PIs, Comments, and Outside of Document Element

• Loss of XML declaration <?xml version="1.0"?>
• Loss of DTD <!DOCTYPE doc SYSTEM "doc.dtd">
• Normalization of whitespace outside of document element (first character of both canonical forms is '<'; single line breaks separate PIs and comments outside of document element)
• Loss of whitespace between PITarget and its data
• Retention of whitespace inside PI data
• Comment removal from uncommented canonical form, including delimiter for comments outside document element (the last character in both canonical forms is '>')

Whitespace in Document Content

• Retain all whitespace between consecutive start tags, clean or dirty
• Retain all whitespace between consecutive end tags, clean or dirty
• Retain all whitespace between end tag/start tag pair, clean or dirty
• Retain all whitespace in character content, clean or dirty

Note: For "After c14n", the input document and canonical form are identical. Both end with '>' character.

• Retain white space inside text, i.e. inbetween A and B but remove all leading and trailing whitespace

Start and End Tags

• Empty element conversion to start-end tag pair. I.e. <e1 /> becomes <e1></e1>
• Normalization of whitespace in start and end tags. I.e.. <e2 ></e2> becomes <e2></e2>
• Relative order of namespace and attribute axes, I.e. in <e5> the namespace declarations precede the attribute declarations.
• Lexicographic ordering of namespace and attribute axes
• Retention of namespace prefixes from original document
• Elimination of superfluous namespace declarations
• Addition of default attribute. I.e. <e9 attr="default">.

Note: Some start tags in the canonical form are very long, but each start tag in this example is entirely on a single line.

Note: In e5, b:attr precedes a:attr because the primary key is namespace URI not namespace prefix, and attr2 precedes b:attr because the default namespace is not applied to unqualified attributes (so the namespace URI for attr2 is empty).

Character Modifications and Character References

• Character reference replacement
• Attribute value delimiters set to quotation marks (double quotes)
• Attribute value normalization
• CDATA section replacement
• Encoding of special characters as character references in attribute values (&amp;, &lt;, &quot;, &#xD;, &#xA;, &#x9;)
• Encoding of special characters as character references in text (&amp;, &lt;, &gt;, &#xD;)

Note: The last element, normId, is well-formed but violates a validity constraint for attributes of type ID. For testing canonical XML implementations based on validating processors, remove the line containing this element from the input and canonical form. In general, XML consumers should be discouraged from using this feature of XML.

Note: Whitespace character references other than &#x20; are not affected by attribute value normalization [[!XML10]].

Note: In the canonical form, the value of the attribute named attr in the element norm begins with a space, an apostrophe (single quote), then four spaces before the first character reference.

Note: The expr attribute of the second compute element contains no line breaks.

Entity References

• Internal parsed entity reference replacement
• External parsed entity reference replacement (including whitespace outside elements and PIs)
• External unparsed entity reference

UTF-8 Encoding

• Effect of transcoding from a sample encoding to UTF-8

Note: The content of the doc element are two octets whose hexadecimal values are C2 and A9, which is the UTF-8 encoding of the UCS codepoint for the copyright sign (©).

Namespaces declarations are pushed down

Default namespace declarations

Sorting namespace declarations

In this example there are four prefixes declarations: "a", "b", "c" and "d". They map to namespace URIs "http://z3", "http://z2", "http://z1" and "http://z0" respectively.

Notice the following things in the default canonicalization ("After c14n"):

• In "a:foo" all the namespace declarations appear before the attribute declarations, even though they were mixed up in the original document.
• In "a:foo" the attributes are sorted in this order - "c:att3", "b:att1", "b:att2". This is because attributes are sorted by namespaceURI as primary key and localName as secondary key. Since "c" is http://z1 and "b" is http://z2, all the "c" attributes appear before "b" attributes. Note that within the "b" attributes "att1" appears before "att2".
• In "a:foo" all the namespace declarations are sorted by prefixes i.e. "a", "b", "c", and NOT by namespace URIs.

• New prefixes are assigned in the order in which they are visibly utilized:
  1. At first "a:foo" visibly utilizes prefixes "a", "b" and "c". These are then sorted by namespace URI, resulting in "c", "b", "a". Then they are assigned prefixes "http://z1"->"n0", "http://z2"->"n1" and "http://z3"->"n2"
  2. After that "c:bar" visibly utilizes prefix "d". This gets assigned prefix "n3"
• In "a:foo" all rewritten prefixes are sorted by prefixes i.e. "n0", "n1", "n2".

Namespace Re-declarations

In this example there are three prefixes "a", "b" and the default prefix. The "foo" element defines them to be "http://z3", "http://z2" and "" respectively. But the "bar" redeclares these prefixes to "http://z2", "http://z3" abd "http://z0" respectively.

Notice the following things in the default canonicalization ("After c14n"):

• The redeclarations of the prefixes by the "bar" element are not removed by canonicalization, since these declarations are not superfluous.

And notice these things in canonicalization with prefix rewriting ("After c14n with PrefixRewrite"):

• With prefix rewriting, the new prefixes are never redeclared. Instead the same prefix gets rewritten to two different values depending on what namespace it currently declares. For example, for element "foo" attribute "a:att1" gets rewritten to "n2:att1", whereas for element "bar" attribute "a:att1" gets rewritten to "n1:att1". Essentially with prefix rewriting, the current prefixes are completely ignored, and canonicalization assigns new prefixes based on namespace values.

Superfluous Namespace declarations

In this example there are five prefixes "a", "b", "c", "d" and the default prefix and they are all declared to the same namespace URI "http://z0". The "a" prefix is defined twice, one in the "foo" element, and then again in "c:bar" element; obviously the definition of "a" in "c:bar" is unnecessary.

Notice the following things in the default canonicalization ("After c14n"):

• It does not attempt to recognize that "a", "b", "c", "d" and default are pointing to the same thing.
• It removes the extra declaration of "xmlns:a" in the "c:bar" as that is superfluous.

• All the file prefixes "a", "b", "c", "d" are rewritten to the same prefix - "z0".

Special namespaces "xml", "xsi", "xsd"

In this example there are are three special namespace declaration the "xml" namespace used in the attribute xml:id="23" and also the "xsi" and "xsd" namespaces used in xsi:type="xsd:string".

Canonicalization only treats "xml" as a special namespace. It is never rewritten by prefix-rewriting. "xsi" and "xsd" are treated as regular namespaces.

Notice the following things in the default canonicalization ("After c14n"):

• It does not consider the "xsd" prefix as visibly utilized since it is embedded in the content. That is why it removes the "xsd" namespace declaration completely.

• It also removes the "xsd" declaration completely.
• It doesn't change the "xml" prefix to anything else, also it doesn't emit the namespace declaration for the "xml" prefix.
• But it does rewrite the "xsi" prefix to "ns1".

• Look at the CanonicalzationMethod in c14nQname.xml. it lists "xsi:type" as a Qualified Attribute in the QNameAware element.
• So now the canonicalization looks into the "xsi:type" for prefixes. It finds the "xsd" element and considers it visibly utilized, and hence emits the "xsd" namespace declaration.

• Look at the CanonicalzationMethod in c14nPrefixQname.xml. it lists "xsi:type" as a Qualified Attribute in the QNameAware element. It also PrefixRewite=sequential
• So now "xsd" is considered visibly utilized, and it is rewritten to "n1". I.e. xsi:type="xsd:string" changes to n2:type="n1:string".

Prefixes in Element content

Notice the following things in the default canonicalization ("After c14n"):

• It does not consider the "xsd", "soap-env" and "b" prefixes as visibly utilized since they are embedded in the content. That is why it removes these namespace declarations completely.

Notice the following things in the default canonicalization ("After c14n with QNameAware <a:bar>"):

• It considers the "xsd", as visbily utilized in element, but not "soap-env" and "b" prefixes as visibly utilized since they are embedded in the XPath content.

Notice the following things in the default canonicalization ("After c14n with QNameAware <a:bar> and <dsig2:IncludedXPath>"):

• All three embedded prefixes "xsd", "soap-env" and "b" prefixes as considered as visibly utilized.
• prefix "c" is not considered visibly utilized since it is inside an XPath sring. Single and double quoted strings inside an XPath expression are not searched for prefixes.
• "child:" is not a prefix, it is an XPath axis, as indicated by the double colon.

Notice the following things in the default canonicalization ("After c14n with QNameAware <a:bar> and <dsig2:IncludedXPath> and PrefixRewrite"):

• All three embedded prefixes "xsd", "soap-env" and "b" prefixes as considered as visibly utilized. They are rewritten to "n1", "n3" and "n2" respectively.
• The text content is modified because of prefix rewriting.
• Even though the "xsd" prefix is rewritten to "n1", the 'xsd:string' is left unchanged, as it inside a single quoted string.