|
This change came up after discussions in:
https://github.com/google/mozc/issues/552
This change is not complex, but it has somewhat complex background. The
following discussion describes it extensively.
1. Six keys somewhat related or unrelated
This change involves six keys: ImeOn, ImeOff, Hangul Latin toggle,
Hangul to Hanja conversion, Eisu toggle / Caps Lock and
Katakana / Hiragana / Romaji.
ImeOn is the name used by Microsoft. Apple calls it "Kana switch".
ImeOff is the name used by Microsoft. Apple calls it "alphanumeric".
Hangul Latin toggle is the name used in keycodes directory. Apple calls
it "Hangul/English toggle".
Hangul to Hanja conversion is the name used in keycodes directory. Apple
calls it "Hanja conversion".
ImeOn and Hangul Latin toggle are mapped to Lang1 in USB HID usage and
physically located right to space key.
ImeOff and Hangul to Hanja conversion are mapped to Lang2 in USB HID
usage and physically located left to space key.
Eisu toggle / Caps Lock and Katakana / Hiragana / Romaji has key top
prints somewhat similar to ImeOff (which says "Eisu") and ImeOn (which
says "Kana"), but they have different semantics for Japanese IMEs like
macOS Japanese Input Method, Microsoft Japanese IME, and Mozc have two
distinct states: IME ON/OFF state and input character type. Eisu_toggle
and Hiragana_Katakana are used to toggle the input character type where
ImeOn and ImeOff keys are used to IME ON/OFF state. Their physical
positions are also different. Eisu toggle / Caps Lock are located where
Caps Lock usually sits on a conventional IBM PC keyboard. The position
of Hiragana / Katakana / Romaji are described in Microsoft's
documentation shown later.
The following pages describe the keys listed above:
https://docs.microsoft.com/en-us/windows-hardware/design/component-guidelines/keyboard-japan-ime
https://developer.apple.com/documentation/uikit/uikeyboardhidusage/uikeyboardhidusagekeyboardlang1
https://developer.apple.com/documentation/uikit/uikeyboardhidusage/uikeyboardhidusagekeyboardlang2
2. Messy situation before this change
The following key codes were defined for xfree86:
<EISU> = 210;
<KANA> = 209;
alias <HNGL> = <FK16>;
alias <HJCV> = <FK17>;
They used to represent ImeOff, ImeOn, Hangul Latin toggle, and Hangul to
Hanja conversion, respectively.
However, apparently HNGL and HJCV were incorrect. The following commits
of xf86-input-keyboard suggest they are actually mapped to 209 and 210:
425c1280439fe37497a33c47b5a8432e59cbfb76
ccf63a61f39e1f107a67c33d6a7ad24ea4c76b7e
evdev only has HNGL and HJCV, and they also represent ImeOn and ImeOff,
respectively.
EISU, KANA, HNGL, HNGL, and HJCV were then used to generate the
following symbols respectively:
Eisu_toggle, Hiragana_Katakana, Hangul, and Hangul_Hanja.
Generating Eisu_toggle for ImeOff and Hiragana_Katakana for ImeOn were
semantically incorrect as they rather provide the functionality of Eisu
toggle / Caps Lock and Hiragana / Katakana / Romaji.
3. Solution
This change solves the situation by always defining HNGL and HJCV key codes and
generating Hangul for Lang1 keys and Hangul_Hanja for Lang2 keys. In
this way, Japanese and Korean IMEs can know keys they need to toggle IME
states. Japanese IMEs can also distinguish ImeOn and ImeOff from
Eisu toggle / Caps Lock and Hiragana Katakana / Romaji.
This change lets Hangul and Hangul_Hanja symbols overload semantics of
Japanese and Korean IMEs. This is based on the following rationales:
* evdev always emitted Hangul and Hangul_Hanja symbols before this
change and therefore it would be least disruptive.
* The key pairs are physically located in the same positions so users
would expect them to perform interchangeably when switching Japanese
and Korean IMEs.
* Distinguishing from key codes is not possible anyway because the
underlying hardware complies USB HID Usage Table, which do not
distinguish those keys.
|
