Brahmic Scripts

[Richard W]

The Uniscribe in your system32 (and my wife's) does not contain the string "dev2", thought it contains many other script tags. (Modern Uniscribe does contain it.) What other version of usp10.dll do you have on your machine? Extrapolating from Word 2002, I would expect you to also have C:\Program Files/Common Files/Microsoft Shared/Office11/USP10.DLL . I would have expected, but could be quite wrong, you to have a version of Uniscribe that came with VOLT.

As an added complication, Firefox for Windows comes with the HarfBuzz layout engine which, it appears, can be used to replace Uniscribe on a script (or script tag?) by script basis. The relevant parameter, accessible by entering about:config in the address bar, is gfx.font_rendering.harfbuzz.scripts, whose value is a bit mask. On our XP machine it is set to 7 (the default), and 23 ( = 16 | 7) and 87 ( = 64 | 16 | 7) have been used to get better Tibetan rendering. On Ubuntu it is set to -1, presumably meaning 'Use HarfBuzz for everything'.

Windows XP becomes wide open to viruses nest year (2014) when security updates cease. Do you have an upgrade plan yet? Would you like me to test your font on Windows 2002 running on Windows 7?

[Morrígan]

Just to interject, this is exactly the kind of problem that discouraged me from doing more work with fonts - there's a lot of effort that goes into getting the layout working, but implementations of Uniscribe were still somehow awful, and Adobe products had basically no support for complex scripts, despite the fact that Adobe was involved in setting up OpenType. That was like 5 years ago though, so I'm sure compatibility is better.

Also hinting was just too much of a headache... and the TrueType hints in Fontforge looked awful on Windows.

[clawgrip]

On a more pleasant note, here is another sample from my hack font of Tocharian:



It's the first two lines of the text used on the Omniglot page (I haven't completed all the glyphs necessary to complete the third line).

There are several stylistic differences between my font and the original manuscript.

[2+3 clusivity]

Looks great so far.

[clawgrip]

Thanks. It's tricky to line up the subscripts using negative-value characters, so I think eventually I'd prefer to make individual glyphs for each conjunct. The problem is that Tocharian seems to have so many conjuncts that defy expectations, like lnta, wla, rcwa, tkma, ypa. Word boundaries are ignored, so something like LAñcäl ṣeṢ\ śtwar dviPAntwā would be written LA-ñcä-lṣe-Ṣ\ śtwa-rdvi-PA-ntwā, introducing weird conjuncts like lṣe and rdvi, and ntwā. On the other hand, I think these unwieldy triple conjuncts are what give Tocharian that appearance of kind of ordered chaos that I like about it.

It also has subscript versions of independent u and i. Subscript i appears in the first character of the second line of that text. I'm not sure what the purpose of this is, but I guess it represents a diphthong. Making independent glyphs of everything except diacritics would take a lot of time (just figuring out what's even necessary would take a lot of time) but it would look cleaner in the end, which is what I want.

[2+3 clusivity]

Are most of the non-vowel conjuncts just vertical stacks of doubled consonants? I seem to remember south Indian scripts doing that as well as the link of khotanese.

[clawgrip]

Yes, they're all vertically, or more often diagonally stacked consonants. In theory it should be simple, but realistically it's of course much more complicated because the subscripts attach directly to the base consonant. Also, since consonants are aligned at the top (like Devanagari) rather than the bottom (like Latin), it's challenging to align a subscript with its base consonant perfectly, especially when you start getting into triple conjuncts and the u vowel signs, which attach on the bottom. Letters with descenders (ka, jha, and ra) have their descenders removed when subscripts are attached. Subscripts also have slightly altered forms so that they can connect smoothly with the consonant above them. Everything I said here is also true for Khotanese as well.

This is why I think just making individual glyphs for conjuncts is easier, since personally I feel that alignment via glyph contour editing is easier than alignment via OpenType coding. Then you only have to worry about the u vowel signs (and the subscript independent u and i letters for Tocharian, and the hook sign for Khotanese).

[Richard W]

My first thought is that the subscript forms should be treated as a kind of mark, and that you should use the mark2mark positioning to stack multiple subscripts. The vowels that are attached to the base consonants (i.e. those aligned with the baseline) can be positioned using mark2base positioning, with the subscript consonants being ignored. (A good Thai font ought to contain examples of this to parrot - more tonight if needed.) Explicit conjuncts can mop up most of the other forms.

[clawgrip]

That sounds like a good idea, as long as I ensure that each base character and each subscript has a place to connect. Tocharian really doesn't have any irregular conjuncts, but Khotanese does, but these can be treated separately.

What's the best tool for decompiling fonts to get the scripts? This is one of my main issues...I don't have much to work with.

[Jipí]

You can view GPOS tables and such in FontForge, as far as I know. However, I'm not familiar with the deep magic of fontmaking.

[clawgrip]

I downloaded and installed fontforge. Unfortunately, I think it decided based on my system that it should set the default language to Japanese. The problem is that it doesn't seem able to display Japanese, so most of the menus are just blank blocks. I have no idea how to change it to English since the menus are completely illegible!

[Morrígan]

I downloaded and installed fontforge. Unfortunately, I think it decided based on my system that it should set the default language to Japanese.

Unfortunate, and weird. I can tell you that the windows builds of Fontforge I've tried are incredibly unstable. FF is kind of a weird tool, but it's ability to handle OT features makes it invaluable. For a while, I was running a dedicated Debian virtual machine just for FF.

[clawgrip]

maybe someone can just tell me which blank spaces to click on so I can see opentype stuff!

[Morrígan]

maybe someone can just tell me which blank spaces to click on so I can see opentype stuff!

The dialogues are a bit complicated, but I'll give it a look. Also, FF reads your environment variables, so you might be able to tweak these for the language issue.
http://fontforge.org/cliargs.html#Environment

EDIT: For OT features, you need to go to Elements > Font Info (first option)

http://fontforge.org/elementmenu.html#Font

[clawgrip]

Thanks. Another person told me something similar. I changed some settings in the command prompt and ran it from there and now it's in English with normal menus.

[Morrígan]

Thanks. Another person told me something similar. I changed some settings in the command prompt and ran it from there and now it's in English with normal menus.

Cool. Pay close attention to the manual, because the FF is weird - it starts to make sense after a while, but it's OpenType support isn't always user-friendly

[clawgrip]

Thanks.

Of course, after all this, Unicode is still missing the following required signs:


Some of them you could work around, like making rra just ra + ra, but some of them I'm not so sure about.

Even though the Tocharian foreign consonants are equivalent to, and as far as I know interchangeable with the regular consonant + ä vowel, They don't always function that way, since they can sometimes form conjuncts, like K(Ä)TÄ, where the first Ä is silenced, resulting in KTÄ.

[Morrígan]

I'm not sure I have anything on this, but I'll have to check my archives. There was a time when I was really into Tocharian, so I'll see if I have anything interesting or informative lying around.

[clawgrip]

Neat, I'm surprised to hear that. I feel like these two scripts are not so well-known, and clear explanations on them are scarce. Khotanese in particular has very little presence on the web. If I put up a detailed explanation on how the script works, I am quite certain would be the first one on the Internet to do it.

And as far as I know, I'm the only person who has attempted to make a complete font of either of them.

[Morrígan]

There is Dunhuang:
http://idp.bl.uk/

Or for Khotanese.
http://idp.bl.uk/database/search_result ... ndom=32556

No sign lists, but lots of good document scans.

[clawgrip]

Yes, that's what I mean. No sign list or explanation of conjuncts, special signs, theorised phonetic values, etc. There is no website anywhere that explains how the script works. I might like to make one so there will be one for people who are interested in this kind of thing. Maybe submit something to Omniglot or something.

The scans on the Dunhuang Project website, in conjunction with the transliterations on the TITUS website, have been invaluable in helping me decipher the script. In fact, the images of jha in the first page of this thread come from a scan from an image on the Dunhuang website.

[Richard W]

That sounds like a good idea, as long as I ensure that each base character and each subscript has a place to connect. Tocharian really doesn't have any irregular conjuncts, but Khotanese does, but these can be treated separately.

At http://www.typophile.com/node/27988 (server currently down), John Hudson notes that abutting characters need to overlap so that they won't be adversely affected by greyscale rendering. Automatic hinting might have similar bad effects.

What's the best tool for decompiling fonts to get the scripts? This is one of my main issues...I don't have much to work with.

TTX will show you what's actually there in terms of data, but no pictures. TTX is close to reversible (except that it corrects errors), which makes it a very good tool for tweaking numbers and strings in a font.

[Morrígan]

So far, I've only found this - even if relevant, uses a very different writing style than you have, than appears in many manuscripts.


At least Dunhuang has some very clean images to base the forms of conjuncts on, even if the transliteration work is tedious.

[clawgrip]

Nice chart. Looks like that's where Omniglot and Ancientscripts got their images from. Central Asian Gupta may be the ancestor of both Tocharian and Khotanese. If I had no responsibilities I'd love to design a font of that as well, but I don't expect that to happen for quite some time, if ever.

Tedious is the word for it, since the Dunhuang and TITUS do not list their documents in the same way most of the time, so it is mostly hit and miss. I'm still trying to find an example of Khotanese independent syllabic r, because I know it exists but have no idea what it looks like. While Khotanese and Tocharian are similar, they are different enough that it's not possible to extrapolate Khotanese syllabic r from Tocharian, so I have to keep looking. Hopefully I'll find it. It's very rare. I guess that the long syllabic r has a vertical long a sign attached to it though, like Khotanese independent au and Tocharian independent long i.

But thanks to Dunhuang, I've actually learned to differentiate two somewhat distinct styles of Khotanese, a rounder one and an angular one. I believe the angular one is the later form of the script. If I ever get a proper Khotanese font working, I might eventually make an angular version as well (since the font I'm making is of the more rounded variety) and apply the same tables to it.

[clawgrip]

It seems like I only need to design between one and six variations of each consonant glyph in order to have all the parts I need to make any conjunct.