Accumulated Feedback on PRI # 392

This page is a compilation of formal public feedback received so far. See Feedback for further information on this issue, how to discuss it, and how to provide feedback.

Date/Time: Mon Dec 3 09:32:52 CST 2018
Name: Christoph Päper
Report Type: Public Review Issue
Opt Subject: PRI392

A Short form for uniform skin tone
----------------------------------

I oppose the proposal to introduce and recommend a shortened form for
group/family emoji sequences where all members have the same skin tone.

Firstly, making it more convenient to have ethnically homogeneous families
is just asking to be criticized for being racist. That is certainly not
worth the bytes saved.

Secondly, I understand that the main reason for proposing variant "Short
Shared" (A2 in the PDF, where c) and d) have been mixed up) is that the
fallback mechanism would be better. This is, however, not really the case.

The "Status Quo" (A1) falls back to separate emoji people with skin tones
applied, unless skin tones are unsupported where it falls back to individual
people each followed by a replacement character.

"Short Shared (A2) falls back to a family ligature emoji followed by a free-
standing skin tone modifier, unless such ZWJ sequences are not supported
where it all falls back to separate people, the last one either being
followed by a replacement character or having a skin tone applied.

The family ZWJ sequences have been introduced with Emoji 2.0 in late 2015.
The skin tone modifiers, after having been encoded in Unicode 8.0, have been
introduced with Emoji 1.0 in mid 2015. Since skin tones came earlier and
color modifications are usually simpler to achieve than compositions of
multiple glyphs, it is more likely that an imperfect implementation of
UTS#51 would support skin tone modification but not composite families.
Therefore, A1 will more often provide an acceptable fallback than A2.

B Mixed skin tone for 3 multi-person emoji characters
-----------------------------------------------------

The Status Quo (B1) is unfortunate. However, only very few existing emojis
are affected. Some of them, i.e. 👪 U+1F46A FAMILY, 💏 U+1F48F KISS and 💑
U+1F491 COUPLE WITH HEART, can already be ZWJ-composed out of other emojis
(predominantly 👨 U+1F468 MAN and 👩 U+1F469 WOMAN). Implementations differ in
the number and gender of people shown, by the way. The three variants of two
people holding hands, i.e. 👫 U+1F46B, 👬 U+1F46C and 👭 U+1F46D, have been
proposed to get similar color-aware sequences, using 🤝 U+1F91D HANDSHAKE
instead of a heart as a combiner. This seems a viable path forward.

Accordingly and since using multiple skin tone modifiers with a single base
is, apparently, not considered a possible option, ZWJ sequences (option B2
or B3) for the few other cases seem appropriate indeed. Ideally, this would
lead to a situation where 👯 U+1F46F WOMAN WITH BUNNY EARS can default to
show as a single person again, as in Android before 7.0, Windows 10 before
the Anniversary Update and Facebook as well as some defunct sets still
today.

For 🤝 U+1F91D HANDSHAKE, I see another variant of "Decompose" B3, let
"Recompose" B4: Introduce a ZWJ sequence using an existing hand emoji, e.g.
🤚 U+1F91A RAISED BACK OF HAND (🤚🤚) or ✋ U+270B RAISED HAND (✋+✋) or a
combination thereof (🤚+✋ or ✋+🤚), alternatively 👍 U+1F44D THUMBS UP SIGN
(👍+👍) and (non-emoji) 🖒 U+1F592 REVERSED THUMBS UP SIGN (🖒+👍 or 👍+🖒). That
notwithstanding, I believe GIVING HAND / PALM FACING DOWN (as in Windows
shared folder icons) and TAKING HAND / PALM FACING UP emojis would be good
additions to the overall emoji set and would be the best candidates to use
for a sequence here. However, option B2 works as well and has better
backwards compatibility than B3, but not than some variants of B4.
Fortunately, Handshake does not need to support gender modification.

For 🤼 U+1F93C WRESTLERS, I think the only candidate for B4 would be 💪
U+1F4AA FLEXED BICEPS (i.e. recomposed 💪+💪 = 🤼), although it does not show a
full person and therefore does not accept gender modifiers yet. As such a
low-use emoji, I believe, the Doubled option B2 is perfectly fine. It is
very similar to 👯 U+1F46F WOMAN WITH BUNNY EARS, except that WRESTLERS was
always intended to show two people, and they should use the same approach
(B2, B3 or B4).

Another emoji that is showing two people, but is not listed on [Draft
Candidates](http://unicode.org/emoji/future/emoji-candidates.html)
currently, is 🤱 U+1F931 BREAST-FEEDING. The recommendation to hide all skin
of the baby is a fowl compromise which makes the emoji less clear than it
could be; many people mistake it for "woman *holding* baby" anyway (and
complain that there is no male variant). It could be recomposed, B4, to a
sequences of 👩 U+1F469 WOMAN (or 🧑 U+1F9D1 ADULT, and 👨 U+1F468 MAN if you
must) and 👶 U+1F476 BABY as well as 🍼 U+1F37C BABY BOTTLE in between perhaps
(👩+👶 or 👩+🍼+👶). There is no need for any new emoji (B3), but since unlike 🤼
U+1F93C and 👯 U+1F46F the two people are not the same height and posture,
Doubled B2 appears inappropriate here; you also could not tell whether the
first or the second one was the baby in a fallback situation.

In conclusion, I prefer recomposing (B4) to existing emojis in general,
decomposing (B3) to new emojis would not be needed at all, unless (for
Handshake) better hand gestures that are useful for other purposes became
available then. Doubling (B2) would be tolerable as well, and actually
preferable for Bunny Dancers and, a little less so, for Wrestlers.

Date/Time: Fri Jan 4 18:44:04 CST 2019
Name: Charlotte Buff
Report Type: Public Review Issue
Opt Subject: PRI #392: Feedback on Multi-Person Emoji

► A. Short form for uniform skin tone

I prefer the Status Quo option, that is: requiring a separate Fitzpatrick
modifier for each person in the sequence. Since their introduction, emoji
modifiers have consistently been treated as only affecting the character
immediately preceding them, not unlike variation selectors. Changing this so
that these modifiers can also apply to an entire ZWJ sequence at once, but
only under certain, rather uncommon circumstances, just to save a few bytes
each in sequences that will see hardly any usage or support anyway, does not
seem worth it. Every component in a ZWJ sequence should still be its own,
independent entity.

I also do believe there is value in allowing modified and unmodified emoji
within the same sequence. In fact, Microsoft’s font support for ZWJ families
already works that way. If modifiers only apply to the previous character,
procedural systems like the one used in the Segoe UI Emoji font can be
comparatively easy to implement because they simply need to consider each
character or modifier sequence in isolation and add them one by one to the
ligated glyph. The proposed “Short Shared” option would introduce all kinds
of exceptions into this well‐established algorithm that would necessitate
deliberately breaking many sequences that worked perfectly fine before, as
well as making some combinations behave counterintuitively and
inconsistently with the rest of the set.

A list of all family sequences currently supported by the Segoe font can be
created by simply iterating through every possible combination of family
members in order. There really is no easier way of doing this. An
alternative algorithm that only produces valid sequences recommended under
the “Short Shared” option would be a right mess.

Furthermore, the discussion of fallback display failed to include two
crucial possibilities: Rendering if no special ligating behaviour is
supported at all, and rendering if only Fitzpatrick modifiers are supported
but no ZWJ mechanism.

The former scenario should be the baseline for all emoji encoding decisions
in my personal opinion. A hypothetical system that does not support any
emoji sequences will see this under option 2:

👨⁠👩⁠👦⁠👧⁠🏾

Or, even worse, this:

👨⁠👩⁠👦⁠👧⁠⛝

Depending on how much the user in question knows about emoji mechanisms,
this would either be read as four people and a glitch, or three people with
indeterminate skin tone and one brown girl, but definitely not as an entire
family all sharing the same skin tone. Option 1 would circumvent this issue
since the modifier is explicit for every base character in the sequence:

👨⁠🏾⁠👩⁠🏾⁠👦⁠🏾⁠👧⁠🏾

Or:

👨⁠⛝⁠👩⁠⛝⁠👦⁠⛝⁠👧⁠⛝

Fitzpatrick modifiers are older than ZWJ sequences and have historically
enjoined greater acknowledgement by the core standard. If modifier sequences
are properly supported, but ZWJ sequences aren’t, the “Short Shared” option
is once again inadequate:

👨⁠👩⁠👦⁠👧🏾

This will never be read as a uniform family.

There is also the question of how generalised this “Short Shared” option is
intended to be. For example, what happens in cases where the Fitzpatrick
modifier is not the last character in the sequence? If the “Short Shared”
approach were to be implemented, how should the glyph for the following
sequence look if vendors choose to support it?

👨⁠[ZWJ]⁠👩⁠[ZWJ]⁠👦⁠[ZWJ]⁠👧⁠🏽⁠[ZWJ]⁠🦰

Would the skin tone still apply to the entire family or just the girl? Would
the hair component also be affected by the same mechanism? What happens if
we throw gender or professional modifiers into the mix? This whole endeavour
seems far too complex for its own good. While there is no real possibility
of these kinds of sequences ever becoming RGI considering the sheer quantity
of combinations, they are still perfectly valid constructs according to UTS
#51, so their precise behaviour must be formally defined.

Lastly, it is not entirely clear whether this “Short Shared” option would
only apply to families or to other kinds of sequences as well. For example,
would the sequence 👨⁠[ZWJ]⁠🤝⁠[ZWJ]⁠👩⁠🏼 produce a glyph that shows both
people with the same skin tone?

► B. Mixed skin tone for 3 multi‐person emoji characters

I fear that there very well may not be any elegant solution to this problem.

The status quo is the most obvious approach and also the least destructive,
but it may also make for some negative publicity. I don’t know how big the
demand for these mixed sequences is; I have seen some complaints concerning
HANDSHAKE in the past and probably also one or two tongue‐in‐cheek
discussions about the segregative nature of WRESTLERS. When it came to WOMAN
WITH BUNNY EARS, the discourse had always been about allowing Fitzpatrick
modifiers at all and never about mixed skin tones, at least in my
experience. Furthermore, there are some emoji whose mixed variants are most
likely never going to be RGI anyway (for example the family sequences), so
monochrome 🤝, 👯, and 🤼 wouldn’t really feel that much out of place in the
grand scheme of things either.

The “Doubled” model strikes me as the worst possible compromise. Calling the
fallback “slightly odd” is an understatement; it is incomprehensible. Four
wrestlers are not two wrestlers. As I have mentioned above, emoji sequences
should always be encoded based on the premise that nothing works as
intended. ZWJ sequences should still be understandable even if all the
joiners are removed; otherwise there is no reason for them to be sequences
in the first place. There is no logic by which two pairs of shaking hands
would be an appropriate substitute for one pair of shaking hands. This is
nothing but a bodge.

Nobody will understand what these sequences mean unless their emoji font
supports the latest updates *and* there aren’t any problems with the
rendering engine or image‐substituting algorithms. This is not good design
for character sequences and no other part of Unicode really works that way.
When ligatures for Devanagari or Tibetan break, the affected user will at
the very least see their individual components and won’t be confronted with
the glyph that was meant to be displayed in the first place, but
reduplicated in a different colour for no reason.

Encoding new characters seems to be the least bad solution. There is
actually no need for a ‘Person with Bunny Ears’ since WOMAN WITH BUNNY EARS
is already singular. A pair of dancing bunny people with differing skin
tones can be trivially encoded as 👯⁠🏻⁠[ZWJ]⁠👯⁠🏿, under the condition that
WOMAN WITH BUNNY EARS officially returns to being treated as if it actually
represented the thing it was encoded to represent all along. While it is
true that many current fonts idiosyncratically display 👯 as two people, it
has already been firmly established in the past that emoji names and
meanings can be changed at any time for any reason, so I see no issue with
WOMAN WITH BUNNY EARS becoming a single‐person emoji once again. The
semantics of the character in the context of how most people use it is
largely unaffected by the number of persons depicted.

It’s not that straightforward for the other two characters, though. I do
think there could be some potential value in an ‘open hand pointing
left/right’ emoji as its own entity, which could then additionally be used
for producing ZWJ handshakes. Encoding another wrestler to complement 🤼
seems rather wasteful, however. If this approach were to be taken, I would
recommend making this character an emoji component that isn’t meant to be
directly present on keyboards by itself. I’m also not sure how recognisable
a singular wrestler who isn’t enganged in a competition would be at standard
font sizes.

One possible solution that the UTC hasn’t considered yet is to allow
multiple Fitzpatrick modifiers per base character. A split handshake would
then be encoded as 🤝⁠🏼⁠🏾. This approach requires no new characters, has very
legible fallback display, and is also quite space‐efficient since no bytes
are wasted on Zero Width Joiners, but it also massively violates the
established semantics of these modifiers. Another issue is defining the
precise syntax of these sequences because 🤝⁠🏽⁠🏽 and 🤝⁠🏽 for example should
theoretically produce identical glyphs, so one or the other would need to be
exceptionally prohibited.