22.4.4 HTML Cross-reference 8-bit Character Expansion

Usually, characters other than plain 7-bit ASCII are transformed into the corresponding Unicode code point(s) in Normalization Form C, which uses precomposed characters where available. (This is the normalization form recommended by the W3C and other bodies.) This holds when that code point is 0xffff or less, as it almost always is.

These will then be further transformed by the rules above into the string `_xxxx', where xxxx is the code point in hex.

For example, combining this rule and the previous section:

     @node @b{A} @TeX{} @u{B} @point{}@enddots{}
     => A-TeX-B_0306-_2605_002e_002e_002e

Notice: 1) @enddots expands to three periods which in turn expands to three `_002e''s; 2) @u{B} is a `B' with a breve accent, which does not exist as a pre-accented Unicode character, therefore expands to `B_0306' (B with combining breve).

When the Unicode code point is above 0xffff, the transformation is `__xxxxxx', that is, two leading underscores followed by six hex digits. Since Unicode has declared that their highest code point is 0x10ffff, this is sufficient. (We felt it was better to define this extra escape than to always use six hex digits, since the first two would nearly always be zeros.)

For the definition of Unicode Normalization Form C, see Unicode report UAX#15, http://www.unicode.org/reports/tr15/. Many related documents and implementations are available elsewhere on the web.