Strings on the Web: Language and Direction Metadata

A string can be created in a number of ways, including a content author typing strings into a plain text editor, text message, or editing tool; or a script scraping text from web pages; or acquisition of an existing set of strings from another application or repository. In the data formats under consideration in this document, many strings come from back end data repositories or databases of various kinds. Sources of strings may provide an interface, API, or metadata that includes information about the string direction and language of the data. Some also provide a suitable default for when the direction or language is not provided or specified. In this document, the producer of a string is the source, be it a human or a mechanism, that creates or provides a string for storage or transmission.

When a string is created, it's necessary to (a) detect or capture the appropriate language and string direction to be associated with the string, and (b) take steps, where needed, to set the string up in a way that stores and communicates the language and string direction.

For example, in the case of a string that is extracted from an HTML form, the string direction can be detected from the computed value of the form's field. Such a value could be inherited from an earlier element, such as the html element, or set using markup or styling on the input element itself. The user could also set the direction of the text by using keyboard shortcut keys to change the direction of the form field. The dirname attribute provides a way of automatically communicating that value with a form submission.

Similarly, language information in an HTML form would typically be inherited from the lang attribute on the html tag, or an ancestor element in the tree with a lang attribute.

If the producer of the string is receiving the string from a location where it was stored by another producer, and where the string direction and language has already been established, the producer needs to understand that the language and string direction has already been set, and understand how to convert or encode that information for its consumers.

A consumer is an application or process that receives a string for processing and possibly places it into a context where it will be exposed to a user. For display purposes, it must ensure that the block direction and language of the string is correctly applied to the string in that context. For processing purposes, it must at least persist the language and direction and may need to use the language and direction data in order to perform language-specific operations.

Proper display of the string involves supplying the string direction and language to the rendering document or process by applying additional markup, adding control codes, or setting display properties. This indicates to rendering software the string direction or language that should be applied to the string in this display context to get the string to appear correctly. For both language and direction, it must make clear the boundaries for the range of text to which the language applies. For text direction, it must also isolate embedded strings from the surrounding text to avoid spill-over effects of the bidi algorithm [UAX9].

Note that a consumer of one document format might be a producer of another document format.

Between any producer and consumer, there needs to be an agreement about what the document format contains and what the data in each field or attribute means. Any time a producer of a string takes special steps to collect and communicate information about the string direction or language of that string, it must do so with the expectation that the consumer of the string will understand how the producer encoded this information.

If no action is taken by the producer, the consumer must still decide what rules to follow in order to decide on the appropriate string direction and language, even if it is only to provide some form of default value.

In some systems or document formats, the necessary behaviour of the producers and consumers of a string are fully specified. In others, such agreements are not available; it is up to users to provide an agreement for how to encode, transmit, and later decode the necessary language or direction information. Low level specifications, such as JSON, do not provide a string metadata structure by default, so any document formats based on these need to provide the "agreement" themselves.

Avoid assigning or requiring language or direction metadata for non-linguistic fields (that is, strings that contain data that is not human language). Note that this includes application-internal data values [INTERNATIONAL-SPECS].

While the value of a syntactic content item or user-supplied value will often use a word-like token that conveys meaning to humans (as an aid in debugging, for example), the values need to consistently be wrapped with localizable display strings for presentation to the user.

For localizable text fields that appear in a single language, use a data structure to represent the value. The recommended representation is an object with three fields. The field value contains the actual string. The field lang contains a valid [BCP47] language tag. The field dir contains the string's string direction (one of the values ltr, rtl, or auto).

Use of heuristics to determine language or string direction will always fail for certain cases, and there needs to be a way to provide the correct outcome for those strings. Assignment of metadata (either as a resource-wide default, or in a string-specific label) is an intentional act that removes the need to guess the outcome by applying heuristics.

The use of metadata for indicating block direction is preferred because it avoids requiring the consumer to interpolate the direction using methods such as first strong or use of methods which require modification of the data itself (such as the insertion of RLM/LRM markers or bidirectional controls).

This combines both language and direction metadata and, if consistently adopted, makes interchange between different formats easier. Consistency between different specifications and document formats allows for the easy interchange of string data. By naming field attributes in the same way and adopting the same semantics, different specifications can more easily extract values from or add values into resources from other data sources.

When a resource contains a number of natural language strings (and particularly if those string are all in the same language), using the localized string representation described above can become inefficient. To reduce the complexity of encoding these strings, specifications can establish a resource-level default for language and string direction. These are separate values, as language does not imply direction. There should still be the ability to override either language or direction on any given string value by using the representation found above.

If your specification defines its own document level defaults, provide two optional fields:

Exceptions to the default are always a possibility, so it needs to be possible for users to override the default on a string-by-string basis.

First-strong heuristics are not applied to strings when the direction has been set externally using metadata. Even if a strongly directional character, such as U+200F RIGHT-TO-LEFT MARK, has been prepended to a string, resource-wide default metadata can override the presentation of the string in ways that result in spillover effects. Therefore content needs to be able to provide string-level metadata to override the default for strings whose string direction does not match the resource-wide default.

For specifications that can make use of the [JSON-LD] @context mechanism, use the @language and @direction fields to supply the document level defaults.

The world is not monolingual. Having documents that contain only a single language would mean providing many iterations of the document, one for each language, in order to localize the content. This also might require language negotiation when requesting the content.

One way to address this is to allow multilingual values for each localizable text field inside the document.

Language selection is not merely the exact matching of language tag string values to the user's preferred locale. The usual object representation of a localizable text field requires that the object be deserialized in order to discover the language tag associated with the value. This can be inefficient when there are many values in a given file. In these cases, the best practice is to use a language map to organize localizable text values. Such a map exposes the language tag for the purposes of selection, but still uses an object representation on the value side of the map, since both language and direction might need to be overridden for a given string value.

A producer doesn't need to do anything.

The string is stored as it is.

Consumers must look for the first character in the string with a strong Unicode directional property, and set the string direction to match it. They then take appropriate action to ensure that the string will be displayed as needed. This is not quite so simple as it may appear, for the following reasons:

1. Characters at the start of a string without a strong direction (eg. punctuation, numbers, etc) and isolated sequences (ie. sequences of characters surrounded by RLI/LRI/FSI...PDI formatting characters) within a string must be skipped in order to find the first strong character.
2. The detection algorithm needs to be able to handle markup at the start of the string. It needs to be able to tell whether the markup is just string text, or whether the markup needs to be parsed in the target location – in which case it must understand the markup, and understand any direction-related information that is carried in the markup.

First-strong detection is only needed where the required string direction is not already known. If direction is indicated for a string by metadata, either string-specific or via a resource-wide declaration, then first-strong heuristics should not be invoked. For example, first-strong heuristics would produce the wrong result for a string such as "HTML و CSS: تصميم و إنشاء مواقع الويب". This can be corrected using metadata, the use of which signifies informed intention, and you would not need or want to apply heuristics that would then make the result incorrect.

However, if there is no mechanism for the application of metadata, or if there is such a mechanism but the content developer omitted to use it, then first-strong heuristics can be helpful to establish base direction in many, though not all, cases. The application of strongly-directional formatting characters can help produce correct results for plain text strings such as the example just quoted, but it is not always possible to apply those (see 4.3 Augmenting first-strong by inserting RLM/LRM markers).

Where it is reliable, information about direction can be obtained without any changes to the string, and without the agreements and structures that would be needed to support out-of-band metadata.

The main problem with this approach is that it produces the wrong result for

1. strings that begin with a strong character with a different directionality than that needed for the string overall (eg. an Arabic tweet that starts with a hashtag)
2. strings that don't have a strong directional character (such as a telephone number), which are likely to be displayed incorrectly in a RTL context.
3. strings that begin with markup, such as span, since the first strong character is always going to be LTR.

In cases where the entire string starts and ends with RLI/LRI/FSI...PDI formatting characters, it is not possible to detect the first strong character by following the Unicode Bidirectional Algorithm. This is because the algorithm requires that bidi-isolated text be excluded from the detection.

If no strong directional character is found in the string, the direction should probably be assumed to be LTR, and the consumer should act on that basis. This has not been tested fully, however.

If a string contains markup that will be parsed by the consumer as markup, there are additional problems. Any such markup at the start of the string must also be skipped when searching for the first strong directional character.

If parseable markup in the string contains information about the intended direction of the string (for example, a dir attribute with the value rtl in HTML), that information should be used rather than relying on first-strong heuristics. This is problematic in a couple of ways: (a) it assumes that the consumer of the string understands the semantics of the markup, which may be ok if there is an agreement between all parties to use, say, HTML markup only, but would be problematic, for example, when dealing with random XML vocabularies, and (b) the consumer must be able to recognise and handle a situation where only the initial part of the string has markup, ie. the markup applies to an inline span of text rather than the string as a whole.

If, however, there is angle bracket content that is intended to be an example of markup, rather than actual markup, the markup must not be skipped – trying to display markup source code in a RTL context yields very confusing results! It isn't clear, however, how a consumer of the string would always know the difference between examples and parseable strings.

Although first-strong detection is outlined in the Unicode Bidirectional Algorithm (UBA) [UAX9], it is not the only possible higher-level protocol mentioned for estimating string direction. For example, X (formerly known as Twitter) and Facebook currently use different default heuristics for guessing the base direction of text — neither use just simple first-strong detection, and one uses a completely different method.

A producer ascertains the string direction of the string and adds that to a metadata field that accompanies the string when it is stored or transmitted.

There are several approaches to using metadata:

1. Label every string with a string direction.
2. Provide a document-level default for block direction and only include metadata for strings whose value is different. The value auto is used when the direction of a string is not known.
3. Rely on the consumer to do first-strong detection, and label only those strings which would produce the wrong result (that is, a right-to-left string that starts with left-to-right strong characters).

If storing or transmitting a set of strings at a time, it helps to have a field for the resource as a whole that sets a global, default string direction which can be inherited by all strings in the resource. Note that in addition to a global field, you still need the possibility of attaching string-specific metadata fields in cases where a string's string direction is not the same as the default value. The string direction set on an individual string must always override the default.

Consumers would need to understand how to read the metadata sent with a string, and would need to apply first-strong heuristics in the absence of metadata.

The use of the Localizable dictionary structure is RECOMMENDED for individual values in JSON-based document formats, as it combines both language and direction metadata and, if consistently adopted, makes interchange between different formats easier.

Passing metadata as separate data value from the string provides a simple, effective and efficient method of communicating the intended string direction without affecting the actual content of the string.

If every string is labelled for direction, or the direction for all strings can be ascertained by applying the global setting and any string-specific deviations, it avoids the need to inspect and run heuristics to determine each separate string's string direction.

Out-of-band information needs to be associated with and kept with strings. This may be problematic for some sets of string data which are not part of a defined framework.

In particular, JSON-LD doesn't allow direction to be associated with individual strings in the same way as it works for language.

A producer ascertains the string direction of the string and adds an marker character (either U+200F RIGHT-TO-LEFT MARK (RLM) or U+200E LEFT-TO-RIGHT MARK (LRM)) to the beginning of the string. The marker is not functional, ie. it will not automatically apply a base direction to the string that can be used by the consumer, it is simply a marker.

There are a number of possible approaches:

1. Add a marker to every string (not recommended).
2. Rely on the consumer to do first-strong detection, and add a marker to only those strings which would produce the wrong result (eg. a RTL string that starts with LTR strong characters).
3. Assume a default of LTR (no marker), and apply only RLM markers.

Consumers apply first-strong heuristics to detect the string direction for the string. The RLM and LRM characters are strongly typed directionally, and should therefore result in detecting the appropriate base direction.

As described in 4.1 First-strong property detection, this approach is not relevant if directional information is provided via metadata.

It provides a reliable way of indicating base direction, as long as the producer can reliably apply markers.

In theory, it should be easier to spot the first-strong character in strings that begin with markup, as long as the correct RLM/LRM is prepended to the string.

If the producer is a human, they could theoretically apply one of these characters when creating a string in order to signal the directionality.

A significant problem with this, especially on mobile devices, is the availability or inconvenience of inputting an RLM/LRM character. The keyboards of mobile devices generally do not provide keys for RLM/LRM characters. Perhaps more important, because the characters are invisible and because Unicode bidi is complicated, it can be difficult for the user to know how to use the character effectively. In fact, a large percentage of users don't actually know what these characters are or what they do.

Furthermore, if a person types information into, say, an HTML form in a RTL page or uses shortcut keys to set the direction for the form field, the strings will look correct without the need to add RLM/LRM. However, used outside of that context, the string would look incorrect unless it is associated with information about the required block direction. Similarly, strings scraped from a web page that has dir=rtl set in the html element would not normally have or need an RLM/LRM character at the start of the string in HTML.

It may be possible for the steps used by a producer to include an examination of the original context of the string for directional information (for example, by testing the computed direction of an HTML form field), followed by automatic insertion of an RLM/LRM mark into the beginning of the string where necessary. An issue with this approach is that it changes the string value and identity. This may also create problems for working with string length or pointer positions, especially if some producers add markers and others don't.

If directional information is contained in markup that will be parsed as such by the consumer (for example, dir=rtl in HTML), the producer of the string needs to understand that markup in order to set or not set an RLM/LRM character as appropriate. If the producer always adds RLM/LRM to the start of such strings, the consumer is expected to know that. If the producer relies instead on the markup being understood, the consumer is expected to understand the markup.

The producer of a string should not automatically apply RLM or LRM to the start of the string, but should test whether it is needed. For example, if there's already an RLM in the text, there is no need to add another. If the context is correctly conveyed by first-strong heuristics, there is no need to add additional characters either. Note, however, that testing whether supplementary directional information of this kind is needed is only possible if the producer has access, and knows that it has access, to the original context of the string. Many document formats are generated from data stored away from the original context. For example, the catalog of books in the original example above is disconnected from the user inputing the bidirectional text.

A producer ascertains the string direction of the string and adds a directional formatting character (one of U+2066 LEFT-TO-RIGHT ISOLATE (LRI), U+2067 RIGHT-TO-LEFT ISOLATE (RLI), U+2068 FIRST STRONG ISOLATE (FSI), U+202A LEFT-TO-RIGHT EMBEDDING (LRE), or U+202B RIGHT-TO-LEFT EMBEDDING (RLE)) to the beginning of the string, and U+2069 POP DIRECTIONAL ISOLATE (PDI) or U+202C POP DIRECTIONAL FORMATTING (PDF) to the end.

There are a number of possible approaches:

1. Add the formatting codes to every string.
2. Rely on the consumer to do first-strong detection, and add a marker to only those strings which would produce the wrong result (eg. a RTL string that starts with LTR strong characters).

Consumers would theoretically just insert the string in the place it will be displayed, and rely on the formatting codes to manage directionality. However, things are not quite so simple (see below).

There are two types of paired formatting characters. The original set of controls provide the ability to add an additional level of bidirectional "embedding" to the Unicode bidirectional Algorithm. More recently, Unicode added a complementary set of "isolating" controls. Isolating controls are used to surround a string. The inside of the string is treated as its own bidirectional sequence, and the string is protected against spill-over effects related to any surrounding text. The enclosing string treats the entire surrounded string as a single unit that is ignored for bidi reordering. This issue is described here.

If paired formatting characters are used, they should be isolating, ie. starting with RLI, LRI, FSI, and not with RLE or LRE.

There are no real advantages to using this approach.

This approach is only appropriate if it is acceptable to change the value of the string. In addition to possible issues such as changed string length or pointer positions, this approach runs a real and serious risk of one of the paired characters getting lost, either through handling errors, or through text truncation, etc.

A producer and a consumer of a string would need to recognise and handle a situation where a string begins with a paired formatting character but doesn't end with it because the formatting characters only describe a part of the string.

Unicode specifies a limit to the number of embeddings that are effective, and embeddings could build up over time to exceed that limit.

Consuming applications would need to recognise and appropriately handle the isolating formatting characters. At the moment such support for RLI/LRI/FSI is far from pervasive.

This approach would disqualify the string from being amenable to UBA first-strong heuristics if used by a non-aware consumer, because the Unicode bidi algorithm is unable to ascertain the base direction for a string that starts with RLI/LRI/FSI and ends with PDI. This is because the algorithm skips over isolated sequences and treats them as a neutral character. A consumer of the string would have to take special steps in such a case to locate the first-strong character.

A producer supplies language metadata for strings, specifying, where necessary, the script in use.

There are a number of possible approaches:

1. Label every string for language, including a script subtag as needed. Consumers may need to compute the script subtag when the producer does not provide one.
2. It might be reasonable to assume a default of LTR for all strings unless marked with a language tag whose script subtag (either present or implied) indicates RTL.
3. Alternatively, limit the use of script subtag metadata to situations where first-strong heuristics are expected to fail — provided that such cases can be identified, and appropriate action taken by the producer (not always reliable). Consumers would then need to use first-strong heuristics in the absence of a script subtag in order to identify the appropriate string direction. The use of script subtags should not, however, be restricted to strings that need to indicate direction; it is perfectly valid to associate a script subtag with any string.
4. Set a default language for a set of strings at a higher level, but provide a mechanism to override that default for a given string where needed.

Consumers extract the script subtag from the language tag associated with each string, computing the string's string direction as necessary. Script subtags associated with RTL scripts are used to assign a direction of RTL to their associated strings.

Language information MUST use [BCP47] language tags. The portion of the language tag that carries the information is the script subtag, not the primary language subtag. For example, Azeri may be written LTR (with the Latin or Cyrillic scripts) or RTL (with the Arabic script). Thus, the subtag az is insufficient to clarify intended block direction. A language tag such as az-Arab (Azeri as written in the Arabic script), however, can generally be relied upon to indicate that the block direction should be RTL.

There is no need to inspect or change the string itself.

This approach avoids the issues associated with first-strong detection when the first-strong character is not indicative of the necessary string direction for the string, and avoids issues relating to the interpretation of markup.

Note that a string that begins with markup that sets a language for the string text content (eg. <cite lang="zh-Hans">) is not problematic here, since that language declaration is not expected to play into the setting of the string direction.

The use of metadata as outlined above is a much better approach, if it is available. This script-related approach is only for use where that approach is unavailable, for legacy reasons.

There are many strings which are not language-specific but which absolutely need to be associated with a particular block direction for correct consumption. For example, MAC addresses inserted into a RTL context need to be displayed with a LTR overall base direction and also be isolated from the surrounding text. It's not clear how to distinguish these cases from others (in a way that would be feasible when using direction metadata). Special language tags, such as zxx (Non-Linguistic), exist for identifying this type of content, but usually data fields of this type omit language information altogether, since it is not applicable.

The list of script subtags may be added to in future. In that case, any subtags that indicate a default RTL direction need to be added to the lists used by the consumers of the strings.

There are some rare situations where the appropriate paragraph direction cannot be identified from the script subtag, but these are really limited to archaic usage of text. For example, Japanese and Chinese text prior to World War 2 was often written RTL, rather than LTR. Languages such as those written using Egyptian Hieroglyphs, or the Tifinagh Berber script, could formerly be written either LTR or RTL, however the default for scholastic research tends to LTR.

The approach outlined here is only appropriate when declaring information about the overall string direction to be associated with a string. We do not recommend use of language data to indicate text direction within strings, since the usage patterns are not interchangeable.

The producer ensures that all strings begin and end with markup which indicates the appropriate base direction for that string. This requires the producer to examine the string. If the string is not bounded by markup with directional information, the producer must add wrap the string with elements that have the dir or its:direction [ITS20] attributes, or other markup appropriate to a given XML application. If the string is bounded by markup, but it is something such as an HTML h1 element, the producer needs to introduce directional information into the existing markup, rather than simply surround the string with a span.

This example uses HTML markup. (Simply to make the example easier to read, it shows the text content of the string as it should be displayed, rather than in the order in which the characters are stored.)

The consumer then relies on the markup to set the base direction around the text content of the string when it is displayed. (Note that, unless additional metadata is provided, the consumer cannot remove the markup before integrating the string in the target location, because it cannot tell what markup has been added by the producer and what was already there. In general, however, such added markup is harmless.)

The benefit for content that already uses markup is clear. The content will already provide complete markup necessary for the display and processing of the text or it can be extracted from the source page context. HTML and XML processors already know how to deal with this markup and provide ready validation.

For HTML, the dir attribute bidirectionally isolates the content from the surrounding text, which removes spillover conflicts. This reduces the work of the consumer.

Markup can also be used for string-internal directional information, something string direction on its own cannot solve.

This is similar to the idea of sending metadata with a string as discussed previously, however the metadata is not stored in a completely separate field (as in 4.2 Metadata), or inserted into the string itself (as in 4.3 Augmenting first-strong by inserting RLM/LRM markers), but is associated with the string as part of the string's serialization format.

Some datatypes, such as [RDF-PLAIN-LITERAL], already exist that allow for language metadata to be serialized as part of a string value. However, these do not include a consideration for direction. This might be addressed by defining a new datatype (or extending an existing one) that document formats could then use to serialize natural language strings that includes both language and direction metadata.

[JSON-LD] 1.1. added the i18n Namespace to permit JSON documents to serialize language and direction metadata directly with a string value. It provides a deserialization to RDF for specifications that need it.

Note that the last string does not include language information because it is an internal data value, but does include direction information because strings of this kind must be presented in the LTR order.

A producer would need to attach the string direction to each string as needed.

Each consumer should use first-strong heuristics for those strings that don't use this approach or do not contain string direction. The producer would then only add string direction information if the first-strong approach would otherwise produce the wrong result. This might simplify the management of strings and the amount of data to be transmittted, because the number of strings requiring metadata is relatively small.

The consumer would look to see whether the string has metadata associated with it, in which case it would set the indicated string direction. Otherwise, it would use first-strong heuristics to determine the string direction of the string.

If a new datatype were added to JSON to support natural language strings, then specifications could easily specify that type for use in document formats. Since the format is standardized, producers and consumers would not need to guess about direction or language information when it is encoded.

Apart from the fact that this currently doesn't work, the downside of adding a datatype is that JSON is a widely implemented format, including many ad-hoc implementations. Any new serialization form would likely break or cause interoperability problems with these existing implementations. JSON is not designed to be a "versioned" format. Any serialization form used would need to be transparent to existing JSON processors and thus could introduce unwanted data or data corruption to existing strings and formats.

A producer ascertains the language of the string (generally from metadata supplied upstream) and includes this information a metadata field that accompanies the string when it is stored or transmitted.

When storing or transmitting a set of strings at a time, it helps to have a field for the resource as a whole that sets a language which can be inherited by all strings in the resource. Note that in addition to a global field, you still need the possibility of attaching string-specific metadata fields in cases where a string's language is not that of the default. The language set on an individual string must override any resource-level value.

A consumer needs to understand how to read the metadata associated with a string and apply it to the display, processing, or data structures that it generates. Note that this might include the need to apply a resource-level default language when serializing or exchanging an individual value.

Using a consistent and well-defined data structure makes it more likely that different standards are composable and will work together seamlessly.

Metadata can be supplied without affecting the content itself.

Where metadata is unavailable, it can be omitted.

Consumers and producers do not have to instrospect the data outside of their normal processing.

Serialized files utilizing the dictionary and its data values will contain additional fields and can be more difficult to read as a result.

For existing document formats, it represents a change to the values being exchanged.

When a document is expected to consist of HTML or XML fragments and will be processed and displayed strictly in a markup context, the producer can use markup to convey the language of the content by wrapping strings with elements that have the lang or xml:lang attributes.

This approach, and thus the advantages, are effectively the same as in this section.

Producers insert Unicode tag characters into the data to tag strings with a language.

Consumers process the Unicode tag characters and use them to assign the language.

Unicode defines special characters that can be used as language tags. These characters are "default ignorable" and should have no visual appearance. Here is how Unicode tags are supposed to work:

Each tag is a character sequence. The sequence begins with a tag identification character. The only one currently defined is U+E0001, which identifies [BCP47] language tags. Other types of tags are possible, via private agreement. The remainder of the Unicode block for forming tags mirrors the printable ASCII characters. That is, U+E0020 is space (mirroring U+0020), U+E0041 is capital A (mirroring U+0041), and so forth. Following the tag identification character, producers use each tag character to spell out a [BCP47] language tag using the upper/lowercase letters, digits, and the hyphen character. A given source language tag, which is composed from ASCII letters, digits and hyphens, can be transmogrified into tags by adding 0xE0000 to each character's code point. Additional structure, such as a language priority list (see [RFC4647]) might be constructed using other characters such as comma or semi-colon, although Unicode does not define or even necessarily permit this.

The end of a tag's scope is signalled by the end of the string, or can be signalled explicitly using the cancel tag character U+E007F, either alone (to cancel all tags) or preceeded by the language tag identification character U+E0001 (i.e. the sequence <U+E0001,U+E007F> to end only language tags).

Tags therefore have a minimum of three characters, and can easily be 12 or more. Furthermore, these characters are supplementary characters. That is, they are encoded using 4-bytes per character in UTF-8 and they are encoded as a surrogate pair (two 16-bit code units) in UTF-16. Surrogate pairs are needed to encode these characters in string types for languages such as Java and JavaScript that use UTF-16 internally. The use of surrogates makes the strings somewhat opaque. For example, U+E0020 is encoded in UTF-16 as 0xDB40.DC20 and in UTF-8 as the byte sequence 0xF3.A0.80.A0.

These language tag characters could be used as part of normal Unicode text without modification to the structure of the document format.

Use of Unicode tag characters for language identification are strongly discouraged by the Unicode Consortium (and thus deprecated). These tag characters were intended for use in language tagging within plain text contexts and are often suggested as an alternate means of providing in-band non-markup language tagging. We are unaware of any implementations that use them as language tags.

Applications that treat the characters as unknown Unicode characters will display them as tofu (hollow box replacement characters) and may count them towards length limits, etc. So they are only useful when applications or interchange mechanisms are fully aware of them and can remove them or disregard them appropriately. Although the characters are not supposed to be displayed or have any effect on text processing, in practice they can interfere with normal text processes such as truncation. line wrapping, hyphenation, spell-checking and so forth.

By design, [BCP47] language tags are intended to be ASCII case-insensitive. Applications handling Unicode tag characters would have to apply similar case-insensitivity to ensure correct identification of the language. (The Unicode data doesn't specify case conversion pairings for these characters; this complicates the processing and matching of language tag values encoded using the tag characters.)

Moreover, language tags need to be formed from valid subtags to conform to [BCP47]. Valid subtags are kept in an IANA registry and new subtags are added regularly, so applications dealing with this kind of tagging would need to always check each subtag against the latest version of the registry.

The language tag characters do not allow nesting of language tags. For example, if a string contains two languages, such as a quote in French inside an English sentence, Unicode tag characters can only indicate where one language starts. To indicate nested languages, tags would need to be embedded into the text not just prefixed to the front.

Although never implemented, other types of tags could be embedded into a string or document using Unicode tag characters. It is possible for these tags to overlap sections of text tagged with a language tag.

Finally, Unicode has recently "recycled" these characters for use in forming sub-regional flags, such as the flag of Scotland (🏴󠁧󠁢󠁳󠁴󠁿󠁧), which is made of the sequence:󠁢󠁳󠁣󠁴󠁿

• 🏴 [U+1F3F4 WAVING BLACK FLAG]
• 󠁧 [U+E0067 TAG LATIN SMALL LETTER G]
• 󠁢 [U+E0062 TAG LATIN SMALL LETTER B]
• 󠁳 [U+E0073 TAG LATIN SMALL LETTER S]
• 󠁣 [U+E0063 TAG LATIN SMALL LETTER C]
• 󠁴 [U+E0074 TAG LATIN SMALL LETTER T]
• 󠁿 [U+E007F CANCEL TAG]

Producers do nothing.

Consumers run a language detection algorithm to determine the language of the text. These are usually statistically based heuristics, such as using n-gram frequency in a language, possibly coupled with other data.

There are no fundamental advantages to this approach.

Heuristics are more accurate the longer and more representative the text being scanned is. Short strings may not detect well.

Language detection is limited to the languages for which one has a detector.

Inclusions, such as personal or brand names in another language or script, can throw off the detection.

Language detection tends to be slow and can be memory intensive. Simple consumers probably can't afford the complexity needed to determine the language.