#

Chapter 23

Special Areas and Format Characters

This chapter describes several kinds of characters that have special properties as well as areas of the codespace that are set aside for special purposes.

The Unicode Standard contains code positions for the 64 control characters and the DEL character found in ISO standards and many vendor character sets. The choice of control function associated with a given character code is outside the scope of the Unicode Standard, with the exception of those control characters specified in this chapter.

Layout controls are not themselves rendered visibly, but influence the behavior of algorithms for line breaking, word breaking, glyph selection, and bidirectional ordering.

Surrogate code points are restricted use. The numeric values for surrogates are used in pairs in UTF-16 to access 1,048,576 supplementary code points in the range U+10000..U+10FFFF.

Variation selectors allow the specification of standardized variants of characters. This ability is particularly useful where the majority of implementations would treat the two variants as two forms of the same character, but where some implementations need to differentiate between the two. By using a variation selector, such differentiation can be made explicit.

Private-use characters are reserved for private use. Their meaning is defined by private agreement.

Noncharacters are code points that are permanently reserved and will never have characters assigned to them.

The Specials block contains characters that are neither graphic characters nor traditional controls.

Tag characters were intended to support a general scheme for the internal tagging of text streams in the absence of other mechanisms, such as markup languages. The use of tag characters for language tagging is deprecated.

#23.1 Control Codes

There are 65 code points set aside in the Unicode Standard for compatibility with the C0 and C1 control codes defined in the ISO/IEC 2022 framework. The ranges of these code points are U+0000..U+001F, U+007F, and U+0080..U+009F, which correspond to the 8-bit controls 00₁₆ to 1F₁₆ (C0 controls), 7F₁₆ (delete), and 80₁₆ to 9F₁₆ (C1 controls), respectively. For example, the 8-bit legacy control code character tabulation (or tab) is the byte value 09₁₆; the Unicode Standard encodes the corresponding control code at U+0009.

The Unicode Standard provides for the intact interchange of these code points, neither adding to nor subtracting from their semantics. The semantics of the control codes are generally determined by the application with which they are used. However, in the absence of specific application uses, they may be interpreted according to the control function semantics specified in ISO/IEC 6429:1992.

In general, the use of control codes constitutes a higher-level protocol and is beyond the scope of the Unicode Standard. For example, the use of ISO/IEC 6429 control sequences for controlling bidirectional formatting would be a legitimate higher-level protocol layered on top of the plain text of the Unicode Standard. Higher-level protocols are not specified by the Unicode Standard; their existence cannot be assumed without a separate agreement between the parties interchanging such data.

#23.1.1 Representing Control Sequences

There is a simple, one-to-one mapping between 7-bit (and 8-bit) control codes and the Unicode control codes: every 7-bit (or 8-bit) control code is numerically equal to its corresponding Unicode code point. For example, if the ASCII line feed control code (0A₁₆) is to be used for line break control, then the text “WX<LF>YZ” would be transmitted in Unicode plain text as the following coded character sequence: <0057, 0058, 000A, 0059, 005A>.

Control sequences that are part of Unicode text must be represented in terms of the Unicode encoding forms. For example, suppose that an application allows embedded font information to be transmitted by means of markup using plain text and control codes. A font tag specified as “^ATimes^B”, where ^A refers to the C0 control code 01₁₆ and ^B refers to the C0 control code 02₁₆, would then be expressed by the following coded character sequence: <0001, 0054, 0069, 006D, 0065, 0073, 0002>. The representation of the control codes in the three Unicode encoding forms simply follows the rules for any other code points in the standard:

UTF-8: <01 54 69 6D 65 73 02>

UTF-16: <0001 0054 0069 006D 0065 0073 0002>

UTF-32: <00000001 00000054 00000069 0000006D

00000065 00000073 00000002>

#Escape Sequences. Escape sequences are a particular type of protocol that consists of the use of some set of ASCII characters introduced by the escape control code, 1B₁₆, to convey extra-textual information. When converting escape sequences into and out of Unicode text, they should be converted on a character-by-character basis. For instance, “ESC-A” <1B 41> would be converted into the Unicode coded character sequence <001B, 0041>. Interpretation of U+0041 as part of the escape sequence, rather than as latin capital letter a, is the responsibility of the higher-level protocol that makes use of such escape sequences. This approach allows for low-level conversion processes to conformantly convert escape sequences into and out of the Unicode Standard without needing to actually recognize the escape sequences as such.

If a process uses escape sequences or other configurations of control code sequences to embed additional information about text (such as formatting attributes or structure), then such sequences constitute a higher-level protocol that is outside the scope of the Unicode Standard.

#23.1.2 Specification of Control Code Semantics

Several control codes are commonly used in plain text, particularly those involved in line and paragraph formatting. The use of these control codes is widespread and important to interoperability. Therefore, the Unicode Standard specifies semantics for their use with the rest of the encoded characters in the standard. Table 23-1 lists those control codes.

Code Point	Abbreviation	ISO/IEC 6429 Name
U+0009	HT	character tabulation (tab)
U+000A	LF	line feed
U+000B	VT	line tabulation (vertical tab)
U+000C	FF	form feed
U+000D	CR	carriage return
U+001C	FS	information separator four
U+001D	GS	information separator three
U+001E	RS	information separator two
U+001F	US	information separator one
U+0085	NEL	next line

The control codes in Table 23-1 have the Bidi_Class property values of S, B, or WS, rather than the default of BN used for other control codes. (See Unicode Standard Annex #9, “Unicode Bidirectional Algorithm.”) In particular, U+001C..U+001E and U+001F have the Bidi_Class property values B and S, respectively, so that the Bidirectional Algorithm recognizes their separator semantics.

The control codes U+0009..U+000D and U+0085 have the White_Space property. They also have line breaking property values that differ from the default CM value for other control codes. (See Unicode Standard Annex #14, “Unicode Line Breaking Algorithm.”)

U+0000 null may be used as a Unicode string terminator, as in the C language. Such usage is outside the scope of the Unicode Standard, which does not require any particular formal language representation of a string or any particular usage of null.

#Newline Function. In particular, one or more of the control codes U+000A line feed, U+000D carriage return, and the Unicode equivalent of the EBCDIC next line can encode a newline function. A newline function can act like a line separator or a paragraph separator, depending on the application. See Section 23.2, Layout Controls, for information on how to interpret a line or paragraph separator. The exact encoding of a newline function depends on the application domain. For information on how to identify a newline function, see Section 5.8, Newline Guidelines.

#23.2 Layout Controls

The effect of layout controls is specific to particular text processes. As much as possible, layout controls are transparent to those text processes for which they were not intended. In other words, their effects are mutually orthogonal.

#23.2.1 Line and Word Breaking

This subsection summarizes the intended behavior of certain layout controls which affect line and word breaking. Line breaking and word breaking are distinct text processes. Although a candidate position for a line break in text often coincides with a candidate position for a word break, there are also many situations where candidate break positions of different types do not coincide. The implications for the interaction of layout controls with text segmentation processes are complex. For a full description of line breaking, see Unicode Standard Annex #14, “Unicode Line Breaking Algorithm.” For a full description of other text segmentation processes, including word breaking, see Unicode Standard Annex #29, “Unicode Text Segmentation.”

#No-Break Space. U+00A0 NO-BREAK SPACE has the same width as U+0020 SPACE, but the NO-BREAK SPACE indicates that, under normal circumstances, no line breaks are permitted between it and surrounding characters, unless the preceding or following character is a line or paragraph separator or space or zero width space. For a complete list of space characters in the Unicode Standard, see Table 6-2.

#Word Joiner. U+2060 WORD JOINER behaves like U+00A0 NO-BREAK SPACE in that it indicates the absence of line breaks; however, the word joiner has no width. The function of the character is to indicate that line breaks are not allowed between the adjoining characters, except next to hard line breaks. For example, the word joiner can be inserted after the fourth character in the text “base+delta” to indicate that there should be no line break between the “e” and the “+”. The word joiner can be used to prevent line breaking with other characters that do not have nonbreaking variants, such as U+2009 THIN SPACE or U+2015 HORIZONTAL BAR, by bracketing the character.

The word joiner must not be confused with the zero width joiner or the combining grapheme joiner, which have very different functions. In particular, inserting a word joiner between two characters has no effect on their ligating and cursive joining behavior. The word joiner should be ignored in contexts other than line breaking. Note in particular that the word joiner is ignored for word segmentation. (See Unicode Standard Annex #29, “Unicode Text Segmentation.”)

#Zero Width No-Break Space. In addition to its primary meaning of byte order mark (see “Byte Order Mark” in Section 23.8, Specials), the code point U+FEFF possesses the semantics of ZERO WIDTH NO-BREAK SPACE, which matches that of word joiner. Until Unicode 3.2, U+FEFF was the only code point with word joining semantics, but because it is more commonly used as byte order mark, the use of U+2060 WORD JOINER to indicate word joining is strongly preferred for any new text. Implementations should continue to support the word joining semantics of U+FEFF for backward compatibility.

#Zero Width Space. The U+200B ZERO WIDTH SPACE indicates a word break or line break opportunity, even though there is no intrinsic width associated with this character. Zero-width space characters are intended to be used in languages that have no visible word spacing to represent word break or line break opportunities, such as Thai, Myanmar, Khmer, and Japanese.

The “zero width” in the character name for ZWSP should not be understood too literally. While this character ordinarily does not result in a visible space between characters, text justification algorithms may add inter-character spacing (letter spacing) between characters separated by a ZWSP. For example, in Table 23-2, the row labeled “Display 4” illustrates incorrect suppression of inter-character spacing in the context of a ZWSP.

This behavior for ZWSP contrasts with that for fixed-width space characters, such as U+2002 EN SPACE. Such spaces have a specified width that is typically unaffected by justification and which should not be increased (or reduced) by inter-character spacing (see Section 6.2, General Punctuation).

In some languages such as German and Russian, increased letter spacing is used to indicate emphasis. Implementers should be aware of this issue.

#Zero-Width Spaces and Joiner Characters. The zero-width spaces are not to be confused with the zero-width joiner characters. U+200C ZERO WIDTH NON-JOINER and U+200D ZERO WIDTH JOINER have no effect on word or line break boundaries, and ZERO WIDTH NO-BREAK SPACE and ZERO WIDTH SPACE have no effect on joining or linking behavior. The zero-width joiner characters should be ignored when determining word or line break boundaries. See “Cursive Connection” later in this section.

#Hyphenation. U+00AD SOFT HYPHEN (SHY) indicates an intraword break point, where a line break is preferred if a word must be hyphenated or otherwise broken across lines. Such break points are generally determined by an automatic hyphenator. SHY can be used with any script, but its use is generally limited to situations where users need to override the behavior of such a hyphenator. The visible rendering of a line break at an intraword break point, whether automatically determined or indicated by a SHY, depends on the surrounding characters, the rules governing the script and language used, and, at times, the meaning of the word. The precise rules are outside the scope of this standard, but see Unicode Standard Annex #14, “Unicode Line Breaking Algorithm,” for additional information. A common default rendering is to insert a hyphen before the line break, but this is insufficient or even incorrect in many situations.

Contrast this usage with U+2027 HYPHENATION POINT, which is used for a visible indication of the place of hyphenation in dictionaries. For a complete list of dash characters in the Unicode Standard, including all the hyphens, see Table 6-3.

The Unicode Standard includes two nonbreaking hyphen characters: U+2011 NON-BREAKING HYPHEN and U+0F0C TIBETAN MARK DELIMITER TSHEG BSTAR. See Section 13.4, Tibetan, for more discussion of the Tibetan-specific line breaking behavior.

#Line and Paragraph Separator. The Unicode Standard provides two unambiguous characters, U+2028 LINE SEPARATOR and U+2029 PARAGRAPH SEPARATOR, to separate lines and paragraphs. They are considered the default form of denoting line and paragraph boundaries in Unicode plain text. A new line is begun after each LINE SEPARATOR. A new paragraph is begun after each PARAGRAPH SEPARATOR. As these characters are separator codes, it is not necessary either to start the first line or paragraph or to end the last line or paragraph with them. Doing so would indicate that there was an empty paragraph or line following. The PARAGRAPH SEPARATOR can be inserted between paragraphs of text. Its use allows the creation of plain text files, which can be laid out on a different line width at the receiving end. The LINE SEPARATOR can be used to indicate an unconditional end of line.

A paragraph separator indicates where a new paragraph should start. Any interparagraph formatting would be applied. This formatting could cause, for example, the line to be broken, any interparagraph line spacing to be applied, and the first line to be indented. A line separator indicates that a line break should occur at this point; although the text continues on the next line, it does not start a new paragraph—no interparagraph line spacing or paragraphic indentation is applied. For more information on line separators, see Section 5.8, Newline Guidelines.

#23.2.2 Cursive Connection and Ligatures

In some fonts for some scripts, consecutive characters in a text stream may be rendered via adjacent glyphs that cursively join to each other, so as to emulate connected handwriting. For example, cursive joining is implemented in nearly all fonts for the Arabic scripts and in a few handwriting-like fonts for the Latin script.

Cursive rendering is implemented by joining glyphs in the font and by using a process that selects the particular joining glyph to represent each individual character occurrence, based on the joining nature of its neighboring characters. This glyph selection is implemented in the rendering engine, typically using information in the font.

In many cases there is an even closer binding, where a sequence of characters is represented by a single glyph, called a ligature. Ligatures can occur in both cursive and noncursive fonts. Where ligatures are available, it is the task of the rendering system to select a ligature to create the most appropriate line layout. However, the rendering system cannot define the locations where ligatures are possible because there are many languages in which ligature formation requires more information. For example, in some languages, ligatures are never formed across syllable boundaries.

On occasion, an author may wish to override the normal automatic selection of connecting glyphs or ligatures. Typically, this choice is made to achieve one of the following effects:

• Cause nondefault joining appearance (for example, as is sometimes required in writing Persian using the Arabic script)
• Exhibit the joining-variant glyphs themselves in isolation
• Request a ligature to be formed where it normally would not be
• Request a ligature not to be formed where it normally would be

The Unicode Standard provides two characters that influence joining and ligature glyph selection: U+200C ZERO WIDTH NON-JOINER and U+200D ZERO WIDTH JOINER. The zero width joiner and non-joiner request a rendering system to have more or less of a connection between characters than they would otherwise have. Such a connection may be a simple cursive link, or it may include control of ligatures.

The zero width joiner and non-joiner characters are designed for use in plain text; they should not be used where higher-level ligation and cursive control is available. (See the W3C specification, “Unicode in XML and Other Markup Languages,” for more information.) Moreover, they are essentially requests for the rendering system to take into account when laying out the text; while a rendering system should consider them, it is perfectly acceptable for the system to disregard these requests.

The ZWJ and ZWNJ are designed for marking the unusual cases where ligatures or cursive connections are required or prohibited. These characters are not to be used in all cases where ligatures or cursive connections are desired; instead, they are meant only for overriding the normal behavior of the text.

#Joiner. U+200D ZERO WIDTH JOINER requests a more connected rendering of adjacent characters. In particular:

• If the two characters could form a ligature but do not normally, ZWJ requests that the ligature be used.
• Otherwise, if either of the characters could cursively connect but do not normally, ZWJ requests that each of the characters take a cursive-connection form where possible.
• A typical use of ZWJ is to show the connected form of a character without a visible neighbor, as shown in Figure 23-2.

In a sequence like <X, ZWJ, Y>, where a cursive form exists for X but not for Y, the presence of ZWJ requests a cursive form for X. Otherwise, where neither a ligature nor a cursive connection is available, the ZWJ has no effect. In other words, given the three broad categories below, ZWJ requests that glyphs in the highest available category (for the given font) be used:

1. Ligated
2. Cursively connected
3. Unconnected

#Non-joiner. U+200C ZERO WIDTH NON-JOINER is intended to break both cursive connections and ligatures in rendering.

ZWNJ requests that glyphs in the lowest available category (for the given font) be used.

For those unusual circumstances where someone wants to forbid ligatures in a sequence XY but promote cursive connection, the sequence <X, ZWJ, ZWNJ, ZWJ, Y> can be used. The ZWNJ breaks ligatures, while the two adjacent joiners cause the X and Y to take adjacent cursive forms (where they exist). Similarly, if someone wanted to have X take a cursive form but Y be isolated, then the sequence <X, ZWJ, ZWNJ, Y> could be used (as in previous versions of the Unicode Standard). Examples are shown in Figure 23-3.

#Cursive Connection. For cursive connection, the joiner and non-joiner characters typically do not modify the contextual selection process itself, but instead change the context of a particular character occurrence. By providing a non-joining adjacent character where the adjacent character otherwise would be joining, or vice versa, they indicate that the rendering process should select a different joining glyph. This process can be used in two ways: to prevent a cursive joining or to exhibit joining glyphs in isolation.

In Figure 23-1, the insertion of the ZWNJ overrides the normal cursive joining of sad and lam.

In Figure 23-2, the normal display of ghain without ZWJ before or after it uses the nominal (isolated) glyph form. When preceded and followed by ZWJ characters, however, the ghain is rendered with its medial form glyph in isolation.

The examples in Figure 23-1 and Figure 23-2 are adapted from the Iranian national coded character set standard, ISIRI 3342, which defines ZWNJ and ZWJ as “pseudo space” and “pseudo connection,” respectively.

#Examples. Figure 23-3 provides samples of desired renderings when the joiner or non-joiner is inserted between two characters. The examples presume that all of the glyphs are available in the font. If, for example, the ligatures are not available, the display would fall back to the unligated forms. Each of the entries in the first column of Figure 23-3 shows two characters in visual display order. The column headings show characters to be inserted between those two characters. The cells below show the respective display when the joiners in the heading row are inserted between the original two characters.

For backward compatibility, between Arabic characters a ZWJ acts just like the sequence <ZWJ, ZWNJ, ZWJ>, preventing a ligature from forming instead of requesting the use of a ligature that would not normally be used. As a result, there is no plain text mechanism for requesting the use of a ligature in Arabic text.

#Transparency. The property value of Joining_Type = Transparent applies to characters that should not interfere with cursive connection, even when they occur in sequence between two characters that are connected cursively. These include all nonspacing marks and most format control characters, except for ZWJ and ZWNJ themselves. Note, in particular, that enclosing combining marks are also transparent as regards cursive connection. For example, using U+20DD COMBINING ENCLOSING CIRCLE to circle an Arabic letter in a sequence should not cause that Arabic letter to change its cursive connections to neighboring letters. See Section 9.2, Arabic, for more on joining classes and the details regarding Arabic cursive joining.

#Joiner and Non-joiner in Indic Scripts. In Indic text, the ZWJ and ZWNJ are used to request particular display forms. A ZWJ after a sequence of consonant plus virama requests what is called a “half-form” of that consonant. A ZWNJ after a sequence of consonant plus virama requests that conjunct formation be interrupted, usually resulting in an explicit virama on that consonant. There are a few more specialized uses as well. For more information, see the discussions in Chapter 12, South and Central Asia-I.

#Implementation Notes. For modern font technologies, such as OpenType or AAT, font vendors should add ZWJ to their ligature mapping tables as appropriate. Thus, where a font had a mapping from “f” + “i” to fi, the font designer should add the mapping from “f” + ZWJ + “i” to fi. In contrast, ZWNJ will normally have the desired effect naturally for most fonts without any change, as it simply obstructs the normal ligature/cursive connection behavior. As with all other alternate format characters, fonts should use an invisible zero-width glyph for representation of both ZWJ and ZWNJ.

#Filtering Joiner and Non-joiner. ZERO WIDTH JOINER and ZERO WIDTH NON-JOINER are format control characters. As such, and in common with other format control characters, they are ordinarily ignored by processes that analyze text content. For example, a spell-checker or a search operation should filter them out when checking for matches. There are exceptions, however. In particular scripts—most notably the Indic scripts—ZWJ and ZWNJ have specialized usages that may be of orthographic significance. In those contexts, blind filtering of all instances of ZWJ or ZWNJ may result in ignoring distinctions relevant to the user’s notion of text content. Implementers should be aware of these exceptional circumstances, so that searching and matching operations behave as expected for those scripts.

#23.2.3 Prepended Concatenation Marks

The Unicode Standard includes a number of format characters that combine with a sequence of one or more characters of a specific category, where the entire sequence is then rendered as a single, ligated display unit. These prepended concatenation marks always have a visible display, despite their status as format characters. An example is U+0601 ARABIC SIGN SANAH, the Arabic year sign, which is followed by a sequence of Arabic-Indic digits to format a year. See “Signs Spanning Numbers” in Section 9.2, Arabic for more discussion of the use and display of these signs in the Arabic script.

A sequence anchored by a prepended concatenation mark behaves somewhat analogously to a combining character sequence, with the prepended concatenation mark standing in as the “base character” of the sequence. The entire sequence is a tightly defined syntactic element and ends with the first character that is not in the specified category (often a digit) required for that particular mark. Inserting any other character, even a default ignorable code point, will disrupt the unit. The entire sequence constitutes a single unit for display; however, unlike a combining character sequence, the sequence introduced by a prepended concatenation mark does not form a grapheme cluster for the purposes of cursor movement and similar editing processes. The overall behavior more closely resembles that of normal ligatures in text.

Many of the prepended concatenation marks occur in the Arabic script, and require an unbroken sequence of decimal digits in the Arabic script. In other instances, as for U+070F SYRIAC ABBREVIATION MARK, the expectation is that they will be followed by an unbroken sequence of Syriac letters.

U+2044 FRACTION SLASH behaves somewhat similarly to prepended concatenation marks. Although it is categorized as an ordinary math symbol and not as a format character, it is intended to result in a ligated unit fraction (such as ¾ instead of 3/4) when positioned between two unbroken sequences of decimal digits (gc=Nd). As for the character sequence following a prepended concatenation mark, the digit sequences for a fraction slash are interrupted even by a default ignorable code point such as U+2060 WORD JOINER.

#23.2.4 Combining Grapheme Joiner

U+034F COMBINING GRAPHEME JOINER (CGJ) is used to affect the collation of adjacent characters for purposes of language-sensitive collation and searching. It is also used to distinguish sequences that would otherwise be canonically equivalent.

Formally, the combining grapheme joiner is not a format control character, but rather a combining mark. It has the General_Category value gc = Mn and the canonical combining class value ccc = 0.

As a result of these properties, the presence of a combining grapheme joiner in the midst of a combining character sequence does not interrupt the combining character sequence; any process that is accumulating and processing all the characters of a combining character sequence would include a combining grapheme joiner as part of that sequence. This differs from the behavior of most format control characters, whose presence would interrupt a combining character sequence.

In addition, because the combining grapheme joiner has the canonical combining class of 0, canonical reordering will not reorder any adjacent combining marks around a combining grapheme joiner. (See the discussion of canonical ordering in Section 3.11, Normalization Forms.) In turn, this means that insertion of a combining grapheme joiner between two combining marks will prevent normalization from switching the positions of those two combining marks, regardless of their own combining classes.

#Blocking Reordering. The CGJ has no visible glyph and no other format effect on neighboring characters but simply blocks reordering of combining marks. It can therefore be used as a tool to distinguish two alternative orderings of a sequence of combining marks for some exceptional processing or rendering purpose, whenever normalization would otherwise eliminate the distinction between the two sequences.

For example, using CGJ to block reordering is one way to maintain distinction between differently ordered sequences of certain Hebrew accents and marks. These distinctions are necessary for analytic and text representational purposes. However, these characters were assigned fixed-position combining classes despite the fact that they interact typographically. As a result, normalization treats differently ordered sequences as equivalent. In particular, the sequence

<lamed, patah, hiriq, finalmem>

is canonically equivalent to

<lamed, hiriq, patah, finalmem>

because the canonical combining classes of U+05B4 HEBREW POINT HIRIQ and U+05B7 HEBREW POINT PATAH are distinct. However, the sequence

<lamed, patah, CGJ, hiriq, finalmem>

is not canonically equivalent to the other two. The presence of the combining grapheme joiner, which has ccc = 0, blocks the reordering of hiriq before patah by canonical reordering and thus allows a patah following a hiriq and a patah preceding a hiriq to be reliably distinguished, whether for display or for other processing.

The use of CGJ with double diacritics is discussed in Section 7.9, Combining Marks; see Figure 7-11.

#CGJ and Collation. The Unicode Collation Algorithm normalizes Unicode text strings before applying collation weighting. The combining grapheme joiner is ordinarily ignored in collation key weighting in the UCA. However, whenever it blocks the reordering of combining marks in a string, it affects the order of secondary key weights associated with those combining marks, giving the two strings distinct keys. That makes it possible to treat them distinctly in searching and sorting without having to tailor the weights for either the combining grapheme joiner or the combining marks.

The CGJ can also be used to prevent the formation of contractions in the Unicode Collation Algorithm. For example, while “ch” is sorted as a single unit in a tailored Slovak collation, the sequence <c, CGJ, h> will sort as a “c” followed by an “h”. The CGJ can also be used in German, for example, to distinguish in sorting between “ü” in the meaning of u-umlaut, which is the more common case and often sorted like <u,e>, and “ü” in the meaning u-diaeresis, which is comparatively rare and sorted like “u” with a secondary key weight. This also requires no tailoring of either the combining grapheme joiner or the sequence. Because CGJ is invisible and has the Default_Ignorable_Code_Point property, data that are marked up with a CGJ should not cause problems for other processes.

It is possible to give sequences of characters that include the combining grapheme joiner special tailored weights. Thus the sequence <c, CGJ, h> could be weighted completely differently from the contraction “ch” or from the way “c” and “h” would have sorted without the contraction. However, such an application of CGJ is not recommended. For more information on the use of CGJ with sorting, matching, and searching, see Unicode Technical Report #10, “Unicode Collation Algorithm.”

#Rendering. For rendering, the combining grapheme joiner is invisible. However, some older implementations may treat a sequence of grapheme clusters linked by combining grapheme joiners as a single unit for the application of enclosing combining marks. For more information on grapheme clusters, see Unicode Technical Report #29, “Unicode Text Segmentation.” For more information on enclosing combining marks, see Section 3.11, Normalization Forms.

#CGJ and Joiner Characters. The combining grapheme joiner must not be confused with the zero width joiner or the word joiner, which have very different functions. In particular, inserting a combining grapheme joiner between two characters should have no effect on their ligation or cursive joining behavior. Where the prevention of line breaking is the desired effect, the word joiner should be used. For more information on the behavior of these characters in line breaking, see Unicode Standard Annex #14, “Unicode Line Breaking Algorithm.”

#23.2.5 Bidirectional Ordering Controls

Bidirectional ordering controls are used in the Bidirectional Algorithm, described in Unicode Standard Annex #9, “Unicode Bidirectional Algorithm.” Systems that handle right-to-left scripts such as Arabic, Syriac, and Hebrew, for example, should interpret these format control characters. The bidirectional ordering controls are shown in Table 23-3.

Code	Name	Abbreviation
U+061C	ARABIC LETTER MARK	ALM
U+200E	LEFT-TO-RIGHT MARK	LRM
U+200F	RIGHT-TO-LEFT MARK	RLM
U+202A	LEFT-TO-RIGHT EMBEDDING	LRE
U+202B	RIGHT-TO-LEFT EMBEDDING	RLE
U+202C	POP DIRECTIONAL FORMATTING	PDF
U+202D	LEFT-TO-RIGHT OVERRIDE	LRO
U+202E	RIGHT-TO-LEFT OVERRIDE	RLO
U+2066	LEFT-TO-RIGHT ISOLATE	LRI
U+2067	RIGHT-TO-LEFT ISOLATE	RLI
U+2068	FIRST STRONG ISOLATE	FSI
U+2069	POP DIRECTIONAL ISOLATE	PDI

As with other format control characters, bidirectional ordering controls affect the layout of the text in which they are contained but should be ignored for other text processes, such as sorting or searching. However, text processes that modify text content must maintain these characters correctly, because matching pairs of bidirectional ordering controls must be coordinated, so as not to disrupt the layout and interpretation of bidirectional text. Each instance of a LRE, RLE, LRO, or RLO is normally paired with a corresponding PDF. Likewise, each instance of an LRI, RLI, or FSI is normally paired with a corresponding PDI.

U+200E LEFT-TO-RIGHT MARK, U+200F RIGHT-TO-LEFT MARK, and U+061C ARABIC LETTER MARK have the semantics of an invisible character of zero width, except that these characters have strong directionality. They are intended to be used to resolve cases of ambiguous directionality in the context of bidirectional texts; they are not paired. Unlike U+200B ZERO WIDTH SPACE, these characters carry no word breaking semantics. (See Unicode Standard Annex #9, “Unicode Bidirectional Algorithm,” for more information.)

#23.2.6 Stateful Format Controls

The Unicode Standard contains a small number of paired stateful controls. These characters are used in pairs, with an initiating character (or sequence) and a terminating character. Even when these characters are not supported by a particular implementation, complications can arise due to their paired nature. Whenever text is cut, copied, pasted, or deleted, these characters can become unpaired. To avoid this problem, ideally both any copied text and its context (site of a deletion, or target of an insertion) would be modified so as to maintain all pairings that were in effect for each piece of text. This process can be quite complicated, however, and is not often done—or is done incorrectly if attempted.

The paired stateful controls recommended for use are listed in Table 23-4.

Characters	Documentation
Bidi Overrides, Embeddings, and Isolates	Section 23.2, Layout Controls; UAX #9
Annotation Characters	Section 23.8, Specials
Musical Beams and Slurs	Section 21.2, Western Musical Symbols

The bidirectional overrides, embeddings, and isolates, as well as the annotation characters are reasonably robust, because their behavior terminates at paragraph boundaries. Paired format controls for representation of beams and slurs in music are recommended only for specialized musical layout software, and also have limited scope.

Bidirectional overrides, embeddings, and isolates are default ignorable (that is, Default_Ignorable_Code_Point = True); if they are not supported by an implementation, they should not be rendered with a visible glyph. The paired stateful controls for musical beams and slurs are likewise default ignorable.

The annotation characters, however, are different. When they are used and correctly interpreted by an implementation, they separate annotation text from the annotated text, and the fully rendered text will typically distinguish the two parts quite clearly. Simply omitting any display of the annotation characters by an implementation which does not interpret them would have the potential to cause significant misconstrual of text content. Hence, the annotation characters are not default ignorable; an implementation which does not interpret them should render them with visible glyphs, using one of the techniques discussed in Section 5.3, Unknown and Missing Characters. See “Annotation Characters” in Section 23.8, Specials for more discussion.

Other paired stateful controls in the standard are deprecated, and their use should be avoided. They are listed in Table 23-5.

Characters	Documentation
Deprecated Format Characters	Section 23.3, Deprecated Format Characters
U+E0001 LANGUAGE TAG	Section 23.9, Tag Characters

The tag characters, originally intended for the representation of language tags, are particularly fragile under editorial operations that move spans of text around. See Section 5.10, Language Information in Plain Text, for more information about language tagging.

#23.3 Deprecated Format Characters

#23.3.1 Deprecated Format Characters: U+206A–U+206F

Three pairs of deprecated format characters are encoded in this block:

• Symmetric swapping format characters used to control the glyphs that depict characters such as “(” (The default state is activated.)
• Character shaping selectors used to control the shaping behavior of the Arabic compatibility characters (The default state is inhibited.)
• Numeric shape selectors used to override the normal shapes of the Western digits (The default state is nominal.)

The use of these character shaping selectors and codes for digit shapes is strongly discouraged in the Unicode Standard. Instead, the appropriate character codes should be used with the default state. For example, if contextual forms for Arabic characters are desired, then the nominal characters should be used, not the presentation forms with the shaping selectors. Similarly, if the Arabic digit forms are desired, then the explicit characters should be used, such as U+0660 ARABIC-INDIC DIGIT ZERO.

#Symmetric Swapping. The symmetric swapping format characters are used in conjunction with the class of left- and right-handed pairs of characters (symmetric characters), such as parentheses. The characters thus affected are listed in Section 4.7, Bidi Mirrored. They indicate whether the interpretation of the term LEFT or RIGHT in the character names should be interpreted as meaning opening or closing, respectively. They do not nest. The default state of symmetric swapping may be set by a higher-level protocol or standard, such as ISO 6429. In the absence of such a protocol, the default state is activated.

From the point of encountering U+206A INHIBIT SYMMETRIC SWAPPING format character up to a subsequent U+206B ACTIVATE SYMMETRIC SWAPPING (if any), the symmetric characters will be interpreted and rendered as left and right.

From the point of encountering U+206B ACTIVATE SYMMETRIC SWAPPING format character up to a subsequent U+206A INHIBIT SYMMETRIC SWAPPING (if any), the symmetric characters will be interpreted and rendered as opening and closing. This state (activated) is the default state in the absence of any symmetric swapping code or a higher-level protocol.

#Character Shaping Selectors. The character shaping selector format characters are used in conjunction with Arabic presentation forms. During the presentation process, certain letterforms may be joined together in cursive connection or ligatures. The shaping selector codes indicate that the character shape determination (glyph selection) process used to achieve this presentation effect is to be either activated or inhibited. The shaping selector codes do not nest.

From the point of encountering a U+206C INHIBIT ARABIC FORM SHAPING format character up to a subsequent U+206D ACTIVATE ARABIC FORM SHAPING (if any), the character shaping determination process should be inhibited. If the backing store contains Arabic presentation forms (for example, U+FE80..U+FEFC), then these forms should be presented without shape modification. This state (inhibited) is the default state in the absence of any character shaping selector or a higher-level protocol.

From the point of encountering a U+206D ACTIVATE ARABIC FORM SHAPING format character up to a subsequent U+206C INHIBIT ARABIC FORM SHAPING (if any), any Arabic presentation forms that appear in the backing store should be presented with shape modification by means of the character shaping (glyph selection) process.

The shaping selectors have no effect on nominal Arabic characters (U+0660..U+06FF), which are always subject to character shaping (glyph selection).

#Numeric Shape Selectors. The numeric shape selector format characters allow the selection of the shapes in which the digits U+0030..U+0039 are to be rendered. These format characters do not nest.

From the point of encountering a U+206E NATIONAL DIGIT SHAPES format character up to a subsequent U+206F NOMINAL DIGIT SHAPES (if any), the European digits (U+0030..U+0039) should be depicted using the appropriate national digit shapes as specified by means of appropriate agreements. For example, they could be displayed with shapes such as the Arabic-Indic digits (U+0660..U+0669). The actual character shapes (glyphs) used to display national digit shapes are not specified by the Unicode Standard.

From the point of encountering a U+206F NOMINAL DIGIT SHAPES format character up to a subsequent U+206E NATIONAL DIGIT SHAPES (if any), the European digits (U+0030..U+0039) should be depicted using glyphs that represent the nominal digit shapes shown in the code tables for these digits. This state (nominal) is the default state in the absence of any numeric shape selector or a higher-level protocol.

#23.4 Variation Selectors

Characters in the Unicode Standard can be represented by a wide variety of glyphs, as discussed in Chapter 2, General Structure. Occasionally the need arises in text processing to restrict or change the set of glyphs that are to be used to represent a character. Normally such changes are indicated by choice of font or style in rich text documents. In special circumstances, such a variation from the normal range of appearance needs to be expressed side-by-side in the same document in plain text contexts, where it is impossible or inconvenient to exchange formatted text. For example, in languages employing the Mongolian script, sometimes a specific variant range of glyphs is needed for a specific textual purpose for which the range of “generic” glyphs is considered inappropriate.

Variation selectors provide a mechanism for specifying a restriction on the set of glyphs that are used to represent a particular character. They also provide a mechanism for specifying variants, such as for CJK ideographs and Mongolian letters, that have essentially the same semantics but substantially different ranges of glyphs.

#Variation Sequence. A variation sequence is a two-character sequence in which a variation selector follows an initial character. Each variation sequence defines a variant of the initial character. The initial character must have the following properties:

• D50: Graphic character (gc=L, M, N, P, S, Zs)
• Not a Variation_Selector
• ccc=0 (does not reorder)
• NFD_QC=Yes (does not decompose)
• NFC_QC=Yes (does not get consumed in composition)

For the definition of NFD_QC and NFC_QC, see Section 5.7.5, “Decompositions and Normalization,” in Unicode Standard Annex #44, “Unicode Character Database.”

For example, the following types of characters cannot be the initial character of a variation sequence: Control codes, format control characters, most diacritics, some Indic dependent vowels, most viramas, Hangul Jamo medial vowels, and canonical composite characters.

These constraints are required because it is important that variation sequences remain stable under normalization, and that the effect of a variation selector can always be characterized as unambiguously applying to a single character. Versions of the Unicode Standard prior to Version 16.0 had a more limited statement of constraints on variation sequences.

In a variation sequence the variation selector affects the appearance of the initial character. Such changes in appearance may, in turn, have a visual impact on subsequent characters, particularly combining characters applied to that initial character. For example, if the initial character changes shape, that should result in a corresponding change in shape or position of applied combining marks. If the initial character changes color, as can be the case for emoji variation sequences, the color may also change for applied combining marks. If the initial character changes in advance width, that would also change the positioning of subsequent spacing characters.

In particular, the emoji variation sequences for digits, U+0023 “#” NUMBER SIGN, and U+002A “*” ASTERISK are intended to affect the color, size, and positioning of U+20E3 ⃣ COMBINING ENCLOSING KEYCAP when applied to those initial characters. For example, the variation sequence <0023, FE0F> selects the emoji presentation variant for “#”. The sequence <0023, FE0F, 20E3> should show the enclosing keycap with an appropriate emoji style, matching the “#” in color, shape, and positioning. Shape changes for variation sequences, with or without additional combining marks, may also result in an increase of advance width; thus, each of the sequences <0023, FE0F>, <0023, 20E3>, and <0023, FE0F, 20E3> may have a distinct advance width, differing from U+0023 alone.

The use of variation selectors is not intended as a general extension mechanism for the character encoding. Combinations of particular initial characters plus particular variation selectors have no effect on display unless they occur in pre-defined lists maintained by the Unicode Consortium. The three sanctioned lists are as follows:

Standardized variation sequences are defined in the file StandardizedVariants.txt in the Unicode Character Database.

Emoji variation sequences are defined in the file emoji-variation-sequences.txt, associated with Unicode Technical Standard #51, “Unicode Emoji.”

Ideographic variation sequences are defined by the registration process defined in Unicode Technical Standard #37, “Unicode Ideographic Variation Database,” and are listed in the Ideographic Variation Database.

Only those three types of variation sequences are sanctioned for use by conformant implementations. In all other cases, use of a variation selector character does not change the visual appearance of the preceding initial character from what it would have had in the absence of the variation selector.

The variation selectors themselves are combining marks of combining class 0 and are default ignorable. Thus, if the variation sequence is not supported, the variation selector should be invisible and ignored. This does not preclude modes or environments where the variation selectors should be given visible appearance. For example, a “Show Hidden” mode could reveal the presence of such characters with specialized glyphs, or a particular environment could use or require a visual indication of an initial character (such as a wavy underline) to show that it is part of a standardized variation sequence that cannot be supported by the current font.

The standardization or support of a particular variation sequence does not limit the set of glyphs that can be used to represent the initial character alone. If a user requires a visual distinction between a character and a particular variant of that character, then fonts must be used to make that distinction. The existence of a variation sequence does not preclude the later encoding of a new character with distinct semantics and a similar or overlapping range of glyphs.

#CJK Compatibility Ideographs. There are 1,002 standardized variation sequences for CJK compatibility ideographs. One sequence is defined for each CJK compatibility ideograph in the Unicode Standard. These sequences are defined to address a normalization issue for these ideographs.

Implementations or users sometimes need a CJK compatibility ideograph to be distinct from its corresponding CJK unified ideograph. For example, a distinct glyphic form may be expected for a particular text. However, CJK compatibility ideographs have canonical equivalence mappings to their corresponding CJK unified ideograph, which means that such distinctions are lost whenever Unicode normalization is applied. Using the variation sequence preserves the distinction found in the original, non-normalized text, even when normalization is later applied.

Because variation sequences are not affected by Unicode normalization, an implementation which uses the corresponding standardized variation sequence can safely maintain the intended distinction for that CJK compatibility ideograph, even in plain text.

It is important to distinguish standardized variation sequences for CJK compatibility ideographs from the variation sequences that are registered in the Ideographic Variation Database (IVD). The former are normalization-stable representations of the CJK compatibility ideographs; they are defined in StandardizedVariants.txt, and there is precisely one variation sequence for each CJK compatibility ideograph. The latter are also stable under normalization, but correspond to implementation-specific glyphs in a registry entry.

#Representative Glyphs for Variants. Representative glyphs for most of the standardized variation sequences are included directly in the code charts. See “Standardized Variation Sequences” in Section 24.1, Character Names List for an explanation of the conventions used to identify such sequences in the code charts. Emoji variation sequences, which often require large, colorful glyphs for their representation, can be found instead in the emoji charts. See Appendix B.3, Other Unicode Online Resources.

Representative glyphs for ideographic variation sequences are located in the pertinent registrations associated with the Ideographic Variation Database.

#Mongolian. For the behavior of older implementations of Mongolian using variation selectors, see the discussion of Mongolian free variation selectors in Section 13.5, Mongolian.

#23.5 Private-Use Characters

Private-use characters are assigned Unicode code points whose interpretation is not specified by this standard and whose use may be determined by private agreement among cooperating users. These characters are designated for private use and do not have defined, interpretable semantics except by private agreement.

Private-use characters are often used to implement end-user defined characters (EUDC), which are common in East Asian computing environments.

No charts are provided for private-use characters, as any such characters are, by their very nature, defined only outside the context of this standard.

Three distinct blocks of private-use characters are provided in the Unicode Standard: the primary Private Use Area (PUA) in the BMP and two supplementary Private Use Areas in the supplemental planes.

All code points in the blocks of private-use characters in the Unicode Standard are permanently designated for private use. No assignment to a particular standard set of characters will ever be endorsed or documented by the Unicode Consortium for any of these code points.

Any prior use of a character as a private-use character has no direct bearing on any eventual encoding decisions regarding whether and how to encode that character. Standardization of characters must always follow the normal process for encoding of new characters or scripts.

#Properties. No private agreement can change which character codes are reserved for private use. However, many Unicode algorithms use the General_Category property or properties which are derived by reference to the General_Category property. Private agreements may override the General_Category or derivations based on it, except where overriding is expressly disallowed in the conformance statement for a specific algorithm. In other words, private agreements may define which private-use characters should be treated like spaces, digits, letters, punctuation, and so on, by all parties to those private agreements. In particular, when a private agreement overrides the General_Category of a private-use character from the default value of gc = Co to some other value such as gc = Lu or gc = Nd, such a change does not change its inherent identity as a private-use character, but merely specifies its intended behavior according to the private agreement.

For all other properties the Unicode Character Database also provides default values for private-use characters. Except for normalization-related properties, these default property values should be considered informative. They are intended to allow implementations to treat private-use characters in a consistent way, even in the absence of a particular private agreement, and to simplify the use of common types of private-use characters. Those default values are based on typical use-cases for private-use characters. Implementations may freely change or override the default values according to their requirements for private use. For example, a private agreement might specify that two private-use characters are to be treated as a case mapping pair, or a private agreement could specify that a private-use character is to be rendered and otherwise treated as a combining mark.

To exchange private-use characters in a semantically consistent way, users may also exchange privately defined data which describes how each private-use character is to be interpreted. The Unicode Standard provides no predefined format for such a data exchange.

#Normalization. The canonical and compatibility decompositions of any private-use character are equal to the character itself (for example, U+E000 decomposes to U+E000). The Canonical_Combining_Class of private-use characters is defined as 0 (Not_Reordered). These values are normatively defined by the Unicode Standard and cannot be changed by private agreement. The treatment of all private-use characters for normalization forms NFC, NFD, NFKD, and NFKC is also normatively defined by the Unicode Standard on the basis of these decompositions. (See Unicode Standard Annex #15, “Unicode Normalization Forms.”) No private agreement may change these forms—for example, by changing the standard canonical or compatibility decompositions for private-use characters. The implication is that all private-use characters, no matter what private agreements they are subject to, always normalize to themselves and are never reordered in any Unicode normalization form.

This does not preclude private agreements on other transformations. Thus one could define a transformation “MyCompanyComposition” that was identical to NFC except that it mapped U+E000 to “a”. The forms NFC, NFD, NFKD, and NFKC themselves, however, cannot be changed by such agreements.

#23.5.1 Private Use Area: U+E000–U+F8FF

The primary Private Use Area consists of code points in the range U+E000 to U+F8FF, for a total of 6,400 private-use characters.

#Encoding Structure. By convention, the primary Private Use Area is divided into a corporate use subarea for platform writers, starting at U+F8FF and extending downward in values, and an end-user subarea, starting at U+E000 and extending upward.

By following this convention, the likelihood of collision between private-use characters defined by platform writers with private-use characters defined by end users can be reduced. However, it should be noted that this is only a convention, not a normative specification. In principle, any user can define any interpretation of any private-use character.

#Corporate Use Subarea. Systems vendors and/or software developers may need to reserve some private-use characters for internal use by their software. The corporate use subarea is the preferred area for such reservations. Assignments of character semantics in this subarea may be completely internal, hidden from end users, and used only for vendor-specific application support, or they may be published as vendor-specific character assignments available to applications and end users. An example of the former case would be the assignment of a character code to a system support operation such as <MOVE> or <COPY>; an example of the latter case would be the assignment of a character code to a vendor-specific logo character such as Apple’s apple character.

Note, however, that systems vendors may need to support full end-user definability for all private-use characters, for such purposes as gaiji support or for transient cross-mapping tables. The use of noncharacters (see Section 23.7, Noncharacters, and Definition D14 in Section 3.4, Characters and Encoding) is the preferred way to make use of non-interchangeable internal system sentinels of various sorts.

#End-User Subarea. The end-user subarea is intended for private-use character definitions by end users or for scratch allocations of character space by end-user applications.

#Allocation of Subareas. Vendors may choose to reserve ranges of private-use characters in the corporate use subarea and make some defined portion of the end-user subarea available for completely free end-user definition. The convention of separating the two subareas is merely a suggestion for the convenience of system vendors and software developers. No firm dividing line between the two subareas is defined in this standard, as different users may have different requirements. No provision is made in the Unicode Standard for avoiding a “stack-heap collision” between the two subareas; in other words, there is no guarantee that end users will not define a private-use character at a code point that overlaps and conflicts with a particular corporate private-use definition at the same code point. Avoiding such overlaps in definition is up to implementations and users.

#23.5.2 Supplementary Private Use Areas

#Encoding Structure. The entire Plane 15, with the exception of the noncharacters U+FFFFE and U+FFFFF, is defined to be the Supplementary Private Use Area-A. The entire Plane 16, with the exception of the noncharacters U+10FFFE and U+10FFFF, is defined to be the Supplementary Private Use Area-B. Together these areas make an additional 131,068 code points available for private use.

The supplementary PUAs provide additional undifferentiated space for private-use characters for implementations for which the 6,400 private-use characters in the primary PUA prove to be insufficient.

#23.6 Surrogates Area

#23.6.1 Surrogates Area: U+D800–U+DFFF

When using UTF-16 to represent supplementary characters, pairs of 16-bit code units are used for each character. These units are called surrogates. To distinguish them from ordinary characters, they are allocated in a separate area. The Surrogates Area consists of 1,024 low-half surrogate code points and 1,024 high-half surrogate code points. For the formal definition of a surrogate pair and the role of surrogate pairs in the Unicode Conformance Clause, see Section 3.8, Surrogates, and Section 5.4, Handling Surrogate Pairs in UTF-16.

The use of surrogate pairs in the Unicode Standard is formally equivalent to the Universal Transformation Format-16 (UTF-16) defined in ISO/IEC 10646. For more information, see Appendix C, Relationship to ISO/IEC 10646. For a complete statement of UTF-16, see Section 3.9, Unicode Encoding Forms.

#High-Surrogate. The high-surrogate code points are assigned to the range U+D800.. U+DBFF. The high-surrogate code point is always the first element of a surrogate pair.

#Low-Surrogate. The low-surrogate code points are assigned to the range U+DC00.. U+DFFF. The low-surrogate code point is always the second element of a surrogate pair.

#Private-Use High-Surrogates. The high-surrogate code points from U+DB80..U+DBFF are private-use high-surrogate code points (a total of 128 code points). Characters represented by means of a surrogate pair, where the high-surrogate code point is a private-use high-surrogate, are private-use characters from the supplementary private use areas. For more information on private-use characters, see Section 23.5, Private-Use Characters.

The code tables do not have charts or name list entries for the range U+D800..U+DFFF because individual, unpaired surrogates merely have code points.

#23.7 Noncharacters

#23.7.1 Noncharacters: U+FFFE, U+FFFF, and Others

Noncharacters are code points that are permanently reserved in the Unicode Standard for internal use. They are not recommended for use in open interchange of Unicode text data. See Section 3.2, Conformance Requirements and Section 3.4, Characters and Encoding, for the formal definition of noncharacters and conformance requirements related to their use.

The Unicode Standard sets aside 66 noncharacter code points. The last two code points of each plane are noncharacters: U+FFFE and U+FFFF on the BMP, U+1FFFE and U+1FFFF on Plane 1, and so on, up to U+10FFFE and U+10FFFF on Plane 16, for a total of 34 code points. In addition, there is a contiguous range of another 32 noncharacter code points in the BMP: U+FDD0..U+FDEF. For historical reasons, the range U+FDD0..U+FDEF is contained within the Arabic Presentation Forms-A block, but those noncharacters are not “Arabic noncharacters” or “right-to-left noncharacters,” and are not distinguished in any other way from the other noncharacters, except in their code point values.

Applications are free to use any of these noncharacter code points internally. They have no standard interpretation when exchanged outside the context of internal use. However, they are not illegal in interchange, nor does their presence cause Unicode text to be ill-formed. The intent of noncharacters is that they are permanently prohibited from being assigned interchangeable meanings by the Unicode Standard. They are not prohibited from occurring in valid Unicode strings which happen to be interchanged. This distinction, which might be seen as too finely drawn, ensures that noncharacters are correctly preserved when “interchanged” internally, as when used in strings in APIs, in other interprocess protocols, or when stored.

If a noncharacter is received in open interchange, an application is not required to interpret it in any way. It is good practice, however, to recognize it as a noncharacter and to take appropriate action, such as replacing it with U+FFFD REPLACEMENT CHARACTER, to indicate the problem in the text. It is not recommended to simply delete noncharacter code points from such text, because of the potential security issues caused by deleting uninterpreted characters. (See conformance clause C7 in Section 3.2, Conformance Requirements, and Unicode Technical Report #36, “Unicode Security Considerations.”)

In effect, noncharacters can be thought of as application-internal private-use code points. Unlike the private-use characters discussed in Section 23.5, Private-Use Characters, which are assigned characters and which are intended for use in open interchange, subject to interpretation by private agreement, noncharacters are permanently reserved (unassigned) and have no interpretation whatsoever outside of their possible application-internal private uses.

#U+FFFF and U+10FFFF. These two noncharacter code points have the attribute of being associated with the largest code unit values for particular Unicode encoding forms. In UTF-16, U+FFFF is associated with the largest 16-bit code unit value, FFFF₁₆. U+10FFFF is associated with the largest legal UTF-32 32-bit code unit value, 10FFFF₁₆. This attribute renders these two noncharacter code points useful for internal purposes as sentinels. For example, they might be used to indicate the end of a list, to represent a value in an index guaranteed to be higher than any valid character value, and so on.

#U+FFFE. This noncharacter has the intended peculiarity that, when represented in UTF-16 and then serialized, it has the opposite byte sequence of U+FEFF, the byte order mark. This means that applications should reserve U+FFFE as an internal signal that a UTF-16 text stream is in a reversed byte format. Detection of U+FFFE at the start of an input stream should be taken as a strong indication that the input stream should be byte-swapped before interpretation. For more on the use of the byte order mark and its interaction with the noncharacter U+FFFE, see Section 23.8, Specials.

#23.8 Specials

The Specials block contains code points that are interpreted as neither control nor graphic characters but that are provided to facilitate current software practices.

For information about the noncharacter code points U+FFFE and U+FFFF, see Section 23.7, Noncharacters.

#23.8.1 Byte Order Mark (BOM): U+FEFF

For historical reasons, the character U+FEFF used for the byte order mark is named ZERO WIDTH NO-BREAK SPACE. Except for compatibility with versions of Unicode prior to Version 3.2, U+FEFF is not used with the semantics of zero width no-break space (see Section 23.2, Layout Controls). Instead, its most common and most important usage is in the following two circumstances:

1. Unmarked Byte Order. Some machine architectures use the so-called big-endian byte order, while others use the little-endian byte order. When Unicode text is serialized into bytes, the bytes can go in either order, depending on the architecture. Sometimes this byte order is not externally marked, which causes problems in interchange between different systems.

2. Unmarked Character Set. In some circumstances, the character set information for a stream of coded characters (such as a file) is not available. The only information available is that the stream contains text, but the precise character set is not known.

In these two cases, the character U+FEFF is used as a signature to indicate the byte order and the character set by using the byte serializations described in Section 3.10, Unicode Encoding Schemes. Because the byte-swapped version U+FFFE is a noncharacter, when an interpreting process finds U+FFFE as the first character, it signals either that the process has encountered text that is of the incorrect byte order or that the file is not valid Unicode text.

In the UTF-16 encoding scheme, U+FEFF at the very beginning of a file or stream explicitly signals the byte order.

The byte sequences <FE₁₆ FF₁₆> or <FF₁₆ FE₁₆> may also serve as a signature to identify a file as containing UTF-16 text. Either sequence is exceedingly rare at the outset of text files using other character encodings, whether single- or multiple-byte, and therefore not likely to be confused with real text data. For example, in systems that employ ISO Latin-1 (ISO/IEC 8859-1) or the Microsoft Windows ANSI Code Page 1252, the byte sequence <FE₁₆ FF₁₆> constitutes the string <thorn, y diaeresis> “þÿ”; in systems that employ the Apple Macintosh Roman character set or the Adobe Standard Encoding, this sequence represents the sequence <ogonek, hacek> “˛ˇ”; in systems that employ other common IBM PC code pages (for example, CP 437, 850), this sequence represents <black square, no-break space> “■ ”.

In UTF-8, the BOM corresponds to the byte sequence <EF₁₆ BB₁₆ BF₁₆>. Although there are never any questions of byte order with UTF-8 text, this sequence can serve as signature for UTF-8 encoded text where the character set is unmarked. As with a BOM in UTF-16, this sequence of bytes will be extremely rare at the beginning of text files in other character encodings. For example, in systems that employ Microsoft Windows ANSI Code Page 1252, <EF₁₆ BB₁₆ BF₁₆> corresponds to the sequence <i diaeresis, guillemet, inverted question mark> “ï » ¿”.

For compatibility with versions of the Unicode Standard prior to Version 3.2, the code point U+FEFF has the word-joining semantics of zero width no-break space when it is not used as a BOM. In new text, these semantics should be encoded by U+2060 WORD JOINER. See “Line and Word Breaking” in Section 23.2, Layout Controls, for more information.

Where the byte order is explicitly specified, such as in UTF-16BE or UTF-16LE, then all U+FEFF characters—even at the very beginning of the text—are to be interpreted as zero width no-break spaces. Similarly, where Unicode text has known byte order, initial U+FEFF characters are not required, but for backward compatibility are to be interpreted as zero width no-break spaces. For example, for strings in an API, the memory architecture of the processor provides the explicit byte order. For databases and similar structures, it is much more efficient and robust to use a uniform byte order for the same field (if not the entire database), thereby avoiding use of the byte order mark.

Systems that use the byte order mark must recognize when an initial U+FEFF signals the byte order. In those cases, it is not part of the textual content and should be removed before processing, because otherwise it may be mistaken for a legitimate zero width no-break space. To represent an initial U+FEFF ZERO WIDTH NO-BREAK SPACE in a UTF-16 file, use U+FEFF twice in a row. The first one is a byte order mark; the second one is the initial zero width no-break space. See Table 23-6 for a summary of encoding scheme signatures.

Encoding Scheme	Signature
UTF-8	EF BB BF
UTF-16 Big-endian	FE FF
UTF-16 Little-endian	FF FE
UTF-32 Big-endian	00 00 FE FF
UTF-32 Little-endian	FF FE 00 00

If U+FEFF had only the semantics of a signature code point, it could be freely deleted from text without affecting the interpretation of the rest of the text. Carelessly appending files together, for example, can result in a signature code point in the middle of text. Unfortunately, U+FEFF also has significance as a character. As a zero width no-break space, it indicates that line breaks are not allowed between the adjoining characters. Thus U+FEFF affects the interpretation of text and cannot be freely deleted. The overloading of semantics for this code point has caused problems for programs and protocols. The new character U+2060 WORD JOINER has the same semantics in all cases as U+FEFF, except that it cannot be used as a signature. Implementers are strongly encouraged to use word joiner in those circumstances whenever word joining semantics are intended.

#U+FEFF as Signature in Other Charsets. An initial U+FEFF also takes a characteristic form in other charsets designed for Unicode text. (The term “charset” refers to a wide range of text encodings, including encoding schemes as well as compression schemes and text-specific transformation formats.) The characteristic sequences of bytes associated with an initial U+FEFF can serve as signatures in those cases, as shown in Table 23-7.

Charset	Signature
SCSU	0E FE FF
BOCU-1	FB EE 28
UTF-7	2B 2F 76 38 or 2B 2F 76 39 or 2B 2F 76 2B or 2B 2F 76 2F
UTF-EBCDIC	DD 73 66 73

Most signatures can be deleted either before or after conversion of an input stream into a Unicode encoding form. However, in the case of BOCU-1 and UTF-7, the input byte sequence must be converted before the initial U+FEFF can be deleted, because stripping the signature byte sequence without conversion destroys context necessary for the correct interpretation of subsequent bytes in the input sequence.

#Guidelines for Use of a BOM in UTF-8. The UTF-8 encoding scheme permits, but does not require, a BOM to be present. This raises the question of when a BOM should or should not be generated or expected when producing or consuming UTF-8 encoded text.

The utility of a BOM in UTF-8 is limited to scenarios in which a byte sequence contains text that may or may not be encoded as UTF-8. In such scenarios, a BOM may be useful to differentiate text encoded in one of a few possible character encodings. However, the presence of a BOM may also complicate text processing.

• Some text processing tools fail to handle BOMs correctly. This is especially true for programs that were historically encoding agnostic and for ad hoc programs written for one-time use purposes.
• A text processing tool must maintain additional state in order to recognize if an observed U+FEFF character is a BOM or whether it should be treated as a zero width no-break space. Such state may or may not be intrinsic to the structure of the program.
• A text generating tool may be required to generate a BOM if the first character to be encoded is U+FEFF and that character is not intended to be used as a BOM. This is only required for compatibility with Unicode versions prior to 3.2; U+2060 WORD JOINER should be used in place of U+FEFF for such purposes with more recent Unicode versions.
• Concatenation of text containing a BOM requires care. When concatenating to an empty text, preservation of a BOM may be warranted, but otherwise, failure to elide the BOM will result in the insertion of a U+FEFF character that becomes part of the concatenated textual content.
• In situations where text is known to be encoded as UTF-8, a BOM consumes storage space unnecessarily. While this is unlikely to be a concern for a single document, it may be a significant concern in situations involving thousands or millions of small text sources.

The following guidelines advise alternative approaches tailored for a few distinct audiences. Except where otherwise noted, these approaches preclude the possibility of a text starting with a U+FEFF character that is not intended as a BOM under the expectation that such text is exceedingly rare and most likely due to a failure to elide a BOM. Text authored for Unicode 3.2 or later should use U+2060 WORD JOINER instead.

Software developers:

• If consuming UTF-8, recognize and discard a BOM.
• If producing UTF-8, include a BOM only if explicitly directed to do so, or if a BOM is known to be required by a protocol.

Text authors:

• Do not use U+FEFF to function as a zero width no-break space character; use U+2060 WORD JOINER instead.
• Include a BOM if one is known to be required by a targeted protocol.
• Otherwise, include a BOM when authoring a UTF-8 text file that contains non-ASCII characters, is not targeting a specific protocol, but which may be opened by applications that will not assume UTF-8 by default. (This is useful on systems like Microsoft Windows where some applications assume text files to be encoded with the Active Code Page.)
• Otherwise, do not include a BOM.

#23.8.2 Specials: U+FFF0–U+FFFF

The nine unassigned Unicode code points in the range U+FFF0..U+FFF8 are reserved for special character definitions.

#23.8.3 Annotation Characters: U+FFF9–U+FFFB

An interlinear annotation consists of annotating text that is related to a sequence of annotated characters. For all regular editing and text-processing algorithms, the annotated characters are treated as part of the text stream. The annotating text is also part of the content, but for all or some text processing, it does not form part of the main text stream. However, within the annotating text, characters are accessible to the same kind of layout, text-processing, and editing algorithms as the base text. The annotation characters delimit the annotating and the annotated text, and identify them as part of an annotation. See Figure 23-4.

The annotation characters are used in internal processing when out-of-band information is associated with a character stream, very similarly to the usage of U+FFFC OBJECT REPLACEMENT CHARACTER. However, unlike the opaque objects hidden by the latter character, the annotation itself is textual.

#Conformance. A conformant implementation that supports annotation characters interprets the base text as if it were part of an unannotated text stream. Within the annotating text, it interprets the annotating characters with their regular Unicode semantics.

U+FFF9 INTERLINEAR ANNOTATION ANCHOR is an anchor character, preceding the interlinear annotation. The exact nature and formatting of the annotation depend on additional information that is not part of the plain text stream. This situation is analogous to that for U+FFFC OBJECT REPLACEMENT CHARACTER.

U+FFFA INTERLINEAR ANNOTATION SEPARATOR separates the base characters in the text stream from the annotation characters that follow. The exact interpretation of this character depends on the nature of the annotation. More than one separator may be present. Additional separators delimit parts of a multipart annotating text.

U+FFFB INTERLINEAR ANNOTATION TERMINATOR terminates the annotation object (and returns to the regular text stream).

#Use in Plain Text. Usage of the annotation characters in plain text interchange is strongly discouraged without prior agreement between the sender and the receiver, because the content may be misinterpreted otherwise. Simply filtering out the annotation characters on input will produce an unreadable result or, even worse, an opposite meaning. On input, a plain text receiver should either preserve all characters or remove the interlinear annotation characters as well as the annotating text included between the INTERLINEAR ANNOTATION SEPARATOR and the INTERLINEAR ANNOTATION TERMINATOR.

When an output for plain text usage is desired but the receiver is unknown to the sender, these interlinear annotation characters should be removed as well as the annotating text included between the INTERLINEAR ANNOTATION SEPARATOR and the INTERLINEAR ANNOTATION TERMINATOR.

This restriction does not preclude the use of annotation characters in plain text interchange, but it requires a prior agreement between the sender and the receiver for correct interpretation of the annotations.

#Lexical Restrictions. If an implementation encounters a paragraph break between an anchor and its corresponding terminator, it shall terminate any open annotations at this point. Anchorcharacters must precede their corresponding terminator characters. Unpaired anchors or terminators shall be ignored. A separator occurring outside a pair of delimiters, shall be ignored. Annotations may be nested.

#Formatting. All formatting information for an annotation is provided by higher-level protocols. The details of the layout of the annotation are implementation-defined. Correct formatting may require additional information that is not present in the character stream, but rather is maintained out-of-band. Therefore, annotation markers serve as placeholders for an implementation that has access to that information from another source. The formatting of annotations and other special line layout features of Japanese is discussed in JIS X 4051.

#Input. Annotation characters are not normally input or edited directly by end users. Their insertion and management in text are typically handled by an application, which will present a user interface for selecting and annotating text.

#Collation. With the exception of the special case where the annotation is intended to be used as a sort key, annotations are typically ignored for collation or optionally preprocessed to act as tie breakers only. Importantly, annotation base characters are not ignored, but rather are treated like regular text.

#Bidirectional Text. Bidirectional processing of text containing interlinear annotations requires special care. This follows from the fact that interlinear annotations are fundamentally nonlinear—the annotations are not part of the main text flow, whereas bidirectional text processing assumes that it is applied to a single, linear text flow. For best results, the Bidirectional Algorithm should be applied to the main text, in which any interlinear annotations are replaced by their annotated text, in each case bracketed by bidirectional format control characters to ensure that the annotated text remains visually contiguous, and then should be separately applied to each extracted segment of annotating text. (See Unicode Standard Annex #9, “Unicode Bidirectional Algorithm,” for more information.)

#23.8.4 Replacement Characters: U+FFFC–U+FFFD

#U+FFFC. The U+FFFC OBJECT REPLACEMENT CHARACTER is used as an insertion point for objects located within a stream of text. All other information about the object is kept outside the character data stream. Internally it is a dummy character that acts as an anchor point for the object’s formatting information. In addition to assuring correct placement of an object in a data stream, the object replacement character allows the use of general stream-based algorithms for any textual aspects of embedded objects.

#U+FFFD. The U+FFFD REPLACEMENT CHARACTER is the general substitute character in the Unicode Standard. It can be substituted for any “unknown” character in another encoding that cannot be mapped in terms of known Unicode characters. It can also be used as one means of indicating a conversion error, when encountering an ill-formed sequence in a conversion between Unicode encoding forms. See Section 3.9, Unicode Encoding Forms for detailed recommendations on the use of U+FFFD as replacement for ill-formed sequences. See also Section 5.3, Unknown and Missing Characters for related topics.

#23.9 Tag Characters

#23.9.1 Tag Characters: U+E0000–U+E007F

This block encodes a set of 97 special-use tag characters to enable the construction of tags using characters that can be strictly separated from ordinary text content characters in Unicode, but that correspond to ASCII-based strings. They can be identified and/or ignored by implementations with trivial algorithms because there is no overloading of usage for these tag characters—they can express only tag values and never textual content itself. One of these 97 characters is the deprecated language tag identification character, U+E0001 LANGUAGE TAG.

The current conformant use of the undeprecated 96 tag characters is specified in Unicode Technical Standard #51, “Unicode Emoji.” See ED-14a. emoji tag sequence (ETS) and Annex C, Valid Emoji Tag Sequences in that specification. UTS #51 does not use nested tag sequences, and the tag sequences it does use are not stateful.

#23.9.2 Deprecated Use for Language Tagging

The tag characters were originally intended for language tagging of plain text, as an alternative to using malformed UTF-8 for language tagging. When it became clear that language tagging using these characters was complicated and unnecessary, the tag characters were deprecated as of Version 5.1. In Version 8.0, all but U+E0001 LANGUAGE TAG and U+E007F CANCEL TAG were un-deprecated, in anticipation of other use for the tag characters. In Version 9.0, U+E007F CANCEL TAG was also un-deprecated, and the tag characters were repurposed for use in emoji tag sequences.

For the full specification of the original use of tag characters, see Section 16.9, Tag Characters in Version 5.0 of The Unicode Standard:

https://www.unicode.org/versions/Unicode5.0.0/ch16.pdf#G17521