This document describes and prioritises gaps for the support of the Devanagari script on the Web and in eBooks. In particular, it is concerned with text layout. It checks that needed features are supported in W3C specifications, in particular HTML and CSS and those relating to digital publications. It also checks whether the features have been implemented in browsers and ereaders. This is a preliminary analysis.
This section describes the status of this document at the time of its publication. Other documents may supersede this document. A list of current W3C publications and the latest revision of this technical report can be found in the W3C technical reports index at https://www.w3.org/TR/.
This document describes and prioritises gaps for the support of the Devanagari script on the Web and in eBooks. In particular, it is concerned with text layout. It checks that needed features are supported in W3C specifications, in particular HTML and CSS and those relating to digital publications. It also checks whether the features have been implemented in browsers and ereaders. This document complements the document Devanagari Layout Requirements, which describes the requirements for areas where gaps appear. It is linked to from the language matrix that tracks Web support for many languages.
The editor's draft of this document is being developed by the India International Program Layout Task Force, part of the W3C Internationalization Interest Group. It is published by the Internationalization Working Group. The end target for this document is a Working Group Note.
This document was published by the Internationalization Working Group as a Working Draft.
GitHub Issues are preferred for discussion of this specification.
Publication as a Working Draft does not imply endorsement by the W3C Membership.
This is a draft document and may be updated, replaced or obsoleted by other documents at any time. It is inappropriate to cite this document as other than work in progress.
This document was produced by a group operating under the 1 August 2017 W3C Patent Policy. The group does not expect this document to become a W3C Recommendation. W3C maintains a public list of any patent disclosures made in connection with the deliverables of the group; that page also includes instructions for disclosing a patent. An individual who has actual knowledge of a patent which the individual believes contains Essential Claim(s) must disclose the information in accordance with section 6 of the W3C Patent Policy.
This document is governed by the 15 September 2020 W3C Process Document.
The W3C needs to make sure that the needs of scripts and languages around the world are built in to technologies such as HTML, CSS, SVG, etc. so that Web pages and eBooks can look and behave as people expect around the world.
This page documents difficulties people encounter when trying to use the Hindi and Marathi languages in the Devanagari script on the Web.
Having identified an issue, it investigates the current status with regards to web specifications and implementations by user agents (browsers, e-readers, etc.), and attempts to prioritise the severity of the issue for web users.
A summary of this report and others can be found as part of the language matrix.
For a description of the Devanagari script and its use for the Hindi language see the (non-W3C) page Devanagari/Hindi, which summarises aspects of the orthography and typographic features, including relevant Unicode characters and their use.
This version of the document is a preliminary analysis
Gap analysis work usually starts with a preliminary analysis, conducted quickly by one or a small group of experts. Then a more detailed analysis is carried out, involving a wider range of experts. The detailed analysis may involve the development of tests, in order to illustrate issues and track results for browsers. The next phase is ongoing maintenance. It is expected that the resulting document will not be frozen: as gaps are fixed, this should be noted in the document. It is also possible that new gaps are noticed or arise, and they can be added to this document when that happens.
This document not only describes gaps, it also attempts to prioritise them in terms of the impact on the local user. The prioritisation is indicated by colour.
It is important to note that these colours do not indicate to what extent a particular features is broken. They indicate the impact of a broken or missing feature on the content author or end user.
Basic styling is the level that would be generally accepted as sufficient for most Web pages. Advanced level support would include additional features one might expect to include in ebooks or other advanced typographic formats. There may be features of a script or language that are not supported on the Web, but that are not generally regarded as necessary (usually archaic or obscure features). In this case, the feature can be described here, but the status should be marked as OK.
The decision as to what priority level is assigned to a described gap is down to the experts doing the gap analysis. It may not always be straightforward to decide. If a given section in this document refers to more than one feature that is broken, each with different impacts on Web users, the priority for the section should be the lowest denominator.
A cell can be scored as OK if the feature in question is specified in an appropriate specification, and is supported by user agents. A specification that is in CR or later and has two implementations in 'major' browsers will count. This means that the feature may not be supported in all browsers yet. (At some point in the future we may try to distinguish, visually, whether support is available in a specification but still pending in major browsers or applications.)
See also General page layout & progression for features such as column layout, page turning direction, etc. that are affected by text direction.
Are the script requirements for vertically oriented text met? What about if you mix vertical text with scripts that are normally only horizontal? Do you need a switch to use different characters in vertical vs. horizontal text? Does the browser support short runs of horizontal text in vertical lines (tate-chu-yoko in Japanese) as expected? Is the orientation of characters and the directional ordering of characters supported as needed? See available information or check for currently needed data.
This issue is applicable to many Indian languages.
Where browsers don't tailor grapheme cluster rules to take into account full orthographic syllables represented as conjuncts, upright characters in vertical lines of text don't break correctly.
Tests & results:
Interactive test, When devanagari characters are rendered upright in vertical text, orthographic syllables containing conjuncts are kept together
Gecko appears to segment on grapheme clusters. Blink, and Webkit keep conjuncts together. But all engines have problems with vowel-signs, especially but not only pre-base vowel-signs.
Priority: Marking this as advanced, since vertical text is not a common use case. In writing sytems like modern Tamil, it is even less common, since conjuncts are rarely used.
If this script runs right-to-left, are there any issues when handling that? Is bidirectional text adequately supported? What about numbers and expressions? Do the Unicode bidi controls and HTML markup provide the support needed? Is isolation of directional runs problematic? See available information or check for currently needed data.
Are there any character repertoire issues preventing use of this script on the Web? Do variation selectors need attention? Are there any other encoding-related issues? See available information or check for currently needed data.
Do the standard fallback fonts used in browsers (eg. serif, sans-serif, cursive, etc.) match expectations? Are special font or OpenType features needed for this script that are not available? See available information or check for currently needed data.
This covers ways of modifying the glyphs, such as for italicisation, bolding, oblique, etc. Do italic fonts lean in the right direction? Is synthesised italicisation problematic? Are there other problems relating to bolding or italicisation - perhaps relating to generalised assumptions of applicability? See available information or check for currently needed data.
Does the script in question require additional user control features to support alterations to the position or shape of glyphs, for example adjusting the distance between the base text and diacritics, or changing the glyphs used in a systematic way? Do you need to be able to compose/decompose conjuncts, or show characters that are otherwise hidden, etc? See available information or check for currently needed data.
If this script is cursive (eg. Arabic, N’Ko, Syriac, etc), are there problems or needed features related to the handling of cursive text? Do cursive links break if parts of a word are marked up or styled? Do Unicode joiner and non-joiner characters behave as expected? See available information or check for currently needed data.
Does the browser support requirements for baseline alignment between mixed scripts and in general? See available information or check for currently needed data.
By and large, Indian scripts in browsers are rendered with default fonts provided by back-end Operating Systems. If there are major differences in physical size for a particular font size, which most of the times are, it shows the mixed-script text in an un-organized way. This obviously affects all the base-line attributes.
Does your script need special text transforms that are not supported? Does your script convert letters to uppercase, capitalised and lowercase alternatives according to your typographic needs? Do you need to to convert between half-width and full-width presentation forms? See available information or check for currently needed data.
This is about how text is divided into graphemes, words, sentences, etc., and behaviour associated with that. Do Unicode grapheme clusters appropriately segment character units for your script? When you double- or triple-click on the text, is the expected range of characters highlighted? When you move through the text with the cursor, or backspace, etc. do you see the expected behaviour? (Some of the answers to these questions may be picker up in other sections, such as line-breaking, or initial-letter styling.) See available information or check for currently needed data.
If you put a span around the first two characters of the 3-character syllable स्तिठड (eg. in order to color or otherwise style it), the i vowel-sign no longer appears at the start of the word, but looks like it modifies the next character (ठ).
Apparently this requirement is a bit difficult to cater to as it requires the knowledge to be transferred by the font to the rendering engine about which part of the glyph is attributed to which code-point in storage. However, if it can be achieved, nothing like it.
When the Devanagari phrase separator । U+0964 DEVANAGARI DANDA (called purna viram in Hindi) or ॥ U+0965 DEVANAGARI DOUBLE DANDA (deergh viram in Hindi) are used, some browsers select them with the preceding word on double-click, while in other browsers they are selected separately.
The properties of purna viram and deergh viram should be the same as the properties of FullStop or other punctuation marks, and a new line should not begin with purna viram and deergh viram.
The Unicode concept of 'grapheme cluster' currently fails to represent syllabic conjuncts (plus vowels, etc) in scripts like Devanagari. This means that various editing operations, line breaking algorithms, vertical text, etc. are likely to break text at the wrong point.
Indic Layout Requirements provides a grammar for indian orthographic syllable boundaries which works for Devanagari, and CSS uses the concept of 'typographic character unit', rather than grapheme cluster, in its specs with the explanation that these cases are beyond the scope of the grapheme cluster concept and that implementations should provide appropriate support. In addition, a modification to the concept of grapheme cluster is currently in development at the Unicode Consortium, which is likely to resolve the problem for a script like Devanagari.
See requirements at: Indic Layout Requirements
Are there specific problems related to punctuation or the interaction of the text with punctuation (for example separation of punctuation from previous text, but allowing no line break between)? Are there issues related to handling of abbreviation, ellipsis, or iteration? Are there problems related to bracketing information or demarcating things such as proper nouns, etc? See available information or check for currently needed data.
This is about ways of marking text (see also specific sections dedicated to quotations and inline notes/annotations). Is it possible to express emphasis or highlight content as expected? Bold, italic and under-/over-lines are not always appropriate, and some scripts have their own unique ways of doing things, that are not in the Western tradition at all. Text delimiters mark certain items or sections off from the main text, such as book names in Chinese, quotations, head markers in Tibetan, etc, and often involve the use of punctuation. Is there any behaviour that isn't well supported, such as overlines for numeric digits in Syriac? Are there issues about the positioning or use of underlines? Some aspects related to the drawing of lines alongside or through text involve local typographic considerations. Do underlines need to be broken in special ways for this script? Do you need support for additional line shapes or widths? Does the distance or position of the lines relative to the text need to vary in ways that are not achievable? Are lines correctly drawn relative to vertical text? See available information or check for currently needed data.
Indian language text has some signs called as Matras which sometimes join above the shirorekha or below the normal baseline.
Applications should ensure that the underline and overline when getting rendered, should adaquetly be taken into consideration. Chrome browser seemed to be perfectly rendering the underline and overline feature by breaking the same where a matra occurs. Internet Explorer also handles it a bit differently by appropriately lowering or heightening the respective lines.
Are there any issues when dealing with quotations marks, especially when nested? Should block quotes be indented or handled specially? See available information or check for currently needed data.
The ruby spec currently specifies an initial subset of requirements for fine-tuning the typography of phonetic and semantic annotations of East Asian text, including furigana, pinyin and zhuyin fuhao systems. Is is adequate for what it sets out to do? What other controls will be needed in the future? What about other types of inline annotation, such as warichu? (For referent-type notes such as footnotes, see below.) See available information or check for currently needed data.
If the script has its own set of number digits, are there any issues in how they are used? Does the script or language use special format patterns that are problematic (eg. 12,34,000 in India)? What about date/time formats and selection - and are non-Gregorian calendars needed? Do percent signs and other symbols associated with number work correctly, and do numbers need special decorations, (like in Ethiopic or Syriac)? How about the management of personal names, addresses, etc. in web pages: are there issues? See available information or check for currently needed data.
Generally Latin Numerals are acceptable in Devanagari text. Most of the user community identifies with them. However, there could be cases where certain web-pages would prefer to have numerals in Devanagari to cater to mono-lingual (mono-script rather) readers. This is particularly required for input types number and date. It would be useful if the input type values are augmented with the script mnemonics e.g. number_deva or date_deva instead of changing it for entire page through some locale setting. There could be cases where a user may require both kinds of numbering in the same web-page.
Does the browser capture the rules about the way text in your script wraps when it hits the end of a line? Does line-breaking wrap whole 'words' at a time, or characters, or something else (such as syllables in Tibetan and Javanese)? What characters should not appear at the end or start of a line, and what should be done to prevent that? See available information or check for currently needed data.
See also hyphenation below.
Is hyphenation used for your script, or something else? If hyphenation is used, does it work as expected? (Note, this is about line-end hyphenation when text is wrapped, rather than use of the hyphen and related characters as punctuation marks.) See available information or check for currently needed data.
When text in a paragraph needs to have flush lines down both sides, does it follow the rules for your script? Does the script need assistance to conform to a grid pattern? Does your script allow punctuation to hang outside the text box at the start or end of a line? Where adjustments are need to make a line flush, how is that done? Do you shrink/stretch space between words and/or letters? Are word baselines stretched, as in Arabic? What about paragraph indents, or the need for logical alignment keywords, such as start/end, rather than left/right? See available information or check for currently needed data.
Some scripts create emphasis or other effects by spacing out the words, letters or syllables in a word. Are there requirements for this script/language that are unsupported? (For justification related spacing, see below.) See available information or check for currently needed data.
For various reasons wherever a word needs to be broken in constituent characters in case of Latin script, Indian language words can and should be broken based on Akshara as given here
As the W3C specification points to Unicode Text Segmentation (TR 29), it is observed that some of the browsers support it (e.g. Chrome and Firefox) whereas Microsoft Edge and Interner Explorer seems to break the words in individual characters.
It has been marked as basic as the Unicode Text Segmentation rules themselves need to be matured enough to cater to nuances of many languages that get written using Devanagari script. Some of the languages like Santali, require some special Nukta rules.
Also, in cases where there is wrong Akshara formation e.g. Consonant+Matra+Matra, the breaking seems to stack ill formed akshara into one set instead of clearly breaking it separate. This breaking behaviour needs to improve.
The CSS Counter Styles specification describes a limited set of simple and complex styles for counters to be used in list numbering, chapter heading numbering, etc.The rules plus more counter styles (totalling around 120 for over 30 scripts) are listed in the document Ready-made Counter Styles. Do these cover your needs? Are the details correct? Are there other aspects related to counters and lists that need to be addressed? See available information or check for currently needed data.
There is no general support for alphabetic counters that use the Devanagari script. There is also no way for users to create their own counter styles.
Known counter styles include
hindi (alphabetic), and
predefined-counter-styles This document contains templates for three Devanagari counter styles that can be applied by users if the custom counter styles spec is supported:
hindi (alphabetic), and
css-counter-styles-3 The numeric style is defined explicitly in this specification. The alphabetic counter styles can be defined by users if the Counter Styles spec's generic mechanism for defining counter styles was implemented. The spec is essentially done.
(In the case of failure, the style falls back to ordinary european decimal numbering.)
Tests & results:
I18n test suite, Simple numeric.
The numeric counter style is supported by Gecko, Blink and Webkit. (IE and Legacy Edge do not support it.)
I18n test suite, Alphabetic styles.
Neither Gecko, Blink, nor Webkit support the alphabetic styles without the user customisation feature.
I18n test suite, Basic custom counter styles.
The spec is essentially done, but so far only Gecko has implemented it. Blink and Webkit do not support it (however, we are hopefully close seeing this feature supported in Blink, too).
Browser bug reports: Blink
Priority: The impact of the lack of support cited here is mediated by the tendency to use western counters, however there appears to be a desire for native styles (at least the numeric) to be made available. So the impact is being marked as advanced.
Does the browser or ereader correctly handle special styling of the initial letter of a line or paragraph, such as for drop caps or similar? How about the size relationship between the large letter and the lines alongide? where does the large letter anchor relative to the lines alongside? is it normal to include initial quote marks in the large letter? is the large letter really a syllable? etc. Are all of these things working as expected? See available information or check for currently needed data.
This issue is applicable to most languages that form conjuncts from consonant clusters using a visible virama.
When the start of a line contains a 2-consonant cluster that uses a visible virama, ::first-letter should highlight only the first consonant+virama. This corresponds to a grapheme cluster, as defined by Unicode.
The Unicode Consortium defines grapheme clusters to, by default, stop after the first virama in a cluster.
Tests & results:
Interactive test, When ::first-letter is applied to Devanagari the browser will NOT select a 2-consonant cluster as a unit if the virama is visible
Interactive test, When ::first-letter is applied to Bengali the browser will only select the first consonant+virama in a cluster if the virama is visible
Gecko only highlights the initial character+virama. Blink, and Webkit select all clusters as a single unit, whether or not they are conjuncts or are rendered with a visible virama.
Priority: This choice needs to be discussed. If a cluster is rendered as a conjunct, it should be selected as a single unit. If, instead, the virama is displayed (ie. it is not a conjunct) then only the initial part of the cluster should be selected. Since the underlying code points are identical if a cluster is displayed as a conjunct or instead with visible viramas, it's not possible to distinguish one situation from another by working with the code points. In the absence of a technical solution that examines what the font used does when rendering, this is a difficult problem to solve. This priority rating says that, in the absence of a perfect solution, it is better to select a whole non-conjunct cluster than to break a conjunct. So it is labelled advanced, rather than basic.
This issue is applicable to most languages that form conjuncts from consonant clusters using an invisible virama.
When the start of a line contains a consonant cluster that uses a conjunct (rather than visible virama), ::first-letter should highlight the whole cluster.
Consonant clusters that form conjuncts using an invisible virama between the component letters need to be selected as a unit. This doesn't work well if segmentation relies on Unicode grapheme clusters, since a conjunct with two consonants will be parsed as two grapheme clusters (the first ending after the virama, and the second starting with the second consonant and including any following vowel-signs or other combining characters).
For these situations it is necessary to tailor the segmentation algorithm, so that it recognises the whole consonant cluster plus any attached vowel-signs or combining characters as a single unit.
For examples see Typographic character units in complex scripts.
css-text-3 CSS uses the concept of 'typographic character unit', rather than grapheme cluster, in its specs with the explanation that the cases just described go beyond the scope of the grapheme cluster concept and that implementations should provide appropriate support. The spec doesn't provide details about the support needed for each language.
The Unicode Consortium made some attempts to address this issue, but it has so far not yielded results. CLDR now flags up a few scripts for which conjuncts are common.
Tests & results:
Interactive test, When ::first-letter is applied to Devanagari the browser will select a 2-consonant conjunct as a unit
Interactive test, When ::first-letter is applied to Bengali the browser will select a conjunct as a unit, if the virama is hidden
Gecko breaks most of the half-form conjuncts (which is the large majority of all conjuncts in Devanagari), and they are broken into an initial consonant with visible virama and a following consonant. Blink, and Webkit fully select all conjuncts as a unit.
I18n test suite, Devanagari text
Priority: Keeping conjuncts together is a pretty basic requirement. Without a fix for this, authors need to manually mark up text to apply initial letter styling, but that isn't a very useful workaround.
How are the main text area and ancilliary areas positioned and defined? Are there any special requirements here, such as dimensions in characters for the Japanese kihon hanmen? The book cover for scripts that are read right-to-left scripts is on the right of the spine, rather than the left. Is that provided for? When content can flow vertically and to the left or right, how do you specify the location of objects, text, etc. relative to the flow? For example, keywords 'left' and 'right' are likely to need to be reversed for pages written in English and page written in Arabic. Do tables and grid layouts work as expected? How do columns work in vertical text? Can you mix block of vertical and horizontal text correctly? Does text scroll in the expected direction? Other topics that belong here include any local requirements for things such as printer marks, tables of contents and indexes. See available information or check for currently needed data.
Does your script have special requirements for footnotes, endnotes or other necessary annotations of this kind in the way needed for your culture? (There is a section above for purely inline annotations, such as ruby or warichu. This section is more about annotation systems that separate the reference marks and the content of the notes.) See available information or check for currently needed data.
Are vertical form controls well supported? In right-to-left scripts, is it possible to set the base direction for a form field? Is the scroll bar on the correct side? etc. See available information or check for currently needed data.
Sometimes a script or language does things that are not common outside of its sphere of influence. This is a loose bag of additional items that weren't previously mentioned. This section may also be relevant for observations related to locale formats (such as number, date, currency, format support).
There are many other CSS modules which may need review for script-specific requirements, not to mention the SVG, HTML, Speech, MathML and other specifications. What else is likely to cause problems for worldwide deployment of the Web, and what requirements need to be addressed to make the Web function well locally?