 Survey of Languages Versioning Strategies

We can examine a variety of languages for their versioning strategies.

Requirement[Docbook 4.5][XHTML 1.0][SVG 1.1][WS-Policy 1.2]
Schema LangRelaxNGXML SchemaDTDXML Schema
3rd party compatibly extendYesYesYesYes
3rd party incompatibly extendNoNoYesYes
stand-aloneYesYes and NoYes or in xhtmlYes
Schema designopen content, patternsModularization, RedefineDTD extensibilitySchema wildcards
Substitution MechanismN/AMust Accept UnknownsMust Accept Unknowns and fallbackMust Accept Unknowns
Component Identification#5#2#2.5#2.5
Incompatible Ext identificationNoN/ArequiredFeatures and requiredExtensions attributesQName of Required assertion
Schema CompletenessN/AComplete including extensionsYespartial

Notes

Requirement[Atom Format][UBL Naming and Design][Reach Interoperability Guidelines][XSLT 2.0]
Schema LangXML Schema, RelaxNGXML Schema, RelaxNGXML SchemaHTML
3rd party compatibly extendYesNoYesYes
3rd party incompatibly extendNoNoYesYes
stand-aloneYesYesYesYes
Schema designWildcards?WildcardsN/A
Substitution MechanismMust Accept UnknownsMust Not Accept UnknownsMust Accept UnknownsMust Accept Unknowns and Fallback provided
Component Identification#4#1#3#5
Incompatible Ext identificationNoneN/A(See Note)MustUnderstand<xsl:message terminate="true">
Schema Completeness?Complete for allComplete for allN/A

10 Unique Particle Attribution

This Finding has spent considerable material describing content models valid under Unique Particle Attribution constraints, and so it is worth describing the W3C XML Schema Unique Particle Attribution constraint in more detail. The reader is reminded that these rules are unique to W3C XML Schema and that other XML Schema languages like RELAX NG do not use these rules and so do not suffer from the contortions one is forced through when using W3C XML Schema. XML DTDs and W3C XML Schema have a rule that requires schemas to have content models valid under the Unique Particle Attribution constraint. From the XML 1.0 specification,

“For example, the content model ((b, c) | (b, d)) is non-deterministic, because given an initial b the XML processor cannot know which b in the model is being matched without looking ahead to see which element follows the b.”

The use of ##any means there are some schemas that we might like to express, but that aren’t allowed.

Good Practice

Follow Unique Particle Attribution constraint: Use of wildcards MUST be follow the Unique Particle Attribution constraint. Location of wildcards, namespace of wildcard extensions, minOccurs and maxOccurs values are constrained, and type restriction is controlled.

As shown earlier, a common design pattern is to provide an extensibility point—not an element - allowing any namespace at the end of a type. This is typically done with <xs:any namespace=”##any”>.

Unique Particle Attribution makes this unworkable as a complete solution in many cases. Firstly, the extensibility point can only occur after required elements in the original schema, limiting the scope of extensibility in the original schema. Secondly, backwards compatible changes require that the added element is optional, which means a minOccurs=”0”. The Unique Particle Attribution constraint prevents us from placing a minOccurs=”0” before an extensibility point of ##any. Thus, when adding an element at an extensibility point, the author can make the element optional and lose the extensibility point, or the author can make the element required and lose backwards compatibility.

11 Other technologies

The W3C XML Schema Working has heard and taken to heart many of these concerns. They have plans to remedy some of these issues in XML Schema 1.1. A Working Draft[@@] and a Guide 2 versioning using the new XML Schema 1.1 features [@@] are available.

A simple analysis of doing compatible extensibility and versioning using RDF and OWL is available [21]. In general, RDF and OWL offer superior mechanisms for extensibility and versioning. RDF and OWL explicitly allow extension components to be added to components. And further, the RDF and OWL model builds in the notion of “Must Ignore Unknowns” as an RDF/OWL processor will absorb the extra components but do nothing with them. An extension author can require that consumers understand the extension by changing the type using a type extension mechanism.  

RELAX NG is another schema language. It explicitly allows extension components to be added to other components as it does not have the Unique Particle Attribution constraint.

12 Conclusion

This Finding describes a number of questions, decisions and rules for using XML, W3C XML Schema, and XML Namespaces in language construction and extension. The main goal of the set of rules is to allow language designers to know their options for language design, and ideally make backwards- and forwards-compatible changes to their languages to achieve loose coupling between systems.

13 References

Atom Format
RFC 4287, the Atom Syndication Format (See http://www.ietf.org/rfc/rfc4287.txt.)
Docbook 4.5
The Docbook Document Type (See http://www.docbook.org/specs/docbook-4.5-spec.html.)
FlexXMLP
Flexible XML Processing Profile. (See http://www.upnp.org/download/draft-goland-fxpp-01.txt.)
WebDAV XMLIgnore post
Yaron GolandXML Ignore proposed for WebDAV (See http://lists.w3.org/Archives/Public/w3c-dist-auth/1997AprJun/0190.html.)
WebDAV
RFC 2518, WebDAV (See http://www.ietf.org/rfc/rfc2518.txt.)
SVG 1.1
W3C Recommendation, SVG 1.1 (See http://www.w3.org/TR/SVG11/.)
UBL Naming and Design
Universal Business Language (UBL) Naming and Design Rules (See http://www.oasis-open.org/committees/download.php/20093/cd-UBL-NDR-2.0.pdf.)
XHTML 1.0
XHTML 1.0. (See http://www.w3.org/TR/xhtml1/.)
Reach Interoperability Guidelines
Reach Interoperability Guidelines (See http://sdec.reach.ie/rigs/rig0006/.)
TBL Mandatory Extensions
Berners-Lee. Web Architecture: Mandatory extensions. (See http://www.w3.org/DesignIssues/Mandatory.html.)
TBL Extensible languages
Berners-Lee. Web Architecture: Extensible languages. (See http://www.w3.org/DesignIssues/Extensible.html.)
TBL Evolution
Berners-Lee. Web Architecture: Evolvability. (See http://www.w3.org/DesignIssues/Evolution.html.)
Web Architecture: Extensible Languages
Berners-Lee and Connolly, ed. Web Architecture: Extensible Languages World Wide Web Consortium, 1998. (See http://www.w3.org/TR/1998/NOTE-webarch-extlang-19980210.)
HTML Document types
Connolly, ed. HTML Document dialects World Wide Web Consortium, 1996. (See http://www.w3.org/MarkUp/WD-doctypes.)
SOAP 1.2
W3C Recommendation, SOAP 1.2 Part 1: Messaging Framework (See http://www.w3.org/TR/SOAP/.)
WSDL 1.1
W3C Note, WSDL 1.1 (See http://www.w3.org/TR/WSDL/.)
WS-Policy 1.2
W3C Note, WS-Policy 1.2 (See http://www.w3.org/Submissions/WS-Policy/.)
XML 1.0
W3C Recommendation, XML 1.0 (See http://www.w3.org/TR/REC-xml.)
xhtmlmodularization
W3C Working Draft, XHTML Modularization (See http://www.w3.org/TR/xhtml-modularization/.)
XInclude
W3C Working Draft, XML Inclusions (See http://www.w3.org/TR-Xinclude.)
XML Namespaces
W3C Recommendation, XML Namespaces (See http://www.w3.org/TR/REC-xml-names.)
XML Schema Part 2
W3C Recommendation, XML Schema, Part 2 (See http://www.w3.org/TR/xmlschema-2.)
XSLT 2.0
W3C Recommendation, XSLT 2.0 (See http://www.w3.org/TR/xslt20/.)
XML Schema Wildcard Test Collection
XML Schema Wildcard Test collection (See http://www.w3.org/XML/2001/05/xmlschema-test-collection/result-ms-wildcards.htm.)
XFront Schema Best Practices
XFront Schema Best Practices (See http://www.xfront.com/BestPracticesHomepage.html.)
XML.com Schema Design Patterns
Dare ObasanjoXML.com Schema design patterns (See http://www.xml.com/pub/a/2002/07/03/schema_design.html.)
Dave Orchard writings on Extensibility and Versioning
Dave Orchard writings on extensibility and versioning (See http://www.pacificspirit.com/Authoring/Compatibility.)

14 Acknowledgements

The author thanks the many reviewers that have contributed to the article, particularly David Bau, William Cox, Ed Dumbill, Chris Ferris, Yaron Goland, Hal Lockhart, Mark Nottingham, Jeffrey Schlimmer, Cliff Schmidt, and Norman Walsh.

A Change Log (Non-Normative)

Changes
WhoWhenWhat
DBO20070704Combined all separate reqs tables into 2 tables. Added version strat/id #2.5