During serialization of XML data, namespace fixup is done as defined in [DOM Level 3 Core] , Appendix B. [DOM Level 2 Core] allows empty strings as a real namespace URI. If the namespaceURI of a Node is empty string, the serialization will treat them as null, ignoring the prefix if any.

LSSerializer accepts any node type for serialization. For nodes of type Document or Entity, well-formed XML will be created when possible (well-formedness is guaranteed if the document or entity comes from a parse operation and is unchanged since it was created). The serialized output for these node types is either as a XML document or an External XML Entity, respectively, and is acceptable input for an XML parser. For all other types of nodes the serialized form is implementation dependent.

Within a Document, DocumentFragment, or Entity being serialized, Nodes are processed as follows

• Document nodes are written, including the XML declaration (unless the parameter "xml-declaration" is set to false) and a DTD subset, if one exists in the DOM. Writing a Document node serializes the entire document.
• Entity nodes, when written directly by LSSerializer.write, outputs the entity expansion but no namespace fixup is done. The resulting output will be valid as an external entity.
• If the parameter " entities" is set to true, EntityReference nodes are serialized as an entity reference of the form " &entityName;" in the output. Child nodes (the expansion) of the entity reference are ignored. If the parameter " entities" is set to false, only the children of the entity reference are serialized. EntityReference nodes with no children (no corresponding Entity node or the corresponding Entity nodes have no children) are always serialized.
• CDATAsections containing content characters that cannot be represented in the specified output encoding are handled according to the " split-cdata-sections" parameter. If the parameter is set to true, CDATAsections are split, and the unrepresentable characters are serialized as numeric character references in ordinary content. The exact position and number of splits is not specified. If the parameter is set to false, unrepresentable characters in a CDATAsection are reported as "wf-invalid-character" errors if the parameter " well-formed" is set to true. The error is not recoverable - there is no mechanism for supplying alternative characters and continuing with the serialization.
• DocumentFragment nodes are serialized by serializing the children of the document fragment in the order they appear in the document fragment.
• All other node types (Element, Text, etc.) are serialized to their corresponding XML source form.

Note: The serialization of a Node does not always generate a well-formed XML document, i.e. a LSParser might throw fatal errors when parsing the resulting serialization.

Within the character data of a document (outside of markup), any characters that cannot be represented directly are replaced with character references. Occurrences of '<' and '&' are replaced by the predefined entities &lt; and &amp;. The other predefined entities (&gt;, &apos;, and &quot;) might not be used, except where needed (e.g. using &gt; in cases such as ']]>'). Any characters that cannot be represented directly in the output character encoding are serialized as numeric character references (and since character encoding standards commonly use hexadecimal representations of characters, using the hexadecimal representation when serializing character references is encouraged).

To allow attribute values to contain both single and double quotes, the apostrophe or single-quote character (') may be represented as "&apos;", and the double-quote character (") as "&quot;". New line characters and other characters that cannot be represented directly in attribute values in the output character encoding are serialized as a numeric character reference.

Within markup, but outside of attributes, any occurrence of a character that cannot be represented in the output character encoding is reported as a DOMError fatal error. An example would be serializing the element <LaCañada/> with encoding="us-ascii". This will result with a generation of a DOMError "wf-invalid-character-in-node-name" (as proposed in " well-formed").

When requested by setting the parameter " normalize-characters" on LSSerializer to true, character normalization is performed according to the definition of fully normalized characters included in appendix E of [XML 1.1] on all data to be serialized, both markup and character data. The character normalization process affects only the data as it is being written; it does not alter the DOM's view of the document after serialization has completed.

Implementations are required to support the encodings "UTF-8", "UTF-16", "UTF-16BE", and "UTF-16LE" to guarantee that data is serializable in all encodings that are required to be supported by all XML parsers. When the encoding is UTF-8, whether or not a byte order mark is serialized, or if the output is big-endian or little-endian, is implementation dependent. When the encoding is UTF-16, whether or not the output is big-endian or little-endian is implementation dependent, but a Byte Order Mark must be generated for non-character outputs, such as LSOutput.byteStream or LSOutput.systemId. If the Byte Order Mark is not generated, a "byte-order-mark-needed" warning is reported. When the encoding is UTF-16LE or UTF-16BE, the output is big-endian (UTF-16BE) or little-endian (UTF-16LE) and the Byte Order Mark is not be generated. In all cases, the encoding declaration, if generated, will correspond to the encoding used during the serialization (e.g. encoding="UTF-16" will appear if UTF-16 was requested).

Namespaces are fixed up during serialization, the serialization process will verify that namespace declarations, namespace prefixes and the namespace URI associated with elements and attributes are consistent. If inconsistencies are found, the serialized form of the document will be altered to remove them. The method used for doing the namespace fixup while serializing a document is the algorithm defined in Appendix B.1, "Namespace normalization", of [DOM Level 3 Core] .

While serializing a document, the parameter "discard-default-content" controls whether or not non-specified data is serialized.

While serializing, errors and warnings are reported to the application through the error handler (LSSerializer.domConfig's " error-handler" parameter). This specification does in no way try to define all possible errors and warnings that can occur while serializing a DOM node, but some common error and warning cases are defined. The types ( DOMError.type) of errors and warnings defined by this specification are:

"no-output-specified" [fatal]

Raised when writing to a LSOutput if no output is specified in the LSOutput.

"unbound-prefix-in-entity-reference" [fatal]

Raised if the configuration parameter " namespaces" is set to true and an entity whose replacement text contains unbound namespace prefixes is referenced in a location where there are no bindings for the namespace prefixes.

"unsupported-encoding" [fatal]

Raised if an unsupported encoding is encountered.

In addition to raising the defined errors and warnings, implementations are expected to raise implementation specific errors and warnings for any other error and warning cases such as IO errors (file not found, permission denied,...) and so on.

See also the Document Object Model (DOM) Level 3 Load and Save Specification.