Reference Label Generation Rules for the Devanagari Script

Overview

This document specifies a reference set of Label Generation Rules (LGR) for the Devanagari script for the second level. The starting point for the development of this LGR can be found in the related Root Zone LGR [RZ-LGR-3-Deva]. For details and additional background on the script, see "Proposal for a Devanagari Script Root Zone Label Generation Rule-Set (LGR)" [Proposal-Devanagari]. The format of this file follows [RFC 7940].

This is a DRAFT document released for public comments and not final. Please see the announcement on the ICANN website for public comments on the Second Level Reference LGRs for details on how to submit comments.

Repertoire

The repertoire contains 83 code points for letters in addition to 29 sequences. The two sequences U+0931 U+094D U+092F (ऱ्य) and U+0931 U+094D U+0939 (ऱ्ह) limit the character U+0931 ( ऱ ) DEVANAGARI LETTER RRA in its own specific context beyond which it does not stand by itself. Accordingly, while U+0931 (ऱ) is not listed by itself, it brings the total of code points available for letters to 84.

A number of other sequences have been defined in connection with the definition of variants (see "Variants" below).

The repertoire includes code points used by languages written in Devanagari that fall within levels 1 to 4 on the [EGIDS] scale. Boro, Braj, Dhundari, Mundari, Kharia are additionally covered. Though listed as level 4 on the EGIDS scale, Saraiki is not covered, because the Devanagari script is "no longer in use" by the Saraiki community. A non-exhaustive list of languages using each code point can be found in the comments.

The repertoire is a subset of [Unicode 6.3]. For more details, see Section 5, "Repertoire" in [Proposal-Devanagari]. (The proposal cited has been adopted for the Devanagari script portion of the Root Zone LGR.)

For the second level, the repertoire has been augmented with the HYPHEN-MINUS, and two sets of decimal digits:

• European (common) digits
• Devanagari digits

for a total of 133 repertoire elements.

Code points outside the Devanagari script that are listed in this file are targets for out-of-script variants and are identified by a reflexive (identity) variant of type "out-of-repertoire-var". They do not form part of the repertoire.

Each code point or range is tagged with the script or scripts that the code point is used with, one or more tag values denoting character category, and one or more references documenting sufficient justification for inclusion in the repertoire, see "References" below. For code points that are part of the repertoire, comments identify the languages using the code point.

Variants

This reference LGR is designed for use in zones shared by more than one script. Where appropriate, cross-script variants have been defined to mutually exclude labels from different scripts that could otherwise be substituted by the users. Once a label has been delegated for one script, any variant labels from other scripts consisting of cross-script variants would be blocked. Any label containing at least one code point that is not a cross-script variant would be considered distinct. Because of that, even large numbers of defined cross-script variants generally do not lead to a high percentage of labels experiencing a collision.

In-script variants: This LGR extends the in-script variants and cross-script variants as described in Section 6, "Variants", in "[Proposal-Devanagari]". The in-script variants in this fall into four broad categories:

• Vowel/Vowel sign followed by Nukta
• Unique Vowels and Vowel Signs required for Kashmiri
• Variants based on Candrabindu and Candra Vowel Signs followed by Anusvara
• Variant based on hyphen vs. eyelash reph

As hyphen is not permitted in some positions on the label, a variant context rule is used to undefine the respective variants in positions in which the seqences containing hyphen cannot occur.

Cross-script variants: In addition to these, cross-script variant analysis of Devanagari identified cross-script variants with the Gurmukhi and Bengali scripts.

Digit Variants: All Devanagari digits are treated as semantic variants of the corresponding common (ASCII) digits. By transitivity, they are also semantic variants of any native digits in scripts that also include the common digits. Such variant relations are deemed to exist implicitly by transitivity but are not listed explicitly in each reference LGR. (Omitting the listing of these other cross script digit variants does not affect index variant calculation, as the ASCII digit variant being smallest would always be the index variant.) Note that in addition to a transitive semantic variant relation, Devanagari digits ZERO, TWO, FOUR, FIVE and EIGHT are cross-script homoglyph or near homoglyph variants of the corresponding Gujarati digits. Devanagari digit ZERO is a cross-script homoglyph or near homoglyph of digit ZERO in many other scripts; these are already implicit semantic variants by transitivity and therefore not listed here. To keep digit variant sets manageable in zones where multiple scripts are present, no attempt has been made at identifying cross-script variants among digits of different numeric value or between a digit in one script and a letter in another, such as between digit zero and Latin letter 'o'.

Variant Disposition: All variants are of type "blocked"; labels that differ only by these variants are mutually exclusive: whichever label containing either of these variants is chosen earlier would be delegated, while any other equivalent label should be blocked.

This LGR does not define allocatable variants.

Context Rules for Variants: some of the variants defined in this LGR are "effective null variants", that is, some code points in the source map to "nothing" in the target with all other code points unchanged. (Because mappings are symmetric, it does not matter whether it is the forward or reverse mapping that maps to "null"). Such variants require a context rule to keep the variant set well behaved. Symmetry requires the same context rule for both forward and reverse mappings.

In other cases, the sequences or code points making up source and target are constrained by explicit context rules on the code points (or by implicit context rules defined for the adjacent code points). In such a case, any variants may require context rules that match the intersection between the effective contexts for both source and target; otherwise, a sequence might be considered valid in some variant label when it would not be valid in an equivalent context in an original label.

The specification of variants in this reference LGR follows the guidelines in [RFC 8228].

Character Classes

Devanagari is an alphasyllabary and the heart of the writing system is the akshar. This is the unit which is instinctively recognized by users of the script. As encoded, the writing system of Devanagari is composed of Consonants, Halant, Vowels, Anusvara, Candrabindu, Nukta and Visarga.

Consonants: Devanagari consonants all contain an implicit schwa /ə/. To make a full syllable, consonants may be followed by certain code points from one or more of the other groups (see "WLE rules" below). See Section 3.3.1, "The Consonants" of the [Proposal-Devanagari].

Halant: Because consonants contain an implicit vowel schwa /ə/, a special sign is needed to denote when this implicit vowel is stripped off. This sign is the Halant, encoded as U+094D ( ् ) DEVANAGARI SIGN VIRAMA. The Halant thus joins two consonants and creates conjuncts, which can be generally from 2 to 4 consonant combinations. In rare cases, it can join up to 5 consonants. However, this LGR will not enforce any length limit. See Section 3.3.2, "The Implicit Vowel Killer: Halant" in [Proposal-Devanagari].

Vowels and Matras: There are separate code points for vowels that are pronounced independently at the beginning of a syllable or after a vowel sound. To indicate a Vowel sound following a consonant other than the implicit shwa sound, a vowel sign (matra) is attached to the consonant. There is an equivalent matra for each vowel excepting U+0905 ( अ ) DEVANAGARI VOWEL A. See Section 3.3.3, "Vowels" in [Proposal-Devanagari].

Anusvara: The Anusvara shows a nasal at the end of a syllable. See Section 3.3.4, "The Anusvara" of the [Proposal-Devanagari].

Candrabindu: A Candrabindu denotes nasalization of the preceding vowel. Present-day Hindi users tend to replace the Candrabindu by the Anusvara. See Section 3.3.5, "Nasalization: Candrabindu" in [Proposal-Devanagari].

Nukta: The nukta sign is placed below a certain number of consonants to represent sounds found only in words borrowed from Perso-Arabic, English and other non-Aryan sources. It is also placed under the consonants U+0921 ( ड ) and U+0922 ( ढ ) to indicate flapped sounds. Apart from this, the Santali language uses Nukta adjoined to certain vowels and vowel signs. See Section 3.3.6, "Nukta" in [Proposal-Devanagari].

Visarga: U+0903 ( ः ) DEVANAGARI SIGN VISARGA, representing an aspiration at the end of a syllable, is frequently used in Sanskrit. See Section 3.3.7, "Visarga and Avagraha" in [Proposal-Devanagari].

Devanagari Digits: U+0966 (०) to U+096F (९) are a set of Devanagari-specific digits. They are used in alternation with the European (common) digits.

Common Digits: U+0030 (0) to U+0039 (9) are the set of digits from the ASCII range.

Whole Label Evaluation (WLE) and Context Rules

Common Rules

• Hyphen Restrictions — restrictions on the allowable placement of hyphens (no leading/ending hyphen and no hyphen in positions 3 and 4). These restrictions are described in section 4.2.3.1 of RFC 5891 [150]. They are implemented here as context rule on U+002D (-) HYPHEN-MINUS.
• Leading Combining Marks — restrictions on the allowable placement of combining marks (no leading combining mark). This rule is described in section 4.2.3.2 of RFC 5891 [150].
• One non-ASCII code point — An IDN label must contain one non-ASCII code point, as described in Section 4.2.4 of RFC 5891 [150].
• Digit-mixing — no mixing between different digit sets (European and Devanagari digits) is allowed; implemented here as a WLE rule with associated action.

Default Actions

Actions include the default actions for LGRs as well as that needed to invalidate labels with misplaced combining marks. They are marked with ⍟. For a description see [RFC 7940].

Devanagari-specific Rules

These rules ensure that the Devanagari label conforms to akshar formation norms for the Devanagari script. These norms are exclusively presented as context rules.

The following symbols are used in the names and comments for WLE rules:

• C → Consonant
• M → Matra
• V → Vowel
• B → Anusvara (Bindu)
• D → Candrabindu
• X → Visarga
• H → Halant / Virama
• N → Nukta
• S → Eyelash Reph (C2HC3) where:
  • C2 is U+0931 (ऱ) DEVANAGARI LETTER RRA
  • H is U+094D (्) DEVANAGARI SIGN VIRAMA
  • C3 comprises
    • U+092F (य) DEVANAGARI LETTER YA
    • U+0939 (ह) DEVANAGARI LETTER HA

The rules are:

• 1. N: must be preceded only by any of C1, V1 and M1
• 2. H: must be preceded by C or CN
• 3. M: must be preceded by C or CN
• 4. X: must be preceded by any of V, C, N or M
• 5. B: must be preceded by any of V, C, N or M
• 6. D: must be preceded by any of V, C, N or M
• 7. V: Can NOT be preceded by H

The following context rules are used for variants where a Nukta maps to a "null", or that are overlapped:

• Variant is not defined if followed by a Nukta
• Variant is not defined unless it is followed by V or C (including RRA) or end of label
• Variant is not defined where sequence starting with HYPHEN-MINUS is disallowed. (This context rule also applies to any sequences starting with HYPHEN-MINUS that are defined as variant targets, so that such sequences do not override the "Hyphen Restrictions" from RFC 5891)

See Section 7, "Whole Label Evaluation Rules (WLE)" of [Proposal-Devanagari].

Methodology and Contributors

This reference LGR for the Devanagari script for the 2nd Level has been developed by Michel Suignard and Asmus Freytag, based on the Root Zone LGR for Devanagari and information contained or referenced therein, see [RZ-LGR-3-Deva]. Suitable extensions for the second level have been applied according to the [Guidelines] and with community input. The original proposal for a Root Zone LGR for the Devanagari script, that this reference LGR is based on, was developed by the Neo-Brahmi Generation Panel (NBGP). For more information on methodology and contributors to the underlying Root Zone LGR, see Sections 4 and 8 in [Proposal-Devanagari], as well as [RZ-LGR-Overview].

References

The following general references are cited in this document:

[Guidelines]

ICANN, "Guidelines for Developing Reference LGRs for the Second Level, (Los Angeles, California: ICANN, 27 May 2020),
https://www.icann.org/en/system/files/files/lgr-guidelines-second-level-06jan16-en.pdf

[EGIDS]

Expanded Graded Intergenerational Disruption Scale,
https://www.ethnologue.com/about/language-status
(Accessed on 13 Nov. 2017)

[Proposal-Devanagari]

Neo-Brahmi Generation Panel, “Proposal for a Devanagari Script Root Zone Label Generation Rule-Set (LGR)”, 22 April 2019,
https://www.icann.org/en/system/files/files/proposal-devanagari-lgr-22apr19-en.pdf

[RFC 7940]

Davies, K. and A. Freytag, "Representing Label Generation Rulesets Using XML", RFC 7940, August 2016,
http://www.rfc-editor.org/info/rfc7940

[RFC 8228]

A. Freytag, "Guidance on Designing Label Generation Rulesets (LGRs) Supporting Variant Labels", RFC 8228, August 2017,
https://www.rfc-editor.org/info/rfc8228

[RZ-LGR-Overview]

Integration Panel, "Root Zone Label Generation Rules - LGR-3: Overview and Summary", 01 March 2020 (PDF),
https://www.icann.org/sites/default/files/lgr/lgr-3-overview-10jul19-en.pdf

[RZ-LGR-3-Deva]

ICANN, Root Zone Label Generation Rules for the Devanagari Script (und-Deva) , 10 July 2019 (XML)
https://www.icann.org/sites/default/files/lgr/lgr-3-devanagari-script-10jul19-en.xml

[Unicode 6.3]

The Unicode Consortium. The Unicode Standard, Version 6.3.0, (Mountain View, CA: The Unicode Consortium, 2013. ISBN 978-1-936213-08-5)
http://www.unicode.org/versions/Unicode6.3.0/

For references consulted particularly in designing the repertoire for the Devanagari script for the second level please see details in the Table of References below. References [0] to [11] refer to the Unicode Standard versions in which the corresponding code points were initially encoded. References [100] and above correspond to sources given in [Proposal-Devanagari] justifying the inclusion of the corresponding code points. Entries in the table may have multiple source reference values.

4.1 Character Classes

The following table lists all named and implicit classes with their definition and a list of their members intersected with the current repertoire (for larger classes, this list is elided).

4.2 Whole label evaluation and context rules

The following table lists all named rules defined in the LGR and indicates whether they are used as trigger in an action or as context (when or not-when) for a code point or variant.

4.3 Actions

The following table lists the actions that are used to assign dispositions to labels and variant labels based on the specified conditions. The order of actions defines their precedence: the first action triggered by a label is the one defining its disposition.