XSLT 2.0 and XQuery 1.0 Serialization

1 Introduction

This document defines serialization of the W3C XQuery 1.0 and XPath 2.0 Data Model, which is the data model of at least [XPath 2.0], [XSLT 2.0], and [XQuery 1.0], and any other specifications that reference it.

Serialization is the process of converting an instance of the [Data Model] into a sequence of octets. Serialization is well-defined for most data model instances.

Ed. Note:

The document assumes the reader already knows generally what serialization is. A brief explanation will be added, especially to disabuse any reader who thinks it might mean Java (or .NET) serialization.

2 Sequence Normalization

An instance of the data model that is input to the serialization process is a sequence. Prior to serializing a sequence using any of the output methods whose behavior is specified by this document (3 Serialization Parameters) the serializer MUST first compute a normalized sequence for serialization; it is the normalized sequence that is actually serialized. [Definition: The purpose of sequence normalization is to create a sequence that can be serialized as a well-formed XML document or external general parsed entity, that also reflects the content of the input sequence to the extent possible.] [Definition: The result of the sequence normalization process is a result tree.]

The normalized sequence for serialization is constructed by applying all of the following rules in order, with the initial sequence being input to the first step, and the sequence that results from any step being used as input to the subsequent step. For any implementation-defined output method, it is implementation-defined whether this sequence normalization process takes place.

Where the process of converting the input sequence to a normalized sequence indicates that a value MUST be cast to xs:string, that operation is as defined in Section 17.1.2 Casting to xs:string and xdt:untypedAtomic^FO of [Functions and Operators]. The steps in computing the normalized sequence are:

1. If the sequence that is input to serialization is empty, create a sequence S₁ that consists of a zero-length string. Otherwise, copy each item in the sequence that is input to serialization to create the new sequence S₁.

2. For each item in S₁, if the item is atomic, obtain the lexical representation of the item by casting it to an xs:string and copy the string representation to the new sequence; otherwise, copy the item, which will be a node, to the new sequence. The new sequence is S₂.

3. For each subsequence of adjacent strings in S₂, copy a single string to the new sequence equal to the values of the strings in the subsequence concatenated in order, each separated by a single space. Copy all other items to the new sequence. The new sequence is S₃.

4. For each item in S₃, if the item is a string, create a text node in the new sequence whose string value is equal to the string; otherwise, copy the item to the new sequence. The new sequence is S₄.

5. For each item in S₄, if the item is a document node, copy its children to the new sequence; otherwise, copy the item to the new sequence. The new sequence is S₅.

6. It is a serialization error [err:SE0001] if an item in S₅ is an attribute node or a namespace node. Otherwise, construct a new sequence, S₆, that consists of a single document node and copy all the items in the sequence, which are all nodes, as children of that document node.

S₆ is the normalized sequence.

The result tree rooted at the document node that is created by the final step of this sequence normalization process is the instance of the data model to which the rules of the appropriate output method are applied. If the sequence normalization process results in a serialization error, the serializer MUST signal the error.

Note:

The sequence normalization process for a sequence $seq is equivalent to constructing a document node using the XSLT instruction: <xsl:document> <xsl:copy-of select="$seq"/> </xsl:document> or the XQuery expression: document { for $s in $seq return if ($s instance of document-node()) then $s/child::node() else $s } This process results in a serialization error [err:SE0001] with sequences containing parentless attribute and namespace nodes.

3 Serialization Parameters

There are a number of parameters that influence how serialization is performed. Host languages MAY allow users to specify any or all of these parameters, but they are not REQUIRED to be able to do so. However, the host language MUST specify all applicable parameters except doctype-public and doctype-system, which are optional even when they are applicable.

The following serialization parameters are defined:

The value of the method parameter is an expanded QName. If the value has a empty namespace URI, then the local name identifies a method specified in this document and MUST be one of xml, html, xhtml, or text; in this case, the output method specified MUST be used for serializing. If the namespace URI is non-empty, then it identifies an implementation-defined output method; the behavior in this case is not specified by this document.

In those cases where they have no important effect on the content of the serialized result, details of the output methods defined by this specification are left unspecified and are regarded as implementation-dependent. Whether a serializer uses apostrophes or quotation marks to delimit attribute values in the XML output method is an example of such a detail.

The detailed semantics of each parameter will be described separately for each output method for which it is applicable. If the semantics of a parameter are not described for an output method, then it is not applicable to that output method.

4 Phases of Serialization

Following the sequence normalization process described in 2 Sequence Normalization, serialization can be regarded as involving three phases of processing.

For an implementation-defined output method, any of these phases MAY be skipped or MAY be performed in a different order than is specified here. For the output methods defined in this specification, these phases are carried out sequentially as follows:

1. Markup generation produces the character representation of those parts of the serialized result that describe the structure of the normalized sequence. In the cases of the XML, HTML and XHTML output methods, this phase produces the character representations of the following:

   • the document type declaration;

   • start tags and end tags (except for attribute values, whose representation is produced by the character expansion phase);

   • processing instructions; and

   • comments.

   In the cases of the XML and XHTML output methods, this phase also produces the following:

   • the XML or text declaration; and

   • empty element tags (except for the attribute values);

   In the case of the text output method, this phase has no effect.

2. Character expansion is concerned with the representation of characters appearing in text and attribute nodes in the normalized sequence. The substitution processes that apply are listed below, in priority order: a character that is handled by one process in this list will be unaffected by processes appearing later in the list, except that a character affected by Unicode Normalization MAY be affected by creation of CDATA sections and by character escaping:

   1. URI escaping (in the case of URI-valued attributes in the HTML and XHTML output methods), as determined by the escape-uri-attributes parameter. [Definition: URI escaping is a process where non-ASCII characters in URI attribute values are escaped using the method defined by Section 5.4 of [XLink].]

   2. Character mapping, as determined by the use-character-maps parameter. Text nodes that are children of elements specified by the cdata-section-elements parameter are not affected by this step.

   3. Unicode Normalization, if requested by the normalization-form parameter. [Definition: Unicode Normalization is the process of removing alternate representations of equivalent sequences from textual data, to convert the data into a form that can be binary-compared for equivalence, as specified in [UAX #15: Unicode Normalization Forms]. For specific recommendations for character normalization on the World Wide Web, see [Character Model for the World Wide Web 1.0: Normalization].]

      The meanings associated with the possible values of the normalization-form parameter are as follows:

      • NFC specifies the serialized result will be in Normalization Form C, using the rules specified in [Character Model for the World Wide Web 1.0: Normalization].

      • NFD specifies the serialized result will be in Normalization Form D, as specified in [UAX #15: Unicode Normalization Forms].

      • NFKC specifies the serialized result will be in Normalization Form KC, as specified in [UAX #15: Unicode Normalization Forms].

      • NFKD specifies the serialized result will be in Normalization Form KD, as specified in [UAX #15: Unicode Normalization Forms].

      • fully-normalized specifies the serialized result will be in fully normalized form.

      • none specifies that no Unicode Normalization will be applied.

      • An implementation-defined value has an implementation-defined effect.

   4. Creation of CDATA sections, as determined by the cdata-section-elements parameter. Note that this is also affected by the encoding parameter, in that characters not present in the selected encoding cannot be represented in a CDATA section.

   5. Escaping according to XML or HTML rules of special characters that cannot be represented in the selected encoding. For example replacing < with <

   6. If a quote (") character is in an attribute, and the attribute is delimited by quote, the character will be changed to an an apostrophe ('). Likewise, if a apostrophe (') character is in an attribute, and the attribute is delimited by apostrophe, the character will be changed to an an quote (").

3. Encoding, as controlled by the encoding parameter, converts the character stream produced by the previous phases into a octet stream.

   Note:

   Serialization is only defined in terms of encoding the result as a stream of octets. However, a serializer may provide an option that allows the encoding phase to be skipped, so that the result of serialization is a stream of Unicode characters. The effect of any such option is implementation-defined, and a serializer is not required to support such an option.

5 XML Output Method

The XML output method serializes the normalized sequence as an XML entity that MUST satisfy the rules for either a well-formed XML document entity or a well-formed XML external general parsed entity, or both. A serialization error [err:SE0003] results if the serializer is unable to satisfy those rules, except for contents modified by the character expansion phase of serialization, as described in 4 Phases of Serialization, which could result in the serialized output being not well-formed but will not result in a serialization error. If a serialization error results, the serializer MUST signal the error.

If the document node of the normalized sequence has a single element node child and no text node children, then the serialized output is a well-formed XML document entity, and the serialized output MUST conform to the appropriate version of the XML Namespaces Recommendation [XML Names] or [XML Names 1.1]. If the normalized sequence does not take this form, then the serialized output is a well-formed XML external general parsed entity, which, when referenced within a trivial XML document wrapper like this:

<?xml version="version"?>
<!DOCTYPE doc [
<!ENTITY e SYSTEM "entity-URI">
]>
<doc>&e;</doc>

where entity-URI is a URI for the entity, and the value of the version pseudo-attribute is the value of the version parameter, produces a document which MUST itself be a well-formed XML document conforming to the corresponding version of the XML Namespaces Recommendation [XML Names] or [XML Names 1.1].

[Definition: A reconstructed tree may be constructed by parsing the XML document and converting it into an instance of the data model as specified in [Data Model].] The result of serialization MUST be such that the reconstructed tree may be different than the result tree as follows:

• If the document was produced by adding a document wrapper, as described above, then it will contain an extra doc element as the document element.

• The order of attribute and namespace nodes in the two trees MAY be different.

• The following properties of corresponding nodes in the two trees MAY be different:

  • the base-uri property of document nodes and element nodes;

  • the document-uri and unparsed-entities properties of document nodes;

  • the type-name and typed-value properties of element and attribute nodes;

  • the nilled property of element nodes;

  • the content property of text nodes, due to the effect of the indent and use-character-maps parameters.

• The reconstructed tree MAY contain additional attributes and text nodes resulting from the expansion of default and fixed values in its DTD or schema.

• The type annotations of the nodes in the two trees MAY be different. Type annotations in a result tree are discarded when the tree is serialized. Any new type annotations obtained by parsing the document will depend on whether the serialized XML document is assessed against a schema, and this MAY result in type annotations that are different from those in the original result tree.

  Note:

  In order to influence the type annotations in the instance of the data model that would result from processing a serialized XML document, the author of the XSLT stylesheet, XQuery expression or other process might wish to create the instance of the data model that is input to the serialization process so that it makes use of mechanisms provided by [XML Schema], such as xsi:type and xsi:schemaLocation attributes. The serialization process will not automatically create such attributes in the serialized document if those attributes were not part of the result tree that is to be serialized. Similarly, it is possible that an element node in the instance of the data model that is to be serialized has the nilled property with the value true, but no xsi:nil attribute. The serialization process will not create such an attribute in the serialized document simply to reflect the value of the property. The value of the nilled property has no direct effect on the serialized result.

• Additional namespace nodes MAY be present in the reconstructed tree if the serialization process did not undeclare one or more namespaces, as described in 5.1.7 XML Output Method: the undeclare-namespaces Parameter, and the starting instance of the data model contained an element node with a namespace node that declared some prefix, but a child element of that node did not have any namespace node that declared the same prefix.

  The result tree MAY contain namespace nodes that are not present in the reconstructed tree, as the process of creating an instance of the data model MAY ignore namespace declarations in some circumstances. See Section 6.2.3 Construction from an Infoset^DM and Section 6.2.4 Construction from a PSVI^DM of [Data Model] for additional information.

• If the indent parameter has the value yes,

  • additional text nodes consisting of whitespace characters MAY be present in the reconstructed tree; and

  • text nodes in the result tree that contained only whitespace characters MAY correspond to text nodes in thereconstructed tree that contain additional whitespace characters that were not present in the result tree

  See 5.1.3 XML Output Method: the indent Parameter for more information on the indent parameter.

• Additional nodes MAY be present in the reconstructed tree due to the effect of character mapping in the character expansion phase, and the values of attribute nodes and text nodes in the reconstructed tree MAY be different from those in the result tree, due to the effects of URI expansion, character mapping and Unicode Normalization in the character expansion phase of serialization.

  Note:

  The use-character-maps parameter can cause arbitrary characters to be inserted into the serialized XML document in an unescaped form, including characters that would be considered to be part of XML markup. Such characters could result in arbitrary new element nodes, attribute nodes, and so on, in the reconstructed tree that results from processing the serialized XML document.

  A consequence of this rule is that certain characters MUST be output as character references, to ensure that they survive the round trip through serialization and parsing. Specifically, CR, NEL and LINE SEPARATOR characters in text nodes MUST be output respectively as "
", "…", and "
", or their equivalents; while CR, NL, TAB, NEL and LINE SEPARATOR characters in attribute nodes MUST be output respectively as "
", "
", "	", "…", and "
", or their equivalents. In addition, the non-whitespace control characters #x1 through #x1F and #x7F through #x9F in text nodes and attribute nodes MUST be output as character references.

  For example, an attribute with the value "x" followed by "y" separated by a newline will result in the output "x
y" (or with any equivalent character reference). The XML output cannot be "x" followed by a literal newline followed by a "y" because after parsing, the attribute value would be "x y" as a consequence of the XML attribute normalization rules.

  Note:

  XML 1.0 did not permit an XML processor to normalize NEL or LINE SEPARATOR characters to a LINE FEED character. However, if a document entity that specifies version 1.1 invokes an external general parsed entity with no text declaration or a text declaration that specifies version 1.0, the external parsed entity is processed according to the rules of XML 1.1. For this reason, NEL and LINE SEPARATOR characters in text and attribute nodes must always be escaped using character references, regardless of the value of the version parameter. XML 1.0 permitted control characters in the range #x7F through #x9F to appear as literal characters in an XML document, but XML 1.1 requires such characters, other than NEL, to be escaped as character references. An external general parsed entity with no text declaration or a text declaration that specifies a version pseudo-attribute with value 1.0 that is invoked by an XML 1.1 document entity MUST follow the rules of XML 1.1. Therefore, the non-whitespace control characters in the ranges #x1 through #x1F and #x7F through #x9F, other than NEL, MUST always be escaped, regardless of the value of the version parameter.

It is a serialization error [err:SE0004] to specify the doctype-system parameter, or to specify the standalone parameter with a value other than 'none', if the instance of the data model contains text nodes or multiple element nodes as children of the root node. The serializer MUST either signal the error, or recover by ignoring the request to output a document type declaration or standalone parameter.

The result of serialization using the XML output method is not guaranteed to be well-formed XML if character maps have been specified (see 9 Character Maps).

<x:foo xmlns:x="http://example.org/x"
       xmlns:y="http://example.org/y"
       xmlns:z="http://example.org/z">
      <x:bar xmlns:z="">...</x:bar>
</x:foo>

6 XHTML Output Method

The XHTML output method serializes the instance of the data model as XML, using the HTML compatibility guidelines defined in the XHTML specification.

It is entirely the responsibility of the person or process that creates the instance of the data model to ensure that the instance of the data model conforms to the [XHTML 1.0] or [XHTML 1.1] specification. It is not an error if the instance of the data model is invalid XHTML. Equally, it is entirely under the control of the person or process that creates the instance of the data model whether the output conforms to XHTML Strict, XHTML Transitional, XHTML Frameset, or XHTML Basic.

The serialization of the instance of the data model follows the same rules as for the XML output method, with the exceptions noted below. These differences are based on the HTML compatibility guidelines published in Appendix C of [XHTML 1.0], which are designed to ensure that as far as possible, XHTML is rendered correctly on user agents designed originally to handle HTML.

• Given an empty instance of an XHTML element whose content model is not EMPTY (for example, an empty title or paragraph) the serializer MUST NOT use the minimized form. That is, it MUST output <p></p> and not <p />.

• Given an XHTML element whose content model is EMPTY, the serializer MUST use the minimized tag syntax, for example <br />, as the alternative syntax <br></br> allowed by XML gives uncertain results in many existing user agents. The serializer MUST include a space before the trailing />, e.g. <br />, <hr /> and <img src="karen.jpg"  alt="Karen" />.

• The serializer MUST NOT use the entity reference ' which, although legal in XML and therefore in XHTML, is not defined in HTML and is not recognized by all HTML user agents.

• The serializer SHOULD output namespace declarations in a way that is consistent with the requirements of the XHTML DTD if this is possible. The DTD requires the declaration xmlns="http://www.w3.org/1999/xhtml" to appear on the html element, and only on the html element. The serializer MUST output namespace declarations that are consistent with the namespace nodes present in the result tree, but it MUST avoid outputting redundant namespace declarations on elements where the DTD would make them invalid.

  Note:

  If the html element is generated by an XSLT literal result element of the form <html xmlns="http://www.w3.org/1999/xhtml"> ... </html>, or by an XQuery direct element constructor of the same form, then the html element in the result document will have a node name whose prefix is "", which will satisfy the requirements of the DTD. In other cases the prefix assigned to the element may be implementation-dependent.

• If the instance of the data model includes a head element in the XHTML namespace, and the include-content-type parameter has the value yes, the XHTML output method MUST add a meta element as the first child element of the head element, specifying the character encoding actually used.

  For example,

  <head>
  <meta http-equiv="Content-Type" content="text/html; charset=EUC-JP" />
  ...
  

  The content type MUST be set to the value given for the media-type parameter if it has been specified.

  Note:

  The recommended default value of the media-type parameter is text/html. The value application/xhtml+xml, registered in [RFC3236], may also be used. However, it is never acceptable to use application/xhtml+xml unless explicitly requested (among other things, some user agents fail to recognize the charset parameter if the type is not text/html). Guidance on which media type to use can be found in [XHTML Media Types W3C Note 1 August 2002].

  If a meta element has been added to the head element as described above, then any existing meta element child of the head element having an http-equiv attribute with the value "Content-Type" MUST be discarded.

  Note:

  This process removes possible parameters in the attribute value. For example, <meta http-equiv="Content-Type" content="text/html;version='3.0'" /> in the data model instance would be replaced by, <meta http-equiv="Content-Type" content="text/html;charset=utf-8" />

• If the escape-uri-attributes parameter has the value yes, the XHTML output method MUST apply URI escaping to URI attribute values, except that relative URIs MUST NOT be absolutized.

  Note:

  This escaping is deliberately confined to non-ASCII characters, because escaping of ASCII characters is not always appropriate, for example when URIs or URI fragments are interpreted locally by the HTML user agent. Even in the case of non-ASCII characters, escaping can sometimes cause problems. More precise control of URI escaping is therefore available by setting escape-uri-attributes to no, and controlling the escaping of URIs by means of the fn:escape-uri function defined in [Functions and Operators].

• If the indent parameter has the value yes, the serializer may add or remove whitespace as it serializes the result tree, so long as it does not change the way that a conforming HTML user agent would render the output.

  Note:

  This rule can be satisfied by observing the following constraints: Whitespace must only be added before or after an element, or adjacent to an existing whitespace character. Whitespace must not be added or removed adjacent to an inline element. The inline elements are those elements in the XHTML namespace in the %inline category of any of the XHTML 1.0 DTD's, in the %inline.class category of the XHTML 1.1 DTD, and elements in the XHTML namespace with local names ins and del if they are used as inline elements (i.e., if they do not contain element children). Whitespace must not be added or removed inside a formatted element, the formatted elements being those in the XHTML namespace with local names pre, script, style, and textarea. The HTML definition of whitespace is different from the XML definition: see section 9.1 of [HTML] 4.01 specification.

Note:

As with the XML output method, the XHTML output method specifies that an XML declaration will be output unless it is suppressed using the omit-xml-declaration parameter. Appendix C.1 of [XHTML 1.0] provides advice on the consequences of including, or omitting, the XML declaration.

Note:

Appendix C of [XHTML 1.0] describes a number of compatibility guidelines for users of XHTML who wish to render their XHTML documents with HTML user agents. In some cases, such as the guideline on the form empty elements should take, only the serialization process itself has the ability to follow the guideline. In such cases, those guidelines are reflected in the requirements on the serializer described above. In all other cases, the guidelines can be adhered to by the instance of the data model that is input to the serialization process. The guideline on the use of whitespace characters in attribute values is one such example. It is the responsibility of the person or process that creates the instance of the data model that is input to the serialization process to ensure it is created in a way that is consistent with the guidelines. No serialization error results if the input instance of the data model does not adhere to the guidelines.

7 HTML Output Method

The HTML output method serializes the instance of the data model as HTML.

<xsl:stylesheet version="2.0"
                xmlns:xsl="http://www.w3.org/1999/XSL/Transform">

<xsl:output method="html" version=”4.0”/>

<xsl:template match="/">
<html>
<xsl:apply-templates/>
</html>
</xsl:template>

...

</xsl:stylesheet>

The version attribute of the xsl:output element indicates the version of the HTML Recommendation [HTML] to which the serialized result is to conform. If the serializer does not support the version of HTML specified by this attribute (which corresponds to the version parameter defined in this specification), it MUST signal a serialization error [err:SE0013].

<script>if (a &lt; b) foo()</script>

<script><![CDATA[if (a < b) foo()]]></script>

<script>if (a < b) foo()</script>

<script type="text/javascript">
      document.write ("<em>This won't work</em>")
</script>

<script type="text/javascript">
      document.write ("<em>This won't work</em>")
</script>

<script type="text/javascript">
      document.write ("&lt;em&gt;This won't work&lt;/em&gt;")
</script>

<script type="text/javascript">
      document.write ("&lt;em&gt;This will work&lt;\/em&gt;")
</script>

<OPTION selected="selected">

<OPTION selected>

<BODY bgcolor='&{{randomrbg}};'>

<BODY bgcolor='&{randomrbg};'>

<HEAD>
<META http-equiv="Content-Type" content="text/html; charset=EUC-JP">
...

8 Text Output Method

The text output method serializes the instance of the data model by outputting the string value of the document node created by sequence normalization, without any escaping.

A newline character in the instance of the data model MAY be output using any character sequence that is conventionally used to represent a line ending in the chosen system environment.

The media-type parameter is applicable for the text output method. See 3 Serialization Parameters for more information.

The encoding parameter identifies the encoding that the text output method MUST use to convert sequences of characters to sequences of bytes. Serializers are REQUIRED to support values of UTF-8 and UTF-16. A serialization error [err:SE0007] occurs if the serializer does not support the encoding specified by the encoding parameter. The serializer MUST signal the error. If the instance of the data model contains a character that cannot be represented in the encoding that the serializer is using for output, the serializer MUST signal a serialization error [err:SE0008].

The normalization-form parameter is applicable for the text output method. The values NFC and none MUST be supported by the serializer. A serialization error [err:SE0011] results if the value of the normalization-form parameter specifies a normalization form that is not supported by the serializer; the serializer MUST signal the error.

The use-character-maps parameter is applicable for the text output method. See 9 Character Maps for more information.

The byte-order-mark parameter is applicable for the text output method. See 3 Serialization Parameters for more information.

9 Character Maps

The use-character-maps parameter is a list of characters and corresponding string substitutions.

Character maps allow a specific character appearing in a text or attribute node in the instance of the data model to be replaced with a specified string of characters during serialization. The string that is substituted is output "as is," and the serializer performs no checks that the resulting document is well-formed. This mechanism can therefore be used to introduce arbitrary markup in the serialized output. See Section 20.1 Character Maps^XT of [XSLT 2.0] for examples of using character mapping in XSLT.

Character mapping is applied to the characters that actually appear in a text or attribute node in the instance of the data model, before any other serialization operations such as escaping or Unicode Normalization are applied. If a character is mapped, then it is not subjected to XML or HTML escaping, nor to Unicode Normalization. The string that is substituted for a character is not validated or processed in any way by the serializer, except for translation into the target encoding. In particular, it is not subjected to XML or HTML escaping, it is not subjected to Unicode Normalization, and it is not subjected to further character mapping.

Character mapping is not applied to characters in text nodes whose parent elements are listed in the cdata-section-elements parameter, nor to characters for which output escaping has been disabled (disabling output escaping is an [XSLT 2.0] feature), nor to characters in attribute values that are subject to the URI escaping defined for the HTML and XHTML output methods, unless URI escaping has been disabled using the escape-uri-attributes parameter in the output definition.

On serialization, occurrences of a character specified in the use-character-maps in text nodes and attribute values are replaced by the corresponding string from the use-character-maps parameter.

Note:

Using a character map can result in non-well-formed documents if the string contains XML-significant characters. For example, it is possible to create documents containing unmatched start and end tags, references to entities that are not declared, or attributes that contain tags or unescaped quotation marks.

If a character is mapped, then it is not subjected to XML or HTML escaping.

A serialization error [err:SE0008] occurs if character mapping causes the output of a string containing a character that cannot be represented in the encoding that the serializer is using for output. The serializer MUST signal the error.

10 Conformance

Serialization is intended primarily as a component that can be used by other specifications. Therefore, this document relies on specifications that use it to specify conformance criteria for Serialization in their respective environments. Specifications that set conformance criteria for their use of Serialization MUST NOT change the semantic definitions of Serialization as given in this specification, except by subsetting and/or compatible extensions.

Specifications that set conformance criteria for their use of Serialization MUST NOT change the semantic definitions of Serialization as given in this specification, except by subsetting and/or compatible extensions. Thus, it is the responsibility of those specifications to avoid any behavior that would conflict with the semantic definition of Serialization.