Modifier Letter Rhotic Hook

Peter Baker

Here's a first go at redesigning 0264 (ram's horns) and 0264+02DE, following John Hudson's suggestion:

I think it needs more work . . .

John Hudson

I think the approach is sound, though. It seems to me the important elements of the rhotic hook are the downward curl to the right, of course, and the sharp corner connecting to the incoming stroke.

André G. Isaak

As someone who regularly works with IPA, that looks fine and unambiguous (assuming it's sitting on the baseline which I assume it is).

Peter Baker

Right. The big question was whether anyone would be confused by the disappearing right-hand horn, but I think not.

Peter Baker

André: yes, the same depth as the o. And I think the rest of Igor's set of vowels could be handled with much simpler V+02DE ligatures or with kerning.

André G. Isaak

Yes. Without context it looked like either a rhoticised ramshorn or an overly-wide gamma with some weird thingy attached to it :-) In context this won't be a problem (out of context the issue only arises because [ɣ] is a much more common phonetic symbol than [ɤ])

[Deleted User]

The user and all related content has been deleted.

Peter Baker

Here is my current set.

It differs somewhat from Igor's in that I include only current IPA symbols: several characters withdrawn from the IPA are omitted. As it turns out, though, it covers the entire IPA vowel chart.

Also, the assumption here is that all vowel + rhotic hook combinations except 025A (ɚ) and 025D (ɝ) will be entered as vowel + 02DE. For most, then, the sequence is simply kerned, while for ten of them a ligature has been made.

I'm happier with the hook glyph but not sure I'm there yet. Suggestions very welcome.

John Savard

I would favor something more conventional, like this crude drawing of mine shows:

Paul Miller

Peter Baker said:

Here is my current set.

It differs somewhat from Igor's in that I include only current IPA symbols: several characters withdrawn from the IPA are omitted. As it turns out, though, it covers the entire IPA vowel chart.

I have decided to re-instate the Rhotic Hook into my font and put anchors to attach it in the correct place for these letters.  So the question is how much of the IPA unicode blocks would it take to be useful without including the obsolete stuff ?

Which characters are obsolete or are not likely to be used ?

Peter Baker

These characters in the IPA extensions range are obsolete (except 029A, which is non-IPA):

0269 (ɩ), 0277 (ɷ), 027C (ɼ), 029A (ʚ, "non-IPA alternate for the preferred 0153 œ"), 029E (ʞ)

Of these, 0269, 2077, 029A are vowels.

This list is based on annotations in the Unicode code chart. I also checked against the official IPA charts, which are available in three different fonts that the curators of the IPA judge to be pretty good (they also have notes on some of the glyphs they find problematic in these fonts).

Another helpful resource is a webpage listing the Unicode values of everything in the IPA, assembled by a Phonetics Prof. at U Coll London back in 2012 (so it may miss a few more recent additions).

The other phonetic ranges in Unicode are (mostly?) not IPA, but uralic and perhaps other systems. I'm restricting myself to IPA for my current project, mainly because this is supposed to be a webfont, and I want to keep it as small as possible.

I'd say do some testing before you rely on anchors for positioning 02DE. It seemed to me when I tried it that some programs ignored those anchors--maybe because 02DE is not a combining mark? That's why I've decided to rely on kerning where possible and ligatures where kerning won't do.

Peter Baker

John Savard said:

I would favor something more conventional, like this crude drawing of mine shows:

A little like Igor's:

which is also very good. I prefer the hook attaching to the bowl rather than the upper stroke, mainly because the slant of the bar seems an essential part of the hook (John Hudson, above, mentions the "sharp corner"), and attaching lower down enables that.

Denis Moyogo Jacquerye

In a way the retroflex hook was nicer to connect to the base letter. The original Kenyon’s symbols were elegant.

@Igor Freiberger I would say your ɤ should have more curvy starting and finishing terminals. The IPA explicitly changed the shape of the symbol so it wouldn’t look like a baby-gamma (small ɣ) anymore to look like a ram’s horn instead in 1989. See https://dx.doi.org/10.1017/S0025100300003868

@John Hudson That’s an interesting ɤ˞. It looks odd at first but it’s already growing on me. I don’t know if it’s clear enough if it’s a rhotic hook or a variant of the horn terminal.

@Peter Baker  œ˞ could have its œ end like æ in æ˞, with the diagonal mid stroke one the e part.

I like the idea of the rhotic hook starting where the stroke of the base letter ends for glyphs whose finishing stroke ends near the median line. Maybe ɤ˞ could have that as well but keeping its right curly terminal, but then the whole think is not very compact.

Peter Baker

Denis: Are you thinking of something like this?

If I understand what you're saying about ɤ˞, I experimented with something a bit like it, but it appeared to sprawl, even if it wasn't really all that wide.

Paul Miller

Peter Baker said:

These characters in the IPA extensions range are obsolete (except 029A, which is non-IPA):

0269 (ɩ), 0277 (ɷ), 027C (ɼ), 029A (ʚ, "non-IPA alternate for the preferred 0153 œ"), 029E (ʞ)

Of these, 0269, 2077, 029A are vowels.

This list is based on annotations in the Unicode code chart. I also checked against the official IPA charts, which are available in three different fonts that the curators of the IPA judge to be pretty good (they also have notes on some of the glyphs they find problematic in these fonts).

Another helpful resource is a webpage listing the Unicode values of everything in the IPA, assembled by a Phonetics Prof. at U Coll London back in 2012 (so it may miss a few more recent additions).

The other phonetic ranges in Unicode are (mostly?) not IPA, but uralic and perhaps other systems. I'm restricting myself to IPA for my current project, mainly because this is supposed to be a webfont, and I want to keep it as small as possible.

I'd say do some testing before you rely on anchors for positioning 02DE. It seemed to me when I tried it that some programs ignored those anchors--maybe because 02DE is not a combining mark? That's why I've decided to rely on kerning where possible and ligatures where kerning won't do.

Thank you for your help.  These are good resources.

There is a problem with kerning also because there is a well known and widely used commercial word processor which switches kerning off by default and makes the dialog to switch it on hard to find.  I wonder if Cursive Anchors are the answer.

Peter Baker

I forgot that well-known word processor. Maybe because I would rather have my fingernails pulled out than actually use it. Got to experiment, I suppose, and figure out what's the best of a number of imperfect options.

Khaled Hosny

I don't think Word disables kerning in particular, it does not enable any OpenType layout at all for the set of scripts it considered "simple", so any other OpenType feature is as good as kerning here.

On the other hand, it feels like kerning is not a good fit for such an essential feature as there might be other legitimate situations where one might want to disable kerning but not break combinations like these, so another more "required" feature tag might be better.

Peter Baker

An experiment (thanks for the idea, Khaled!): I moved my ligatures involving 02DE from liga to rlig. The result, in Word (for the Mac), was that those ligatures showed up if kerning was turned on. I remember reading somewhere that no OpenType features are applied in Word unless kerning is turned on. This is not exactly right: mark positioning works in any case. But it is true of rlig.

Craig Eliason

My impulse would be to look to the cedilla in the font as guidance, particularly with regard to contrast. 

Peter Baker

Another little experiment: I added an anchor class for attaching 02DE relative to some problematic glyphs and tried it out in Word and Firefox. Both those apps ignored the anchors. If anyone can get this working please report--but when I tested SIL Doulos, which takes this approach, I got the same result.

Reclassifying 02DE as a combining mark might work, but it can't be non-spacing, and the idea is frankly terrifying (e.g. would it fail font validation? would there be unintended side effects?)

Peter Baker

Craig Eliason said:

My impulse would be to look to the cedilla in the font as guidance, particularly with regard to contrast. 

Here are some different approaches.

Times and Gentium have weight similar to that of the cedilla (Gentium seems especially well done here), while Brill and Doulos do not. To my eye, 02DE in Doulos looks too light. The IPA Kiel font has some contrast, but not nearly as much as in some marks (but I haven't seen the cedilla for that font).

Paul Miller

I have just downloaded Doulos to take a look.  The Rhotic Hook seems to piggy back on the 'Hook' anchor but in a strange position, however it works.  There also seems to be a lot of Cursive Anchor kerning for attaching various sloped line segments together which I believe is something to do with IPA.  Are the line segments really necessary for implementation of IPA ? it seems like a lot of work.

Khaled Hosny

Peter Baker said:

Another little experiment: I added an anchor class for attaching 02DE relative to some problematic glyphs and tried it out in Word and Firefox. Both those apps ignored the anchors. If anyone can get this working please report--but when I tested SIL Doulos, which takes this approach, I got the same result.

Since U+02DE is not a mark glyph, you will need cursive anchors not mark anchors, since the later applies only to base+mark combinations.

Reclassifying 02DE as a combining mark might work, but it can't be non-spacing, and the idea is frankly terrifying (e.g. would it fail font validation? would there be unintended side effects?)

Classifying it as mark glyph will cause some layout engines (at least Uniscribe and HarfBuzz) to zero its width, so it will not add to the advance width of the base glyph.

Peter Baker

Is it okay to use cursive anchors in Latin script? That won't make trouble with any layout engines?

Khaled Hosny

It should be OK in theory, never tried it myself though.

Peter Baker

Cursive anchors appear to work (in my Latin font) in Firefox and LibreOffice (yay Harfbuzz!), but not in InDesign or Word (for Mac). So I'm afraid cursive anchors aren't the solution we're looking for.

So far, anyway, I have yet to see a better solution than a combination of kerning for combos that can be kerned, and ligatures (via rlig not liga, per Khaled) for those that can't.

Perhaps Word users savvy enough to input 02DE at all are also savvy enough to turn on kerning?

Claudio Piccinini

James Montalbano said:

Maybe 35 people in the world are interested in this. I've done it many times. It does not sell.

I assume there is already a number of professional fonts including IPA support, so I guess if one needs to typeset phonetics, he can choose a typeface which goes well with the one he's maybe using for main text. Brill is a great example.
If one cares enough to have the whole phonetic area covered, good, otherwise I would not worry so much over single IPA subsets, unless they're really needed.

Theunis de Jong

Peter Baker said:

Cursive anchors appear to work (in my Latin font) in Firefox and LibreOffice (yay Harfbuzz!), but not in InDesign […]

Even though Adobe is (or was) one of the driving forces behind OpenType and InDesign served as their flagship in supporting it, they still took a shortcut and made the default Paragraph Composer only support a handful of common Latin-based features. It would not surprise me if you have to use the World Ready Composer for your cursive attachments to work.

Peter Baker

Shifting to the World-Ready Paragraph (or, for that matter, Line) Composer has, alas, no effect. Cursive attachment points still do not work.

Isn't this odd? It seems to me that this feature would be extremely useful for the billion or so cursive Latin fonts that are out there--but it is in effect unavailable.

Khaled Hosny

What if the kerning was put under a required feature tag, say mark, would this make Word turn it on by default. Feature tags vs lookup types are arbitrary in OpenType, and apart from a handful of features that really require a specific lookup type, I tend to use tags based on how required or optional I want the lookup to be, regardless of its type.