bugFree UCS Outline Fonts - Bugs: bug #40081, Ottoman GAF


bug #40081: Ottoman GAF

Submitted by:  Rosella Capriotti <rosy58>
Submitted on:  Sat 21 Sep 2013 03:21:05 PM UTC  
Category:  individual character(s) Severity:  3 - Normal
Item Group:  glyph(s) missing Status:  Need info
Privacy:  Public Assigned to:  Steve White <Stevan_White>
Open/Closed:  Open Release:  2012-05-03

Add a New Comment(Rich Markup)

You are not logged in

Please log in, so followups can be emailed to you.


( Jump to the original submission)

Sun 29 Sep 2013 08:52:09 PM UTC, comment #27:

My understanding is that it is a glyph variant and as such not eligible for encoding. Some characters encoded in the Arabic block are border line glyph variants, like the U+06CC vs U+064A or U+06A9 vs U+0643 but I think they have a history of being encoded as separate characters in pre-Unicode encodings and have the support of national standards bodies considering them different characters, which cannot be said about this (rarely used) variant of Gaf. Wikipedia talk page of Ottoman Turkish has some interesting opinions https://en.wikipedia.org/wiki/Talk%3AOttoman_Turkish_language#Comment.

That being said, I hardly know anything about Turkish orthography, so my opinion should be taken with a big grain of salt.

Khaled Hosny <khaledhosny>
Sun 29 Sep 2013 08:15:39 PM UTC, comment #26:

I agree with Kaled on all the points he's made, indeed I've been saying pretty much the same thing here. I'd like to ask him whether and where it's possible to ask the implementation of such letter, however one might call it, in Unicode. Does he deem it appropriate?

Rosella Capriotti <rosy58>
Sun 29 Sep 2013 01:29:34 PM UTC, comment #25:

I'm not sure where this bug report is going. It seems the nature of the problem is unclear. I provided several suggestions, but they were evidently unsatisfactory. When the nature of the probem becomes more concrete, please let us know!

Regarding the bugs with various PDF-generating software: You have my response elsewhere, I believe.

Steve White <Stevan_White>
Project AdministratorIn charge of this item.
Sun 29 Sep 2013 09:31:36 AM UTC, comment #24:

I think calling that variant of Gaf an “Ottoman” Gaf is a bit of stretch; the use of Gaf was introduce very late in the history of Ottoman orthography and never gained wide spread use, and it seems Ottoman calligraphers just derived it from Naskh calligraphy rather than from Nastaliq as it is the case with Persian, and this is not even completely true, in the very few Ottoman uses of Gaf I have seen, both forms were used.

Older version of Amiri font used the Naskh form because it looked more natural fit for the style of the font, but it turned out to be problematic in practice (the horizontal stroke hangs to much on the right side resulting in many collisions, that would require many complex substitution rules to handle) so I ended up making the other form the default and keeping the Naskh form as a stylistic set for people who might prefer it. I wouldn’t go far to making it a locale-specific variant, because I think it is rather a stylistic different.

Khaled Hosny <khaledhosny>
Sat 28 Sep 2013 05:51:41 PM UTC, comment #23:

Hello Steve,

A diacritic is NOT a letter, as the name says, the diacritic would be added to ARABIC KAF to form OTTOMAN GAF but from a strictly linguistical point of view OTTOMAN GAF is no letter on its own, i.e. it doesn't get its entry on the dictionary.

By the way, this a low-priority issue for me, if you want to do something for me, try to solve https://savannah.gnu.org/bugs/?40058 or https://savannah.gnu.org/bugs/?39806#comment23


Rosella Capriotti <rosy58>
Sat 28 Sep 2013 02:20:10 PM UTC, comment #22:

polyglossia? Oh the LaTeX package. I never really knew what it did, and I was just trying to clean up the test file.

I do not know the reasoning for adding 0762 etc. I can look it up, but as was discussed before, there are various ways a character could get into Unicode. That doesn't change my statement that normally glyph variants are not accepted.

Regarding your new views on the Ottoman gaf, the question would be, is this diacritic an existing letter? If so, maybe we can do something for you.

Steve White <Stevan_White>
Project AdministratorIn charge of this item.
Thu 26 Sep 2013 01:33:31 PM UTC, comment #21:


I forgot to state clearly that GAF is not really a letter in Ottoman Turkish as it is in Persian: in manuscripts you'll find KAF used for both the voiceless and the voiced palatal stop. GAF appeared only in the late Ottoman period, that is the beginning of the 20th century and even then failed to gain widespread usage: it was never deemed to be a letter on its own but rather the one stroke in the initial and medial position, and the two strokes in the final and isolated position were something of a DIACRITIC, a sign to mark voicedness, in cases when confusion might arise.

So, if we were to conform ourselves to REAL Ottoman usage, we should speak of a diacritic to combine with ARABIC KAF, not a glyph variant of PERSIAN GAF, and not even a letter on its own.

Rosella Capriotti <rosy58>
Thu 26 Sep 2013 11:56:02 AM UTC, comment #20:

Hello Steve,

I suppose you got better results because you dispensed with polyglossia, I usually use polyglossia to set the language.

No, the fundamental issue is not software bugs, but the fact that up to version 6.1 they've been adding characters like U+0762 and U+0763 that are essentially GLYPH VARIANTS, and what's more, they involve the difference of shape between Arabic and Persian-style KAFs: it's clear to me the Unicode consortium for some reasons unknown to me doesn't consider this to be a matter of style.

Now, Ottoman GAF is nothing more than an Arabic-style GAF: I don't see why it should be confined to a glyph variant, when you see the policy the Unicode Consortium has been CONSISTENTLY adopting even in its newest versions.

Rosella Capriotti <rosy58>
Thu 26 Sep 2013 12:12:49 AM UTC, comment #19:


This issue with XeLaTeX/fontspec is new, since your posting before last. I haven't looked into it, and scarcely know what you're talking about. I assume you've discussed this with the proper XeLaTeX/fontspec people already.

I attach a couple of experiments, in HTML as well as XeLaTeX and LuaLaTeX, just of typesetting some Ottoman Turkish in Arabic script (near as I can tell -- got it from Wikipedia). There is a lot written on Arabic in LaTeX, and... it's not encouraging. Anyway, I got LuaLaTeX to at least display the letters in the right order. With XeLaTeX, bizarrely, the words are reversed, but not the letters within the words! What's up with that???

But this has nothing to do with OpenType or Unicode or FreeFont as such.

Of course it is possible that somebody implemented something that assumes that Turkish has always been written in Latin script, but if that's the case, it sounds like a bug to me.

A bug in some software package is a very very poor reason to bring the Unicode consortium for inclusion of a glyph. I think they would ask you, why not just get the bug fixed in the software? It would certainly be far easier than changing Unicode.

I may have written to some people involved with these efforts over the years, but I really don't know exactly who is most involved in Arabic at this time. And as I said, I've never made such a proposal myself. You would be best off looking at the similar Unicode proposals, and writing to one of those people.

(file #29215, file #29216)

Steve White <Stevan_White>
Project AdministratorIn charge of this item.
Wed 25 Sep 2013 07:20:24 PM UTC, comment #18:

Hello Steve,

I understand what you're saying. However, in practice, implementation is such as to render your proposal impracticable. Even if I turn the switch RTL on, letters won't still get connected because Turkish is supposed (by fontspec) to be written in Latin Script. Only Arabic enables the contextual variants to be correctly displayed.

I'm still wondering whether you know someone or some forum where I can find people interested in forwarding a request for the addition of OTTOMAN GAF to Unicode.

Rosella Capriotti <rosy58>
Wed 25 Sep 2013 07:00:12 PM UTC, comment #17:


I'm afraid you're mixing up some issues.

The XeLaTeX/fontspec behavior is one issue. These are complex systems with a lot of switches. I suspect there is a LTR/RTL switch that will resolve the issue you mention.

As far as OpenType and Unicode are concerned, no, Turkish language is not "deemed" to be written in Latin. In fact, it is very much assumed to be written in Latin or Arabic script.

Again, as far as those standards are concerned, language and script are independent properties of text. In Unicode, the script is determined by the block of Unicode the letters come from. The OpenType language tag is prescribed by the document or application, and may turn on features for text with that language.

Steve White <Stevan_White>
Project AdministratorIn charge of this item.
Wed 25 Sep 2013 05:56:46 PM UTC, comment #16:

Hello Steve,

Problem is, you can't apply the "Turkish" tag to text written in a complex script (like the Arabic script) in LO, because Turkish is deemed to be written in Latin script. In XeLaTeX you can do so but it entails switching the text flow from right-to-left to left-to-right; it also hinders contextual forms, which basically means letters aren't connected to each other anymore and all of them appear in their isolated state. In a word, a disaster.

Rosella Capriotti <rosy58>
Tue 24 Sep 2013 02:49:06 PM UTC, comment #15:


I can't parse your first sentence.

There may be some confusion there. The language and script of OpenType features are in principle independent. So Turkish language tags are meaningful for both Arabic script and Latin.

As far as the standards are concerned, Turkish may be written in Arabic or in Latin script (or for that matter, in any script -- there just isn't any precedent for it).

Private code points are not in the scope of the FreeFont project of course. Even if the project were not Unicode-based from the start, a private code point by definition isn't standard. So users would have to at least know where it is to make use of it.

How on earth would that be better than using a feature to activate the desired glyph?

Steve White <Stevan_White>
Project AdministratorIn charge of this item.
Tue 24 Sep 2013 11:51:21 AM UTC, comment #14:

Because Turkish is written ONLY in Latin Script and does NOT make allowance for another script, like Serbian, which can be written either in Cyrillic or Latin.

The best workaround I can think of would be to use some private code point to encode OTTOMAN GAF.

Rosella Capriotti <rosy58>
Mon 23 Sep 2013 10:05:53 PM UTC, comment #13:

I thought we were talking about Arabic script here. Why do you bring Latin script into it?

There is indeed no OpenType language tag for Ottoman Turkish (just as there is no tag for Old English).

I'm not sure what would happen if you marked some text in a web page as 'lang="OTA"'. Probably they wouldn't recognize it. This is not the way things should be -- probably it should be passed down to the font rendering layers as Turkish. But this is a separate, more general question -- how to tell systems that the language is Ottoman Turkish.

None of this rules out my suggestion though.

Turkish language + Historical forms would work, provided it is OK to mark the text as Turkish (given there seems to be no better way to mark it.) The "Historical forms" would serve to indicate the ancient form of Turkish.

Steve White <Stevan_White>
Project AdministratorIn charge of this item.
Mon 23 Sep 2013 09:11:51 PM UTC, comment #12:

Dear Steve,

Problem is when you say "Turkish", we mean a very different language, with different morphology, syntax and lexicon, NOT mutually intelligible with "Ottoman Turkish", and what's more it's written in a Latin script, there's no 1on1 correspondence of letters like Latin Serbian/Cyrillic Serbian.

Rosella Capriotti <rosy58>
Mon 23 Sep 2013 07:23:59 PM UTC, comment #11:


I did not mean to suggest their policy was unbending. That is why I allowed there were many exceptions.

As I said, I could implement an alternate glyph in several ways. Just not with a 'lang' tag, which seemed natural. Maybe "Historical forms" + Turkish.

Steve White <Stevan_White>
Project AdministratorIn charge of this item.
Mon 23 Sep 2013 11:51:20 AM UTC, comment #10:

But IOTIFIED A was implemented in Unicode 5.1, so it's a recent development, obviously the rule is not as unbending as you suggest. Anyway, to implement an alternate glyph, you need a language tag which does not exist at the moment.

Rosella Capriotti <rosy58>
Sun 22 Sep 2013 11:02:02 PM UTC, comment #9:

>I'm going to answer you by this: by the same token, why didn't
> they put a Serbian BE and a Russian BE on Serbian keyboard?

They who? The designers of Persian typewriters?

What connection do you see between Persian keyboard design and Serbian? Of course, these standards arose in different countries, from different perceived needs, and were done by different people.

I think you're missing the point I have tried to make in my past two messages.

Some code points for the same letter are distinct in Unicode because they were distinct in past standards (for example, your Persian and Arabic gaf).

But for new code points, Unicode applies a criterion (*like this* I'm paraphrasing) that new code points should correspond to conceptually different letters not just letters that are shaped differently.

They are trying to avoid the open ended acceptance of any possible letter shape or style.

In the end politics may be involved. But I can tell you, it will be a weak argument, that Unicode should break the criterion in all cases because they broke it before.

Again, I still don't know what you think we can do for you here.

Steve White <Stevan_White>
Project AdministratorIn charge of this item.
Sun 22 Sep 2013 08:42:27 PM UTC, comment #8:

I'm going to answer you by this: by the same token, why didn't they put a Serbian BE and a Russian BE on Serbian keyboard?

So, yes, I'm convinced a codepoint is needed, but I don't know how to forward a request or start a group supporting the introduction of a separate codepoint, but I do think there's a precedent in Unicode: for a long time they used to turn down appeals to separate codepoint for CYRILLIC IOTIFIED A, they justified this by saying it was just an Old Slavonic variant glyph for CYRILLIC YA, but in the end they caved in.

Rosella Capriotti <rosy58>
Sun 22 Sep 2013 08:05:41 PM UTC, comment #7:


It may be puzzling. but the Persian keyboard is that way for practical reasons -- they often enough typed words from both languages, and needed to keep them distinct, I guess.
(With computers and OpenType this might have been done differently.)

Practically speaking, nobody is going to make a keyboard with an Ottoman gaf, right? Is anybody in need of a code point?

I still don't know what you are hoping we can do here.

Steve White <Stevan_White>
Project AdministratorIn charge of this item.
Sun 22 Sep 2013 07:38:48 PM UTC, comment #6:

My answer is: NO, because...

1. Ottoman Turkish doesn't have a language tag, as far as I can see from LO.
2. The Arabic range's logic, no matter how inconsistent, patchily and puzzling it may be, doesn't seem to make allowance for alternate glyphs: the Persian keyboard has two KAF's: one Arabic, the other strictly Persian, and two YEH's: one Arabic and the other strictly Persian.

Rosella Capriotti <rosy58>
Sun 22 Sep 2013 07:16:58 PM UTC, comment #5:


No I don't see what you're getting at.

First, in your image, do I understand rightly that the Ottoman letters are the ones with a dot below and a little curl on the right horizontal?
How does this curl connect on with a preceding character?

The Unicode standards are a complex thing. Some of it is an assemblage of older standards (such as ASCII and the main Arabic range you mention). As you pointed out to me in another discussion, the Arabic Presentation forms are from other older standards, that have been now encorporated into Unicode.

And as I complained elsewhere, the Greek range is one of the worst messes in Unicode. I don't know whom to blame for that. But... this is not the place to look for consistency.

They try to restrict new code points to conceptually distinct characters (whatever that means -- many fights ensue).

Again, the question is: can this letter be regarded as a style of the letter gaf?

Steve White <Stevan_White>
Project AdministratorIn charge of this item.
Sun 22 Sep 2013 06:19:12 PM UTC, comment #4:

Hi Steve!

Problem is, the boundary between letter shapes and actual characters appears to be different in the Arabic range, compared to the Cyrillic or Greek one.

For instance, U+06A9 KEHEH is just the Persian localised glyph for U+0643 KAF, U+06A2 FEH WITH DOT MOVED BELOW and U+06A7 QAF WITH DOT ABOVE are just Maghribi localised glyphs for U+0641 FEH and U+0642 QAF. Furthermore, there is a whole range of Persian numerals, U+06F0 to U+06F9 which again, are just localised glyphs, not characters in the strict sense.

I hope you get my drift.

Rosella Capriotti <rosy58>
Sun 22 Sep 2013 11:35:46 AM UTC, comment #3:

Not sure what you want from us Rosella.

As I said, if the letter is simply a different shape, it is possible to replace the glyph using a font feature.

But you indicated that you thought the letter should be a Unicode code point. The Unicode consortium is a big international concern, with the usual procedures and politics. I can only point you to possible contacts.

If it's just a style of the letter, we will need an image of the letter, and some supporting evidence for it, and then we can consider doing something in the font.

Steve White <Stevan_White>
Project AdministratorIn charge of this item.
Sat 21 Sep 2013 07:19:12 PM UTC, comment #2:

Umpf... seeing that link has really dampened my pep! They require loads of papers, advocate speakers, conferences, adhoc meetings and what not :( I thought it was as easy as walking into here and saying "Hey, Steve, I need that feature, are you going to implement it for me?" :)

Rosella Capriotti <rosy58>
Sat 21 Sep 2013 06:56:56 PM UTC, comment #1:


Is the letter different just in shape, or is it considered a different letter?

If it's just a difference of shape, we could resolve it in the font. But there is no OpenType language tag for Ottoman Turkish, so that route isn't immediately open.

To really make a proposal to Unicode is going to be some work. I have never done it. It would be best to get in touch with somebody who has worked on the Arabic ranges, and ask them about how to proceed.

Recently I have been looking at some proposals listed in
You might getsome ideas there.

Steve White <Stevan_White>
Project AdministratorIn charge of this item.
Sat 21 Sep 2013 03:21:05 PM UTC, original submission:

OK, this isn't really about FreeFont, it's about encoding a particular character in UNICODE, I don't know where the request can be forwarded, so I posted it here, maybe you can be of help to me.

Ottoman Turkish requires its own GAF, different than the standard (Persian) GAF, encoded in U+06AF. The two differ in their final and isolated form, the Ottoman GAF behaving just like KAF but with two horizontal strokes on the right of its main body.

Now, the Amiri font, to my astonishment features the Ottoman GAF on that very code point, even though this is indeed incorrect. See attached image for a comparison.

Unfortunately Ottoman GAF, being a historical glyph, has not been encoded, I was wondering where and whether I can post a feature request to the Unicode Consortium.

Thank you very much for your cooperation!

Rosella Capriotti <rosy58>


(Note: upload size limit is set to 16384 kB, after insertion of the required escape characters.)

Attach Files:

Attached Files
file #29216:  Ottoman-Turk.lua.pdf added by Stevan_White (5KiB - application/force-download)


Depends on the following items: None found

Items that depend on this one: None found


Carbon-Copy List
  • -email is unavailable- added by khaledhosny (Posted a comment)
  • -email is unavailable- added by Stevan_White (Posted a comment)
  • -email is unavailable- added by rosy58 (Submitted the item)

    Do you think this task is very important?
    If so, you can add your encouragement to it.
    This task has 0 encouragements so far.

    Only logged-in users can vote.




    Follow 5 latest changes.

    Date Changed by Updated Field Previous Value => Replaced by
    2013-09-26 Stevan_White Attached File- => Added Ottoman-Turk.html, #29215
        Attached File- => Added Ottoman-Turk.lua.pdf, #29216
    2013-09-21 Stevan_White StatusNone => Need info
        Assigned toNone => Stevan_White
    2013-09-21 rosy58 Attached File- => Added Ottoman vs Persian GAF.png, #29192

    Back to the top

    Powered by Savane 3.4