silnrsi / font-gentium

Fonts for languages and writing systems that use Latin, Cyrillic, and Greek scripts
https://software.sil.org/gentium/
SIL Open Font License 1.1
74 stars 5 forks source link

Deseret Typography in Gentium #7

Closed sigilante closed 2 years ago

sigilante commented 2 years ago

Deseret Typography in Gentium

Motivation

The Deseret alphabet was produced as an English-language spelling reform from 1847 and used somewhat actively from 1854 until 1869. The 40-letter basic Deseret alphabet occupies U+10400–U+1044f.

Most Deseret typefaces are either sans-serif (Noto Sans Deseret, TuBeeRound) or heavy slab-serif types (QueenBee, Times Bee). As Deseret letterforms are not terribly different from other Gentium alphabets (Cyrillic, Latin, and Greek), their addition to Gentium is a natural extension of the font's mandate and will make for an easy-reading serif implementation for running copy.

Letterforms

Deseret glyphs are Latin-like, and in a few instances overlap directly with coincident Latin letterforms, such as Deseret 𐐛 Eth with Latin L. In order to distinguish letterforms, I will always refer to a Deseret alphabet character with its symbol and name, e.g. 𐐃 Long Ah or 𐐄 Long O.

As with another type-first alphabet, Cherokee, Deseret possesses several features that would likely have been eroded with a long period of handwriting first. For instance, Deseret frequently requires rightward serifs, which are a bit awkward and heavy in serif type. Deseret was apparently visually conceived as being written with a horizontally-held pen, leading to difficulties in working with a conventionally-held (45Β°) pen and the serifs and letterforms. (For instance, the slash in 𐐆 Short I feels as if it runs the wrong direction from upper left to lower right; the letter pairs 𐐝 Es/𐐠 Zhee and 𐐀 En/π₯ Eng are difficult to distinguish from each other.)

Here is a comparison of "Twinkle, Twinkle, Little Star" in short samples from five other fonts and my current work in Gentium:

Screenshot_20220102_093737

Running text is much easier on the eyes in Gentium; the same text as running copy. (Note that I have not done any serious work on kerning yet, so please forgive the uneven spacing.)

Screenshot_20220102_095043

Similar to the Latin alphabet, structurally-related letterforms are not phonetically related to each other. Deseret letterforms can be divided into the following typographic series:

Series I:

Series II:

Series III:

Series IV:

Series V:

Series VI:

Series VII:

Series VIII:

Series IX:

Letterforms within a series must be sufficiently distinct to allow immediate visual discrimination in reading running text.

The most common modifier is a sort of hook or loop at the top of some glyphs. Unfortunately, due to the relative clutter of some letterforms (notably 𐐍 Ow), it is impossible to adopt a single glyph form as the loop modifier.

Another common modifier is the rightwards vertical, treated as a capital I in many typefaces. However, this makes it necessary to clearly distinguish the following three letters from each other:

(We set aside for now ligature glyphs and their variant forms.)

Font Design

Using Gentium

As Gentium began life with calligraphic letterforms, I have adapted the Deseret letters in a way as compatible as I think possible. For instance, consider the letters 𐐁 Long E and 𐐚 Vee; since Gentium has a properly-formed analogue in both U+0510 Cyrillic Capital Letter Reversed Ze and U+0190 Latin Capital Letter Open E, adjusted for the proper shape (rather than horizontally flipped), I can adopt either directly as 𐐁 Long E. In this case, because Deseret tends to be heavy in letterforms already, I adopt the lighter of the two, U+0190, which does not include the lower serif. For 𐐚 Vee, I adapt the U+a7ab Latin Capital Letter Reversed Open E which has the corresponding form.

Continuing in this way, I have hewed to using existing Gentium letterforms in producing Deseret letters. U+0277 Latin Small Letter Closed Omega, U+029a Latin Small Letter Closed Open E, and U+025e Latin Small Letter Closed Reversed Open E have been employed for some of the medium-complexity Deseret glyphs.

Resolving Series

The following entries are given as DESERET Sound (majuscule source/miniscule source).

Series I:

Deseret clearly conceives of a vertical line as sufficient to distinguish one letter from another, e.g. 𐐂 Long A/𐐒 Bee. This leaves us with some visually heavy majuscules (in particular 𐐒 Bee) but leaves pleasant room to work with ascenders and descenders in the miniscules.

Series II:

These are primarily architected around the letter O structurally.

Series III:

Although e.g. 𐐈 Short A is always vertical at the right-hand side in previous fonts, in handwriting it assumes a form much like a checkmark, motivating the tilted version here (and making for a less cluttered 𐐉 Short Ah).

Miniscules with a vertical component (such as 𐐆 Short I) use a small-capital block letter I as their basis, but could also use a dotless i.

Series IV:

The obvious thing for these is to base them on modifications of phi and psi. I added ascenders on the majuscules.

Series V:

Series VI:

These are perhaps the most whimsical letterforms and present the most challenge aside from the 𐐌 Ay/𐐐 H/𐐑 Er and 𐐝 Es. I was conservative on 𐐏 Yee and 𐐜 Thee (particularly as this latter is very commonly used as a standalone "the" in English) but opted for a whimsical connected descender on the miniscule 𐐖 Jee.

Series VII:

Series VIII:

𐐝 Es in particular is an uncharacteristically ornate character, which I embraced in the majuscule and subdued in the minuscule.

Series IX:

While miniscule 𐐀 En and π₯ Eng are distinguishable by a difference in ascender and rightward serif, the majuscules are likely too similar in the current proposal, and I welcome suggestions.

Miniscules

Many of the thorniest design problems involve the lower-case letters, which in conventional Deseret are simply smaller forms of the majuscules. This is aesthetically unsatisfactory and makes for difficult reading in long-form text. The introduction of ascenders and descenders goes against the initial intent of the designers (who wished to preserve metal type by not wearing out ascenders and descenders) but follows the lead of other Deseret font designers such as John Jenkins in the Deseret Alphabet Classics series.

Ligatures

The original Deseret alphabet specification from 1855 included two ligatures:

By 1859, these were replaced with two variant forms. (Discussion)

I have currently included the 1859 variants at the Unicode code points for each ligature. The options include:

  1. Revert to 1855 ligatures throughout. (The Unicode solution)
  2. Retain the 1859 ligatures and deviate in keeping with most Deseret practice. (My current solution)
  3. Adopt 1855 letterforms but keep 1859 ligatures as variant glyphs in a private-use area. (Seems the most Gentium solution)

I want the letterforms to look great first, then I'll fix kerning and maybe add a few fi-style ligatures. Once all design questions for the regular typeface have been resolved satisfactorily, I intend to work on the italic and bold variants in the same vein.

sigilante commented 2 years ago

Screenshot_20220102_112954

All letters in current implementation.

sigilante commented 2 years ago

To be more Gentium-compatible, I need to play with the ascenders more on e.g. Eth and Short O. There's an ascender on capital Bee to either lower or otherwise replace.

jvgaultney commented 2 years ago

Wow - you've done a lot of work on this. It does feel, however, like it's too much like Gentium and not enough like Deseret. Although the Deseret forms are distantly inspired by Latin (and Greek) the writing system is considerably different. Cherokee has a similar history.

Many of your capital forms are significantly different from the traditional ones and those in other modern fonts. The lowercase seems experimental. It looks and sounds like a proposal to change the writing system rather than a simple application of Gentium-style. That's good to consider, but should probably be done in cooperation with the Deseret community.

It would likely be more effective to create a unique Deseret font more loosely based on Gentium proportions, forms, and features, than to try to put together bits and pieces of Gentium glyphs to form Deseret ones, or to grab some extended Latin or Greek form and apply it unchanged. That has many advantages for you. You can get some of the benefits of Gentium (e.g. high readability and clarity) without trying to make it 'Gentium'. You can create something that is Gentium-compatible, and even use Gentium for the Latin glyphs, but is uniquely Deseret. A good example of this approach is Phoreus Cherokee by Mark Jamra.

For these reasons we would not be interested in supporting Deseret in the main Gentium project. Of course you would be very welcome to take Gentium and use it as the basis for your own Deseret font and release it separately (with a different name). That has many advantages for you. You could revise and improve it without needing to tie it in with our Gentium releases, and work on refining the lowercase proposal with the community before rushing to complete the other weights and styles.

sigilante commented 2 years ago

Fair enough. I'll aim for a standalone in this vein. Gentium is one of the most readable fonts so def. sticking with its inspiration here.