[css-text] I18N-ISSUE-334: 'letter-spacing' and Indic

State: OPEN WG Comment Product: CSS3-text Raised by: Addison Phillips Opened on: 2014-01-23 Description: Section 8.2: The section on "tracking" ('letter-spacing') may need to consider the effect on scripts such as Indic. The I18N WG asked for guidance from various Indic language contributors [1]. It was reported to use that 'letter-spacing' is not a native typographic style to these languages, however it is used in several languages, such as Hindi, for visual effect--if only rarely. The guidance we received, which is consistent with our understanding, is that letter-spacing should "break" the joining "bar" (shirorekha) in those scripts that use a "bar" and that the separation should be on syllable boundaries. These boundaries do not necessarily correspond to Unicode's default grapheme cluster boundaries, making proper description more complicated. [1] See thread: http://lists.w3.org/Archives/Public/public-i18n-indic/2013OctDec/

On Fri, Jan 24, 2014 at 06:32:28PM +0000, Phillips, Addison wrote: > that letter-spacing should "break" the joining "bar" (shirorekha) in Is it proposed that the specification add that this bar "should" be broken? My information is that this practice is common (or even universal outside of hand-formed word shapes), but that it is nevertheless undesirable in some circumstances (such as justification in a newspaper, where the spacing isn't semantically significant, and isn't intended to be noticeable). I am also told that handwriting would not break this bar [other than in the sense that आरम्भ might be described as having a broken bar]. (As I understand it, these gaps have become acceptable because they were largely unavoidable when using moveable type and typewriters.) I would thus be careful about specifying that this bar should always have visible gaps when letter-spacing is non-zero. (There's also the case of negative letter-spacing to consider, if we were to look at adding such a clause to the spec.) I can try to find some more information on this if that's useful. pjrm.

On Sat, Jan 25, 2014 at 10:30:00AM +1100, I wrote: > I would thus be careful about specifying that this bar should always have > visible gaps when letter-spacing is non-zero. More practically in terms of how to word the spec when addressing this issue (if indeed we don't wish to specify that this bar should always have visible gaps): the spec might merely say where letter-spacing is applied (namely at syllable boundaries), without necessarily even mentioning the bar. If gaps in the bar are mentioned, then the requirement might be on where gaps should/must not appear. pjrm.

On Jan 25, 2014, at 9:07, Peter Moulder <pjrm@mail.internode.on.net> wrote: > On Sat, Jan 25, 2014 at 10:30:00AM +1100, I wrote: > >> I would thus be careful about specifying that this bar should always have >> visible gaps when letter-spacing is non-zero. > > More practically in terms of how to word the spec when addressing this issue > (if indeed we don't wish to specify that this bar should always have visible > gaps): the spec might merely say where letter-spacing is applied (namely at > syllable boundaries), without necessarily even mentioning the bar. If gaps in > the bar are mentioned, then the requirement might be on where gaps should/must > not appear. Hm, I agree to leave things that we’re not certain undefined. Is the “syllable boundaries” different from the current grapheme cluster definitions? If no, I suppose there’s no need to update the spec. If different, is it possible to change Unicode grapheme cluster, or is it something that letter-spacing point needs to be different from grapheme cluster boundaries? /koji

[[ forwarding to both lists above, so that the thread can be brought back within the purview of the i18n tracker. Please keep both lists in cc going forward. missing emails from the thread on www-international beside the one included below are http://lists.w3.org/Archives/Public/www-style/2014Jan/0453.html http://lists.w3.org/Archives/Public/www-style/2014Jan/0454.html ]] -------- Original Message -------- Subject: Re: [css-text] I18N-ISSUE-334: 'letter-spacing' and Indic Resent-Date: Wed, 23 Apr 2014 08:10:43 +0000 Resent-From: www-style@w3.org Date: Wed, 23 Apr 2014 08:10:01 +0000 From: Koji Ishii <kojiishi@gluesoft.co.jp> To: Peter Moulder <pjrm@mail.internode.on.net> CC: www-style@w3.org <www-style@w3.org> On Jan 25, 2014, at 9:07, Peter Moulder <pjrm@mail.internode.on.net> wrote: > On Sat, Jan 25, 2014 at 10:30:00AM +1100, I wrote: > >> I would thus be careful about specifying that this bar should always have >> visible gaps when letter-spacing is non-zero. > > More practically in terms of how to word the spec when addressing this issue > (if indeed we don't wish to specify that this bar should always have visible > gaps): the spec might merely say where letter-spacing is applied (namely at > syllable boundaries), without necessarily even mentioning the bar. If gaps in > the bar are mentioned, then the requirement might be on where gaps should/must > not appear. Hm, I agree to leave things that we’re not certain undefined. Is the “syllable boundaries” different from the current grapheme cluster definitions? If no, I suppose there’s no need to update the spec. If different, is it possible to change Unicode grapheme cluster, or is it something that letter-spacing point needs to be different from grapheme cluster boundaries? /koji

> On Jan 25, 2014, at 9:07, Peter Moulder <pjrm@mail.internode.on.net> wrote: > >> On Sat, Jan 25, 2014 at 10:30:00AM +1100, I wrote: >> >>> I would thus be careful about specifying that this bar should always >>> have visible gaps when letter-spacing is non-zero. >> >> More practically in terms of how to word the spec when addressing this >> issue (if indeed we don't wish to specify that this bar should always have >> visible gaps): the spec might merely say where letter-spacing is applied >> (namely at syllable boundaries), without necessarily even mentioning the bar. If >> gaps in the bar are mentioned, then the requirement might be on where gaps >> should/must not appear. > > Hm, I agree to leave things that we’re not certain undefined. > > Is the “syllable boundaries” different from the current grapheme cluster > definitions? If no, I suppose there’s no need to update the spec. > > If different, is it possible to change Unicode grapheme cluster, or is > it something that letter-spacing point needs to be different from > grapheme cluster boundaries? I have sent a note to the Indic Layout Task Force list asking for clarification, and will report back what they say. My current understanding is that the units around which letter-spacing applies are different depending on which font is used. If the font indicates consonant clusters without showing the virama, then the unit is the whole consonant cluster plus vowel-signs, diacritics etc. This is very often likely to NOT coincide with extended grapheme clusters. If the font does show the virama explicitly, then the stretching appears just after the virama. I think that this DOES map to grapheme clusters as defined by Unicode. As i said, I'll let you know if the indic experts get back to me to confirm this one way or another. RI

>>>> I would thus be careful about specifying that this bar should always >>>> have visible gaps when letter-spacing is non-zero. >>> >>> More practically in terms of how to word the spec when addressing this >>> issue (if indeed we don't wish to specify that this bar should always have >>> visible gaps): the spec might merely say where letter-spacing is applied >>> (namely at syllable boundaries), without necessarily even mentioning the bar. If >>> gaps in the bar are mentioned, then the requirement might be on where gaps >>> should/must not appear. >> >> Hm, I agree to leave things that we’re not certain undefined. >> >> Is the “syllable boundaries” different from the current grapheme cluster >> definitions? If no, I suppose there’s no need to update the spec. >> >> If different, is it possible to change Unicode grapheme cluster, or is >> it something that letter-spacing point needs to be different from >> grapheme cluster boundaries? > > I have sent a note to the Indic Layout Task Force list asking for clarification, and will report back what they say. > > My current understanding is that the units around which letter-spacing applies are different depending on which font is used. If the font indicates consonant clusters without showing the virama, then the unit is the whole consonant cluster plus vowel-signs, diacritics etc. This is very often likely to NOT coincide with extended grapheme clusters. > > If the font does show the virama explicitly, then the stretching appears just after the virama. I think that this DOES map to grapheme clusters as defined by Unicode. > > As i said, I'll let you know if the indic experts get back to me to confirm this one way or another. CSS Text already has this text: > In some scripts such as Thai or Lao, the UA may apply the additional spacing within a character sometimes along with decompositions. In other scripts such as Myanmar, the UA may disallow applying the additional spacing between a specific pair of characters such as within a syllable. and I don’t think it’s right to talk about the bar in CSS. On the other hand, I had informal conversation with Unicode folks but it’s unlikely that Unicode defines letter-spacing point. If we want to make this information available, I think it’s best for the Indic Layout Task Force and I18N WG to publish a note, so that CSS can point to when it’s available. It’s great if we could get something like JLREQ/KLREQ, but a note specifically for Indic letter spacing could also work, and I guess it’s much easier to work on. Is this reasonable? /koji

On 01/24/2014 10:32 AM, Phillips, Addison wrote: > State: > OPEN WG Comment > Product: > CSS3-text > Raised by: > Addison Phillips > Opened on: > 2014-01-23 > Description: > Section 8.2: The section on "tracking" ('letter-spacing') may > need to consider the effect on scripts such as Indic. The I18N > WG asked for guidance from various Indic language contributors [1]. > > It was reported to use that 'letter-spacing' is not a native > typographic style to these languages, however it is used in > several languages, such as Hindi, for visual effect--if only rarely. > > The guidance we received, which is consistent with our understanding, > is that letter-spacing should "break" the joining "bar" (shirorekha) > in those scripts that use a "bar" and that the separation should be > on syllable boundaries. These boundaries do not necessarily correspond > to Unicode's default grapheme cluster boundaries, making proper > description more complicated. > > [1] See thread: http://lists.w3.org/Archives/Public/public-i18n-indic/2013OctDec/ Indeed, and this is why the spec explicitly allows for tailoring the units used for letter-spacing. Since we have no references on exactly how such units should be tailored, we cannot add any normative guidance on this. Unless you can provide such references, I am marking this issue as Closed No Change, since it does not seem to reflect a deficiency in the spec. ~fantasai

On 10/05/2014 19:39, fantasai wrote: > On 01/24/2014 10:32 AM, Phillips, Addison wrote: >> State: >> OPEN WG Comment >> Product: >> CSS3-text >> Raised by: >> Addison Phillips >> Opened on: >> 2014-01-23 >> Description: >> Section 8.2: The section on "tracking" ('letter-spacing') may >> need to consider the effect on scripts such as Indic. The I18N >> WG asked for guidance from various Indic language contributors [1]. >> >> It was reported to use that 'letter-spacing' is not a native >> typographic style to these languages, however it is used in >> several languages, such as Hindi, for visual effect--if only rarely. >> >> The guidance we received, which is consistent with our >> understanding, >> is that letter-spacing should "break" the joining "bar" (shirorekha) >> in those scripts that use a "bar" and that the separation should be >> on syllable boundaries. These boundaries do not necessarily >> correspond >> to Unicode's default grapheme cluster boundaries, making proper >> description more complicated. >> >> [1] See thread: >> http://lists.w3.org/Archives/Public/public-i18n-indic/2013OctDec/ > > Indeed, and this is why the spec explicitly allows for tailoring the units > used for letter-spacing. > > Since we have no references on exactly how such units should be tailored, > we cannot add any normative guidance on this. > > Unless you can provide such references, I am marking this issue as > Closed No Change, since it does not seem to reflect a deficiency > in the spec. For additional information on this topic see the thread that starts at http://lists.w3.org/Archives/Public/public-i18n-indic/2013OctDec/0001.html In summary, letter-spacing does apparently occur in Indic scripts such as Hindi, and the line is broken. There is still no authoritative pronouncement on whether the text unit around which spacing occurs is different when the virama is visible, but certainly when the virama is not visible the appropriate text unit is a syllable/akshar, which is not always equivalent to a grapheme cluster. There is a section in the draft of Indic Layout Requirements doc that refers to letter spacing (http://www.w3.org/International/docs/indic-layout/#letter-spacing), and it points to the section which describe indic syllables (http://www.w3.org/International/docs/indic-layout/#abnf-valid-segmentation-proposed-solution-for-layout-issues-in-indian-languages). hth ri

On 24/01/2014 18:32, Phillips, Addison wrote: > State: > OPEN WG Comment > Product: > CSS3-text > Raised by: > Addison Phillips > Opened on: > 2014-01-23 > Description: > Section 8.2: The section on "tracking" ('letter-spacing') may need to consider the effect on scripts such as Indic. The I18N WG asked for guidance from various Indic language contributors [1]. > > It was reported to use that 'letter-spacing' is not a native typographic style to these languages, however it is used in several languages, such as Hindi, for visual effect--if only rarely. > > The guidance we received, which is consistent with our understanding, is that letter-spacing should "break" the joining "bar" (shirorekha) in those scripts that use a "bar" and that the separation should be on syllable boundaries. These boundaries do not necessarily correspond to Unicode's default grapheme cluster boundaries, making proper description more complicated. > > [1] See thread: http://lists.w3.org/Archives/Public/public-i18n-indic/2013OctDec/ The i18n WG discussed this at our last telecon. We agree that the text around the definition of typographic character units helps address this issue. We suggest however, in view of current implementations of letter-spacing[1], the addition of a short note near the top of the letter-spacing section to remind implementers that "typographic charactere units" are not glyphs. thanks ri [1] http://192.168.0.4/International/tests/repository/css3-text/letter-spacing/css3-text-i18n-letter-spacing-001.html

On 10/23/2014 03:03 AM, Richard Ishida wrote: > > We suggest however, in view of current implementations of letter-spacing[1], the addition of a short note near the top of the > letter-spacing section to remind implementers that "typographic charactere units" are not glyphs. Given current implementations are based on earlier specs like CSS2.1 and my impression that “typographic character unit” is sufficiently scary that an implementer reading *this* spec will want to look it up, I think this reminder is not really necessary here, and I would therefore prefer not to add it. :) ~fantasai