Re: unicode values for keyboard keys?

From: Erik Fortune (
Date: Thu Jul 31 1997 - 14:32:30 EDT

> Correct me if I'm wrong, but I believe the original poster was thinking of
> sets of symbols commonly used on keycaps (engravings), not some hardware
> independent key identifier approach (keysyms). My apologies for not being
> literal enough, but I assumed that this point was obvious. But, as we all
> know, on the Internet, assumptions can kill :-)

I didn't save the earlier messages, but I believe that (at least part of)
the discussion was about representing keystrokes in plain text, which is
not the same thing.
> X11 keysyms are not very useful for representing keyboards in documentation,
> hardcopy or on-line. All the keysym strings are encoded in 8859-1 and I'm
> sure the Japanese would be annoyed (I know I would be) if "Henkan" keys had
> to be documented in Romaji all the time because of keysym string limitations.

Where on earth did you get the idea that I wanted to use the ASCII names
of the keysyms to display the keystrokes?

> In addition, it is really annoying to see keysym strings like "Multi_key"
> inscribed on a keyboard representation as opposed to the actual engraving on
> the keycap ("AltGraph" on my Sun).
If we're representing keystrokes in text with a single codepoint per
keystroke, all of:
        AltGr, AltGraph, Compose and Multi_key
should be represented by a single codepoint. Choosing between the various
possible appearances is a font issue.

> XKeycaps goes a quite a ways to starting a catalog of symbols that are
> actually used on physical keyboards; not complete, but an excellent
> beginning.

IMO, cataloging the ways that keycaps are drawn is the wrong approach.

I have about thirty keyboards in my office. The variety of symbols engraved
on keys is staggering; for example, most of the "Enter" keys have an arrow
of some kind, but the relative proportions of the vertical and horizontal
strokes vary widely. Do you really want to add code points for each of the
possible variations?

How about "menu" keys, which also have several renditions (ranging from
fairly detailed icons to the word "Menu")?

What about command keys? Is the Macintosh command key different than
the "Windows" logo keys on the Microsoft Natural keyboard?

If we're going to set aside an area to refer to keystrokes, we should
identify the keys by function, not appearance. It's certainly more Unicode
in spirit to do so.

>Erik> Xkeycaps might know how to display several different renditions for
>Erik> a single logical symbol, but the underlying X keysyms _must_ be
>Erik> defined either in keysymdef.h or in the XKeysymDB.
> Nat ürlich. But not the point.

Disagree. It is precisely the point. If we want to represent keystrokes
in Unicode, we should represent the keys by function, not by appearance.
If you're identifying keys by appearance, X Keysyms are the place to look.

If you just want to draw reasonably accurate pictures of keyboards, most
of the characters you need are already there. Some of them are even
identified as "keyboard symbols."

-- Erik
| Erik Fortune ( | |
| SGI Core I18n Architecture | Calm down! It's only ones and zeroes. |
| ++1 415 933 1922 | |
       All opinions are my own, so leave my employer out of this

This archive was generated by hypermail 2.1.2 : Tue Jul 10 2001 - 17:20:36 EDT