Web addresses in HTML 5

• all characters with codepoints less than or equal to U+0020 (i.e. the C0 control characters)
• all characters with codepoints greater than or equal to U+007% (i.e. U+007?F and all non-ASCII characters in the w)
• U+0022 double quotation mark
• U+0025 percent sign
• U+003C less-than sign
• U+003E greater-than sign mark
• U+005C reverse solidus (backslash)
• U+005E circumflex accent
• U+0060 grave accent
• U+007B left curly bracket
• U+007C vertical line
• U+007D right curly bracket

Otherwise, percent-encode all left and right square brackets.

Percent-encode all occurrences of U+0023 (Number sign, "#") after the first.

Parse w using the grammar in RFC 3986.

If w doesn't match the <URI-reference> production, even after the above changes are made to it, then parsing the Web address fails with an error. [RFC3986]

Let w be the Web address being resolved.

Let encoding be the character encoding of the Web address.

If encoding is UTF-16, then change it to UTF-8.

If the algorithm was invoked with an absolute Web address to use as the base Web address, let base be that absolute Web address.

Otherwise, let base be the base URI of the element, as defined by the XML Base specification, with the base URI of the document entity being defined as the document base Web address of the Document that owns the element. [XMLBASE]

For the purposes of the XML Base specification, user agents must act as if all Document objects represented XML documents.

It is possible for xml:base attributes to be present even in HTML fragments, as such attributes can be added dynamically using script. (Such scripts would not be conforming, however, as xml:base attributes are not allowed in HTML documents.)

Parse w into its component parts.

If parsing w resulted in a <host> component, then replace the matching subtring of w with the string that results from expanding any sequences of percent-encoded octets in that component that are valid UTF-8 sequences into Unicode characters as defined by UTF-8.

If any percent-encoded octets in that component are not valid UTF-8 sequences, then return an error and abort these steps.

Apply the IDNA ToASCII algorithm to the matching substring, with both the AllowUnassigned and UseSTD3ASCIIRules flags set. Replace the matching substring with the result of the ToASCII algorithm.

If ToASCII fails to convert one of the components of the string, e.g. because it is too long or because it contains invalid characters, then return an error and abort these steps [RFC3490].

1. Encode the character into a sequence of octets as defined by UTF-8.
2. Replace the character with the percent-encoded form of those octets. [RFC3986]

1. If the character in question cannot be expressed in the encoding encoding, then replace it with a single 0x3F octet (an ASCII question mark) and skip the remaining substeps for this character.
2. Encode the character into a sequence of octets as defined by the encoding encoding.
3. Replace the character with the percent-encoded form of those octets. [RFC3986]

Apply the algorithm described in RFC 3986 section 5.2 Relative Resolution, using w as the potentially relative URI reference (R), and base as the base URI (Base). [RFC3986]

Apply any relevant conformance criteria of RFC 3986 and RFC 3987, returning an error and aborting these steps if appropriate. [RFC3986] [RFC3987]

For instance, if an absolute URI that would be returned by the above algorithm violates the restrictions specific to its scheme, e.g. a data: URI using the "//" server-based naming authority syntax, then user agents are to treat this as an error instead.

Let result be the target URI (T) returned by the Relative Resolution algorithm.

If result uses a scheme with a server-based naming authority, replace all U+005C REVERSE SOLIDUS (\) characters in result with U+002F SOLIDUS (/) characters.

Return result.