notofonts / devanagari

Noto Devanagari
SIL Open Font License 1.1
1 stars 2 forks source link

Devanagari: vowels over the hanging line+candrabindu intersect in document fonts #3

Closed marekjez86 closed 3 months ago

marekjez86 commented 9 years ago

copied from https://github.com/googlei18n/noto-alpha/issues/206

Imported from Google Code issue notofonts/noto-fonts#206 created by roozbeh@google.com on 2014-04-18T01:57:00.000Z:


The sequence DDA+AU+CANDRABINDU (0921,094C,0901) in Noto Sans Devanagari results in the AU vowel and the CANDRABINDU intersecting.

See sample rendering from Noto, together with renderings from SIL's Annapurna and Microsoft's Utsaah and Nirmala.

This issue is especially important for Nepali, as it commonly uses candrabindu in its normal orthography.

This also happens with the other vowels going over the hanging line: II, SHORT E, E, AI, CANDRA O, SHORT O, and O.

In the UI fonts, there is another bug (issue notofonts/khmer#12) that predicts me from testing this at the moment, but after fixing that issue, the UI fonts should also be tested to make sure the glyphs don't intersect.

Comment #2 originally posted by jelle.bosma@monotypeimaging.com on 2014-04-18T07:17:59.000Z:


This is not a bug of Noto, but a design decision to make a ligature of the marks. Similar to the reph, which joins with the vowel in order not to end up on the next syllable. In the other three fonts the candrabindu is too far to the right and might be covered by an i-vowel or another top mark on the following syllable. In Noto it might still touch an i-vowel or a mark to the right, but you can always recognise it. Which is definitely not the case in the other three fonts!

nirmala.png nirmala

noto.png noto

sil.png sil

utsaah.png utsaah --- dougfelt commented on Jul 10 This is still an issue.

It seems that one could let the candrabindu be farther to the right when there was no mark for it to interfere with, and only use the squeezed position when necessary. This would involve taking more context into account. What do the other fonts do? Does text with interfering marks occur frequently? --- JelleBosmaMT commented on Jul 11 This is NOT an issue!

marekjez86 commented 9 years ago

I presented the following characters to several Nepali users (note the chars were fully presented in the attached files - not cut like here):

nep-char

I got two responses:

from ST: "Hi Marek, All four are readable, and I would think of them as different Nepali fonts. However, the second one (noto-20150807.png) has the chandrabindu overlapping, which does not look good."

from GSK: "agreed. noto-20150807.png is still readable at this size but there shouldn't be any overlap"

The respondent sample 100% voted for non-overlapping candrabindu

JelleBosmaMT commented 9 years ago

Item notofonts/devanagari#3 ?? Item notofonts/noto-fonts#206 ?? Was this not item notofonts/noto-fonts#6 ???

On 10-aug.-15 15:17, "Marek Jeziorek" notifications@github.com wrote:

copied from googlei18n/noto-alpha#206 https://github.com/googlei18n/noto-alpha/issues/206 Imported from Google Code issue notofonts/noto-fonts#206 https://code.google.com/p/noto-alpha/issues/detail?id=206 created by roozbeh@google.com https://code.google.com/u/roozbeh@google.com/ on 2014-04-18T01:57:00.000Z:


The sequence DDA+AU+CANDRABINDU (0921,094C,0901) in Noto Sans Devanagari results in the AU vowel and the CANDRABINDU intersecting. See sample rendering from Noto, together with renderings from SIL's Annapurna and Microsoft's Utsaah and Nirmala. This issue is especially important for Nepali, as it commonly uses candrabindu in its normal orthography. This also happens with the other vowels going over the hanging line: II, SHORT E, E, AI, CANDRA O, SHORT O, and O. In the UI fonts, there is another bug (issue notofonts/khmer#12 https://github.com/googlei18n/noto-fonts/issues/4) that predicts me from testing this at the moment, but after fixing that issue, the UI fonts should also be tested to make sure the glyphs don't intersect. Comment notofonts/khmer#11 https://code.google.com/p/noto-alpha/issues/detail?id=206#c2 originally posted by jelle.bosma@monotypeimaging.com on 2014-04-18T07:17:59.000Z:


This is not a bug of Noto, but a design decision to make a ligature of the marks. Similar to the reph, which joins with the vowel in order not to end up on the next syllable. In the other three fonts the candrabindu is too far to the right and might be covered by an i-vowel or another top mark on the following syllable. In Noto it might still touch an i-vowel or a mark to the right, but you can always recognise it. Which is definitely not the case in the other three fonts!

nirmala.png

https://cloud.githubusercontent.com/assets/9939212/8630608/ca82cb30-271c- 11e5-9e5f-19e37b35a1f1.png noto.png

https://cloud.githubusercontent.com/assets/9939212/8630610/d5dca6c2-271c- 11e5-90a4-ff6524c046ff.png sil.png

https://cloud.githubusercontent.com/assets/9939212/8630613/df6d3760-271c- 11e5-9f7a-849b079280da.png utsaah.png

https://cloud.githubusercontent.com/assets/9939212/8630615/ede5e29c-271c- 11e5-84df-ae641c0f5be5.png --- dougfelt commented on Jul 10 This is still an issue. It seems that one could let the candrabindu be farther to the right when there was no mark for it to interfere with, and only use the squeezed position when necessary. This would involve taking more context into account. What do the other fonts do? Does text with interfering marks occur frequently? --- JelleBosmaMT commented on Jul 11 This is NOT an issue! — Reply to this email directly or view it on GitHub https://github.com/googlei18n/noto-fonts/issues/474.

marekjez86 commented 9 years ago

It may have been item notofonts/noto-fonts#6. I just joined the project. I moved it from a private bug list to a public list. Hence it became item notofonts/devanagari#3.

JelleBosmaMT commented 9 years ago

As the designer of the font I took the decision to have candrabindu’s overlap with top vowels. It does not matter if that is an i, ii, e, ai or one of the others. To me the best compromise, between relative size, position in relation to the base glyph, fitting into the design language of a sans-serif typeface, as well as working with bold and extra bold fonts. If it is considered a feature that needs to be in a bug list, one item should be sufficient.

On 10-aug.-15 16:49, "Marek Jeziorek" notifications@github.com wrote:

It may have been item notofonts/noto-fonts#6 https://github.com/googlei18n/noto-fonts/issues/6. I just joined the project. I moved it from a private bug list to a public list. Hence it became item

notofonts/devanagari#3 https://github.com/googlei18n/noto-fonts/issues/474. — Reply to this email directly or view it on GitHub https://github.com/googlei18n/noto-fonts/issues/474#issuecomment-12947964 3.

marekjez86 commented 9 years ago

Thank you for the reply. You are correct that "one item should be sufficient." It is https://github.com/googlei18n/noto-fonts/issues/474 (all other items were private and will be closed, this is the only one that's public - I will copy though the information from googlei18n/noto-alpha#6, I do not want to lose the comments from there). Marek

On Mon, Aug 10, 2015 at 8:58 AM, Jelle Bosma notifications@github.com wrote:

As the designer of the font I took the decision to have candrabindu’s overlap with top vowels. It does not matter if that is an i, ii, e, ai or one of the others. To me the best compromise, between relative size, position in relation to the base glyph, fitting into the design language of a sans-serif typeface, as well as working with bold and extra bold fonts. If it is considered a feature that needs to be in a bug list, one item should be sufficient.

On 10-aug.-15 16:49, "Marek Jeziorek" notifications@github.com wrote:

It may have been item notofonts/noto-fonts#6 https://github.com/googlei18n/noto-fonts/issues/6. I just joined the project. I moved it from a private bug list to a public list. Hence it became item

notofonts/devanagari#3 https://github.com/googlei18n/noto-fonts/issues/474. — Reply to this email directly or view it on GitHub < https://github.com/googlei18n/noto-fonts/issues/474#issuecomment-12947964 3>.

— Reply to this email directly or view it on GitHub https://github.com/googlei18n/noto-fonts/issues/474#issuecomment-129506313 .

Marek Z Jeziorek [ 老马 ] | marekj@google.com | 312 725-6958

jungshik commented 9 years ago

Note that in googlei18n/noto-alpha#6, @cibu wrote that it's better to use the current glyphs/designs for Noto Sans UI Devanagari.

The remaining question from that bug was what to do about 'Noto Sans Devanagari' (non-UI / Devangari). @JelleBosmaMT : what do you think of @cibu's question?

Anyway, copying over the conversation from googlei18n/noto-alpha#6 would be necessary to collect all the information in one place.

marekjez86 commented 9 years ago

the following info is from https://github.com/googlei18n/noto-alpha/issues/6

Imported from Google Code issue notofonts/noto-fonts#6 created by roozbeh@google.com on 2013-10-16T00:02:31.000Z:


In the sequence लीँ <la+ii+candrabindu> (U+0932, U+0940, U+0901), the candrabindu touches the ii in Noto Sans Devanagari. It shouldn't.

The problem doesn't exist in Microsoft's Mangal font (although it's made too small in the Mangal font, which is not good).

Attaching screenshots from Noto Sans Devanagari and Mangal.

(See also issue notofonts/khmer#12 )

Magal: mangal

Noto Sans Deva: noto 2


Comment #3 originally posted by jelle.bosma@monotypeimaging.com on 2013-11-11T20:26:35.000Z:


It is the understanding of the designer that in normal usage it is the anusvara dot that is used in combination with top vowel marks, including the ii. That is why in the UI font the candrabindu ligates into the anusvara. In the document font it overlaps with the ii vowel. Making a tiny candrabindu seems a very bad idea. Moving it up it may be clipped. Move it to the right it may overlap with a mark on the next syllable. That is much worse than forming a ligature with the vowel it belongs to.


Comment #4 originally posted by vvasuki@google.com on 2013-11-11T21:14:26.000Z:


Is "normal" usage a proxy for modern hindi usage? Chandrabindu and anusvAra are distinct in Sanskrit, so that this assumption of interchangeability does not hold, where devanAgarI is also used.

Attaching lohit rendering if it helps the designer.

lii-lohita


Comment #5 originally posted by cibu@google.com on 2013-11-11T23:01:18.000Z:


The overlapping in Noto looks better than the solutions in Mangal or Lohit. Mangal is using thinner strokes for Candrabindu and Lohit has shifted it up using more vertical space. Thinner strokes are ugly; using more vertical space is not ideal for at least the UI version.


Comment #7 originally posted by cibu@google.com on 2013-12-10T02:13:41.000Z:


Hi Jelle,

You had mentioned that if you move up the candrabindu clipping might occur. Could you elaborate in the context of this body font? Since Lohit is using such approach, I am wondering how it is handling the clipping.


Per hindi speaker I (marekjez86) asked:

HOWEVER, he claimed that this combination is not pronounceable in hindi (i.e. unlikely to occur in the modern hindi).

In what language this combination would be pronounceable?


JelleBosmaMT responded:

Evidently some combinations may occur in Nepali, and some very rare cases in Sanskrit. But in Hindi the candrabindu is used only if there is no other mark occupying the room above the base glyph. So hindi readers do not have an expectation regarding the characters to touch or not to touch, they would not expect a candrabindu at all.

This seems to me a type design issue, where being a Hindi speaker does not necessarily grant a greater expertise than being a type designer or typographer who is used to arrange typographic elements. If in the Devanagari script we would exclude characters from touching each other, there would be very few characters left.

On 04-aug.-15 03:20, "Marek Jeziorek" notifications@github.com wrote:

Per hindi speaker I asked:

  • both lohit rendering (the green one) and Mangal font have equivalent readability
  • Noto Sans Deva rendering is wrong ("no touching" :-))

HOWEVER, he claimed that this combination is not pronounceable in hindi (i.e. unlikely to occur in the modern hindi). In what language this combination would be pronounceable? — Reply to this email directly or view it on GitHub https://github.com/googlei18n/noto-alpha/issues/6#issuecomment-127446520 .


roozbehp asked:

@marekjez86, would you please ask our internal Nepali expert? @xiangyexiao has the contact.

JelleBosmaMT commented 9 years ago

There is no difference between UI and document font (If that is the question.)

On 10-aug.-15 23:45, "jungshik" notifications@github.com wrote:

Note that in googlei18n/noto-alpha#6 https://github.com/googlei18n/noto-alpha/issues/6, @cibu https://github.com/cibu wrote that it's better to use the current glyphs/designs for Noto Sans UI Devanagari.

The remaining question from that bug was what to do about 'Noto Sans Devanagari' (non-UI / Devangari). @JelleBosmaMT https://github.com/JelleBosmaMT : what do you think of @cibu https://github.com/cibu's question? Anyway, copying over the conversation from googlei18n/noto-alpha#6 https://github.com/googlei18n/noto-alpha/issues/6 would be necessary to collect all the information in one place.

— Reply to this email directly or view it on GitHub https://github.com/googlei18n/noto-fonts/issues/474#issuecomment-12962072 2.

cibu commented 9 years ago

As a Hindi reader, I can say that overlapping does not make it unreadable or illegible. Personally, I think overlapping is better design - just an opinion.

JelleBosmaMT commented 8 years ago

The comment that Cibu made is to the point. This is not a "priority-high" issue, it is not an issue at all.

roozbehp commented 8 years ago

The bug is based on feedback from our Nepali expert, Saroj. Marek and Xiangye can contact him again and ask.

xiangyexiao commented 8 years ago

@marekjez86 please contact saroj if you haven't. The wanted fixing date is April/May for Android.

JelleBosmaMT commented 8 years ago

Bear in mind this is about type design, not about language.

Jelle

On 28-feb.-16 21:59, "Xiangye Xiao" notifications@github.com wrote:

@marekjez86 https://github.com/marekjez86 please contact saroj if you haven't. The wanted fixing date is April/May for Android. — Reply to this email directly or view it on GitHub https://github.com/googlei18n/noto-fonts/issues/474#issuecomment-18994568 5.

nizarsq commented 3 years ago

@marekjez86 what is the status of this issue? is this still high priority issue?

simoncozens commented 2 years ago

To summarise the discussion so far:

I don't think any issue which has been open for seven years can reasonably be called "high priority". But we should get input from Nepali experts, and then get it fixed if needed.

simoncozens commented 3 months ago

I believe this has now been fixed in v2.006.