On Mon, 3 Apr 2017 14:12:52 +0200 Michael Everson <[email protected]> wrote:
> On 2 Apr 2017, at 18:27, Richard Wordingham > <[email protected]> wrote: > I think you are seriously going the wrong way with this thinking. The > immediate parallel that comes to mind are things like: > > 1000 MYANMAR LETTER KA > ⁓ 1000 FE00 dotted form > > where the character can still be read if the variation selector’s > glyph can’t be shown. Uniform width is a feature of CJK, sure, but > that’s the nature of the writing system. Chess pieces for setting > withing in ordinary text do NOT have to be an em wide, and they don’t > in fonts. Chess pieces on a white square or on a black square do have > to have a uniform width in order to produced the board matrix. Nobody said the glyphs for use in ordinary text had to be a fixed width. What I am saying is that the glyphs for the two new variants you are proposing need to harmonise with the block elements such as U+2581 LOWER ONE EIGHTH BLOCK. That requires uniform width *for those variants*. That is a key part of the glyph family's essence. There is no such requirement on the glyphs for normal text use as at present. > > U+00A0 makes a lot of sense as the base character. > > What? NBSP and SP are whitespace characters, with complex behaviours, > and chessboards, whether set in lead type or digitally, are sets of > simple symbol glyphs. NBSP glues two things together. SP separates > things. Chessboards are not collections of black squares glued > together by white spaces with white spaces at the alternating ends of > lines. I reject this analysis. If one had a row of squares in flowing text, one would want the row to act like a word. One might have to resort to gluing it together using CGJ or WJ. > > Also having variants of U+25A1 and U+25A8 that match the game > > square filter modifiers seems quite legitimate. > > Um, wait… What are you proposing NBSP for? I'm confused now. If you > like these two characters (and I am glad you do) there’s no need for > U+00A0 at all. To be pedantic, I said that the proposed variants were legitimate, not that I liked them. > > Secondly, the mechanism can only look for a substitute if it knows > > that the glyph is missing. > The macOS does this quite reliably. If Baskerville has no chess > piece, but Ludus does, then a text in Baskerville wlll usually > display the Ludus glyph. You can override this by selecting the Ludus > gyph and forcing it back to Baskerville and then you get a box or > other substitution glyph. I'm talking about looking for a U+2654 glyph for ordinary text when all the first font tried has is: 2654 FE01; Chesspiece on white; # WHITE CHESS KING 2654 FE02; Chesspiece on black; # WHITE CHESS KING I must confess I am now wondering what the format 4 cmap should say about U+2654. Should it give a glyph for U+2654 or not? I'm also wondering about Windows behaviour. There was a time when Windows 7 only supported variation sequences if they appeared in the cmap 14 subtable. > > If it's looking for an OpenType font for a glyph of the family > > <U+82A6, U+E0100>, > > Or any OpenType substitution string. Most won't be recognised as needed. If the first font lacks a ligature for <f, i>, fallback won't be used for it. Grapheme clusters and variation sequences get special treatment. Richard.
