Cursive Anchors in Latin scripts

I have been experimenting with the kerning of glyphs using 'Cursive Anchors'.  Although these seem to work as expected in the font editor when the font is exported and installed in Windows the cursive positioning works in some applications but not in most.  Many applications including Microsoft Word just seem to ignore the anchors and position the glyphs as if the anchors did not exist.
Is there something I am missing ?
Are these features only reserved for Arabic scripts or something like that ?
Are they supposed to function correctly for Latin scripts but haven't been implemented for most applications ?
:s
Tagged:

Comments

  • In the spec it states 'curs' feature is only applied when the script is recognized as cursive. It doesn't specify which scripts are cursive; I'm sure Latin is not a one. But how about writing the cursive positioning rule in `mark` or another feature, does it solve the issue?
  • Khaled HosnyKhaled Hosny Posts: 289
    Some applications, in a misguided pursuit of optimization or out of laziness, act as a font policy and decide what features should be activated for what script, and are often mislead by outdated statements in the OpenType feature registry.

    Ideally, only a handful of features are script-specific since they are applied based on text analysis performed by the layout engine and can't be activated globally. Everything else should either be on by default of off by default and be applied globally when activated. This is what "modern" OpenType engines like HarfBuzz (and I think also Core Text) do.
  • Khaled HosnyKhaled Hosny Posts: 289
    Some "outdated" OpenType engines also restrict one types of lookups can be used with certain features, but again apart from a handful features that really need such restriction, there is no valid reasons to restrict what kind of lookup can be used with any given feature tag.
  • Peter BakerPeter Baker Posts: 178
    I just reported in another thread that cursive anchors appear to work with Latin script in apps using HarfBuzz (e.g. LibreOffice, Firefox) and in InDesign, but not in Word (for Mac—I can't test on Windows right now). But I'd like very much to hear from people more knowledgeable than I am.
  • Paul MillerPaul Miller Posts: 273
    edited July 2019
    I just reported in another thread that cursive anchors appear to work with Latin script in apps using HarfBuzz (e.g. LibreOffice, Firefox) and in InDesign, but not in Word (for Mac—I can't test on Windows right now). But I'd like very much to hear from people more knowledgeable than I am.
    My experiments were initiated due to the discussion in the Rhotic Hook thread.  It would be nice to use Cursive Anchors for this purpose but the sad truth is a large number of users for any font I produce will potentially be using it in Microsoft Word. :(
    The implementation of Cursive Anchors seems to be patchy at best so this is probably not the way to solve the Rhotic Hook problem.
  • Word doesn't do glyphing on Latin by default but you can force it to do that — by enabling at least one OT feature and it should work.
  • Paul MillerPaul Miller Posts: 273
    Word doesn't do glyphing on Latin by default but you can force it to do that — by enabling at least one OT feature and it should work.
    But most people who download my font won't know they must do this (even if it is in the documentation) and won't know how to do this.
  • Belleve InvisBelleve Invis Posts: 269
    edited July 2019
    Word doesn't do glyphing on Latin by default but you can force it to do that — by enabling at least one OT feature and it should work.
    But most people who download my font won't know they must do this (even if it is in the documentation) and won't know how to do this.
    Put that instructions in docs is a workaround for now.

    The real problem is still performance. Shaping is still really slow compared to "simple" text route.

    Here are some discussions about shaping performance from Sergey Malkin (who directly works on Word): https://github.com/harfbuzz/harfbuzz/issues/1463#issuecomment-505592189
Sign In or Register to comment.