Junicode-font
Junicode-font copied to clipboard
psb1558
An idea for ⟨ĥ⟩; ligatures and fine details
Dear psb1558,
Thank you very much for the correction of the small details I proposed you, which you implemented so fast. You’re very gentle. 🙂
I’m now writing you again, for other esthetic suggestions on fine details.
- I’m currently writing a text about Esperanto, and I tried to typeset it in Junicode. I noticed that for
ĥ(which frequently poses a challenge to designers) you put the circumflex in the middle of the blank space on top of the letter:While common, sincerely speaking I’ve never found this shape attractive. To my eyes, it appears unbalanced in itself and not in harmony with the other circumflexed letters of the language:
Beyond the personal tastes, which are inevitably subjective, I think there’s a problem for legibility: the diacritic is so high and far from the main “body” of the letter that it can easily be overlooked, especially in linguistic texts that rely on many symbols, diacritics, abbreviations, etc. and so often are quite “dense”.
The rarer variant with the circumflex on the side of the ascender in my opinion solves both issues beautifully. See a comparison with Libertinus Serif:
For Junicode, if you like the idea it may need the circumflex on the h to be a little narrower. A fast mockup:
...But maybe it could work also if it was instead wider (like it is now on top of the letter) and touched the ascender. 🤔
2. When writing in IPA, for the phonemic stress I follow the rare but more precise convention of writing it not before the syllable (/afˈfa/) but directly on the vowel (/affa̍/) for the many languages in which syllabic division is not phonemic but only phonetic (for example Italian: menò /menɔ̍*/ [meˈnɔ]). For such usage, to avoid a collision or an excessive narrowness, it may be useful to have the top terminal of
f, ʃ and ʧ to “retract”, or the space between letters to enlarge, before i̍, as it happens for fí, ffí.
3. The same thing could be beautiful for
ʃí, ʃî. (I don’t know if and how it could be done for ʃï…).
4. In my very personal opinion,
fí and ffí would look better and be more coherent if the i was actually joined with the f(s), as it happens in fi and ffi: it is strange and not very pleasant that the addiction of an accent breaks the ligature (an i is still an i, with or without an accent). The same can be said of other diacritics. See a comparison with other typefaces:
Befter:
Barracas:
Marseille:
These are only ideas, personal preferences! And for fine details. I hope I'm not annoying. 🙂
Thank you for you work again.
Not everyone is happy with Libertinus Serif the way it is now ant I think it will eventually get an alternate.
c.f. https://github.com/alerque/libertinus/issues/329
Hi Alerque. 🙂 I read that discussion some time ago. For what I know of Esperanto, which I’ve been actively interested in for some years now (to give a measure, I’ve got a full shelf in my library only for Esperanto books) I think the matter has been summed adequately by Georgd:
There's no right or wrong on any of the three forms. They can all be seen throughout the history of Esperanto typesetting, though admittedly the present version is the least common. I'd generally refrain from categorising in right and wrong upon such questions as this is about design which is about aesthetics, a highly subjective matter.
And I wanted to let Psb1558 know that other design choices are possible and (IMHO) both esthetically and legibly preferable. As for Libertinus, alternates are good, but (in general) I suggest you don’t feel forced to change your (beautiful!) design choices just because of one over-aggressive commentator... 😛
Other fonts with the circumflex on the side of the ascender:
Libre Baskerville:
Caveat:
Shadows Into Light:
Indie Flowers:
Domine:
Frank Ruhl Libre:
Righteous:
Gruppo:
Libre Caslon Text:
And many others surely can be found.
Some other typefaces with the circumflex on the side, which I found just scrolling through Google Fonts:
BioRhyme, Titan One, Marck Script, Audiowide, Aclonica, Quintessential, Petit Formal Script, Caudex, Besley, Fondamento, Peralta, Ribeye, Charmonman, Glass Antiqua, Rum Raisin, Cantora One, Uncial Antiqua, Spicy Rice, Milonga, Margarine, Risque, New Rocker, Galindo, Plaster, Ribeye Marrow, Special Elite, Mouse Memoirs, Bruno Ace
ĥ: ss20 substitutes low-riding diacritics for some that normally sit high above letters like h and b. I have added the combining circumflex to that feature, so with ss20 you'll get, for the sequence h U+034F U+0302:
This will not work with the precomposed character U+0125, and you need U+034F to prevent the h+mark sequence being turned into U+0125.
f + lowercase letter + mark: This sequence will trigger substitution of narrow f for standard f. So you'll get:
and many more combinations will be affected.
Narrow version of esh U+0283: Good idea. I will do this.
Ligatures of f with i+mark: I am satisfied with the way Junicode is handling these sequences now.
These changes will be in the next release (whenever that comes along).
This will not work with the precomposed character U+0125
Why not? This doesn't sound like a valid technical limitation from the font engineering side of things.
Just because I haven't implemented it that way. Right now ss20 manages 18 combining marks. If I get it into the business of also managing composites like U+0125, it will grow to become one of the font's bigger maintenance headaches. This will sound selfish, but I'm fine with asking folks who want something a little unusual to do a little more to get it.
And then, Junicode's GSUB features are already so huge that they are causing overflow errors in Glyphs, so I can no longer test within the font editor--a major pain. I would love to put the brakes on the growth of this font, or even find a way to compress or cut back a little.
Okay, that's fair. It just sounded like you thought there was some reason you could not, not some reason you would not. "Won't" is fair, the "can't" just sounded wrong to me.
Still the trouble you describe sounds like a Glyphs problem more than anything else. I'm not a Glyphs user and know more about the backend how fonts are stored and used, but this shouldn't result in more work for a font designer, both the precomposed and combining sequences should be covered by the same feature set across the board. It sounds like something (likely Glyphs not you) has jinxed this and created two paths that need fiddling with even though they eventually merge and could have been changed at a single point.
Since fontmake compiles without complaining about overflow errors in GSUB, I do expect that I'm running into a bug in Glyphs. Last time I tried to report an overflow error (in GPOS) as a bug, though, I was told it was my fault for putting in too many anchors.
So I'm not going to report this one as a bug.
I was told it was my fault for putting in too many anchors.
How many anchors do you have? You aren't putting unique anchors for each possible diacritic are you?
How many anchors do you have? You aren't putting unique anchors for each possible diacritic are you?
No, but in addition to the usual collection (the things Glyphs adds automatically), U+1DD2 combining -us has its own anchor (since I eccentrically insist that it must be positioned differently from other marks), and U+1DCE combining ogonek above, and some variants of U+035B combining zigzag. And then I have propagation of anchors in composites turned on, and lots of anchors on glyphs that aren't down in the Glyphs database as requiring them. I don't think there are so many as to cause problems, but rather something about my rather complex GPOS is triggering an obscure Glyphs bug.
In the last weeks I’ve bought or had scanned some Esperanto books or magazine articles (all of them about linguistics) for the text I’m preparing. Less than ten in total, yet in at least four of them I found the h with the low circumflex (just by chance: I’m not researching that). I’m posting here the pictures not to ask for any change in Junicode (I already tried the new solution and it displays perfectly: thank you very much again, @psb1558 🤗), but to provide useful examples to anybody who may find and read these discussions, whether they are type-designers or Esperantists, about this letter-shape.
Ebbe Vilborg, Esperantigo de grekaj nomoj, in Esperantologio, July 1961:
Tadeusz J. Michalski, Kiel esperantigi polajn geografiaĵojn, in Pola Esperantisto, January-April 1975:
Josef Kavka, Esperantigo de la ĉeĥoslovakaj loknomoj, in Scienca revuo, vol. 29, 1978:
Bernard Golden, Principoj por la esperantigo de hispanlingvaj vortoj kaj propraj nomoj, in Serta gratulatoria in honorem Juan Régulo, vol. II, 1987:
It's a classic problem in font design that a glyph for one language looks a little different from a glyph for another. One might say, Well, so what? but it means something to readers and writers of the affected languages, and so it has to be dealt with. OpenType has an excellent mechanism for dealing with the issue, namely to swap out one glyph for another when the language is this or that. Junicode does this for English, Irish, Turkish, Romanian, and a number of other languages.
Of course, that only works when there's substantial agreement about usage within a language community, and this is not the case with Esperanto. So you get the manual solution I provided.
With some glylphs, there's a preferred form, but printers compromise when the preferred form is absent from a font. I was wondering what old Zamenhof would prefer. In Fundamento de Esperanto (1905), one finds this:
But I can't tell where this copy was printed—not sure whether Zamenhof had much to do with the printing. In Dr. Esperanto's International Language, trans. R. H. Geohegan and printed in Warsaw 1889, so presumably under Zamenhof's supervision or influence, one finds this (p. 36):
But I couldn't find much more, at least in early prints that seem at all authoritative (lots of early copies were printed far from Warsaw).
It's quite possible that Zamenhof didn't care about this at all--after all, he seems to have had nothing to say about it. It would be interesing to get hold of some of his manuscripts, but I haven't managed that.
I think this is as complete as it's going to be (for now anyway), so I will close.