Imporve Dinka langauge support in Noto Sans and Noto Serif families

[andjc]

The Dinka language like many African languages uses the n-form of U+014A LATIN CAPITAL LETTER ENG. The Noto fonts default to the N-form. Noto Sans provides three versions of the glyph, but only supplied access to the alternative glyphs using the aalt feature. This is useful for applications such as Indesign, but isn't really suitable for web development.

I would suggest:

1. add a cvNN feature to access the variant glyphs for U+014A
2. add the DNK OT language system (for applications that provide adequate locl support) - requires Eng.alt1 glyph

[brawer]

@andjc, do you know which other languages (beyond Dinka) use a capital Eng letter with an n instead of N shape?

[miguelsousa]

It may be better to make the n-shape the default and then implement the N-shape via locl for Sami languages (NSM and SKS).

[brawer]

See also https://github.com/googlei18n/noto-fonts/issues/808#issuecomment-298403290 about Eng (and other letters).

[andjc]

@brawer Many languages use n-shaped Eng, there is also some languages that use the n-shape without descender (Eng.alt3 I think)

The key problem is that the OT language system tags are limited. There are more languages that would need to access alternative Eng glyphs than there are OT language tags. One of the reasons in the request I suggested adding a cvNN or ssNN feature to access the alternative Eng shapes.

[andjc]

@miguelsousa changing default glyph although useful for African languages would just shift the problem. You would still need to identify languages that require the N-form and figure out how to support them, either through ssNN or cvNN features, there are more than just the Sami languages, and like the African languages, there are unlikely to be sufficient OT language tags to support the languages in question.

well thought out application of OpenType features combined with good documentation is critical for language support.

[andjc]

@brawer I am working on a list of OT language tags that need Eng.alt1. But it will take time since I will also need to determine if any of those languages also require other changes in Noto fonts. Pointless implementing a language system if support is only partial.

[brawer]

Hi @andjc, can you have a look at this PDF file? It is the Universal Declaration of Human Rights in Northeastern Dinka, rendered with a draft version for the next release of Noto Sans.

Currently, the font has four variants for the uppercase Ŋ letter. For Dinka, it currently uses the default variant, which is the one named Eng in the picture below. Because of your feedback (much appreciated!), we now know that the Eng glyph is the wrong variant for Dinka, and it sounds like Dinka should use the variant Eng.alt1 instead. Can you confirm that Dinka really uses that specific glyph, and not Eng.alt3? (Not sure why our font vendor made an Eng.alt2 glyph which seems identical to Eng; let’s ignore this here).

Also, can you tell us if any other glyphs than Ŋ look wrong in the attached PDF file? The PDF contains the same text in normal casing; in small caps; and in uppercase; thanks for checking all three.

For later reference, here’s the HTML file from which the above PDF was rendered.

[andjc]

@brawer The PDF seems to be using the Eng (Eng.alt2) - N-form, which is wrong for Dinka and most African languages. The correct glyph for Dinka, is Eng.alt1, as it is for most African languages that use Eng. From memory, Eng.alt3 is an archaic form, but may be in use, not sure, I would need to delve into my archives for further information.

There would appear to be four identifiable glyphs for Eng.

• The N-form (Noto's Eng and Eng.alt2)
• The n-form with descender (Noto's Eng.alt1)
• The n-from without descender (Not present in Eng)
• The turned-G form (Noto's Eng.alt3)

The U+00EF U+00EF sequence may need some work, maybe adjusting the combining diaeresis glyph or via kerning. This combination doesn't occur in the PDF, but if the dialect shown is similar to the ones I am more familiar with, there is an odd character sequence that might supposed to be U+00EF U+00EF.

I Would also look at the kerning of II as well.

Other character sequences seem to be acceptable.

[andjc]

This issue doesn't appear to be going anywhere, and as is , Noto Sans and Noto Serif are far form ideal for Dinka and various other African languages. It is best to use some other font at this point in time. And as the person who raised this issue initially, I'd like to close it.

[fitojb]

@andjc The fact you have given up does not mean the bug has magically disappeared. It is still worth tracking.

[andjc]

@fitojb I do not specifically consider it a bug, more correctly speaking it was a flawed project dsign/parameter from early period of the project.

Even if the issue was fixed tomorrow. The existing problematic fonts will be on Android devices many years into the future. For web developers, other solutions need to be used.

Probably self hosting appropriate web fonts and if Noto Sans is used ensuring lcal versions on device are not used. Since tbere is no way for a web dveloper to specify version of font to be used.

[c933103]

Android can use specific Noto font variant according to languages as you can see in Japanese apps versus Chinese apps. The problem is that the font that support different glyph have to come with the device, which can be solved by first implementing it into the font and then wait for a few years before most customers replaced their old phone with new one that have the new font embedded, if we disregard system updates. It is certainly a slow process and web developers should use web fonts and such to help before most users picked new font instead, but then it's still better than stuck in the situation eternally

[simoncozens]

This appears to be another Eng selection issue, which we'll track at #161. We will fix this, but part of fixing it is getting organised about the scope of changes required.

[andjc]

@simoncozens for proper Dinka support you also need to tick off ïï and ÏÏ as well, probably also the high tone equivalents as well as per unified orthography and grammar. But that probably needs to be separate issues.

[simoncozens]

We have those, but they need kerning. I'm aware of it!