Escaping (cont.)

With the reference processing model, escapes become unambiguous:

• Independent of encoding, always refer to UCS
• No need to parse HTML/XML/CSS when transcoding (e.g. at a proxy)
• Escapes allow use of any Unicode character in any document (even when encoded in US-ASCII only)

NCRs are decimal (A is A) or, in HTML 4.0 and XML, hexadecimal (A is A); now in SGML corrigendum

One character = one escape, not two for surrogate pairs