- From: John Hudson <tiro@tiro.com>
- Date: Sat, 11 May 2013 12:52:00 -0700
- CC: "www-style@w3.org" <www-style@w3.org>
On 10/05/13 9:15 PM, fantasai wrote: > So, somehow, we need to have a concept of the synthesized text's > size and baseline, and be keying the decorations off of that. If a font maker has built the font carefully, it will contain recommendations for super- and subscript scaling and positioning for synthesised results that correspond as closely as possible to the designed super- and subscript glyphs in the font. However, I suspect the number of fonts with reliable data of this kind is quite small: font tools provide default settings that many developers don't change, and the accurate settings are difficult to calculate. Reliable information on how to correctly set these values is hard to come by. [I need to do some further checking to confirm where in the TT/OT font structure this data is stored; I have not looked at it for a while.] There is a longstanding issue in text layout regarding interaction of real and synthesised styled text (not just super- and subscripts, but also smallcaps), and it seems to me that text decoration is simply another aspect of this. The general issue regards what should happen when an author wants to style text that includes characters that are not supported by the font's glyph substitution tables. So, for example, let's say that a font contains a full set of a-z superior letter glyphs, accessible via the {sups} feature, but an Icelandic author applies a superscript styling to text that includes the letter ð. What should be done? The options are: A) Use synthetic styling for the whole stylised string and ignore the glyph substitution feature (implies analysing the string, registering inclusion that is not included in {sups} feature); B) Use the substitution feature for all characters for which mappings are available, and use synthetic superscript for the ð character (best results if able to use font superscript scaling and positioning recommendation data and if that data is accurate; implies keeping track of which glyphs are substituted by feature and which are not); C) Use the substitution feature for all characters for which mappings are available, and fail in styling of ð character; D) Throw up hands in despair and use synthetic styling everywhere. If one takes option B as providing the best chance of optimal results, even though actual results will vary, then it seems to me that one could approach the question of text decoration along similar lines. That is, one could approach styled text, regardless of how it is arrived at, as if it were synthesised and apply text decoration accordingly. So, you would have a string of superscript glyphs resulting from a substitution feature, and you would apply text decoration at the height of synthesised superscripts. My guess is that the results would be sub-optimal unless making use of accurate font recommendations for synthesis, but that more often than not they would be in an acceptable range. Given than decorating of superscript and subscript styled text is likely to be pretty rare, is this a Good Enough solution? JH
Received on Saturday, 11 May 2013 19:52:52 UTC