Adobe ventures into open fonts [LWN.net]

Adobe ventures into open fonts

Posted Aug 8, 2012 15:11 UTC (Wed) by bronson (subscriber, #4806)
In reply to: Adobe ventures into open fonts by alankila
Parent article: Adobe ventures into open fonts

Just that scaling antialiased text is HARD. And italics have a lot more antialiasing.

Click on the big pic and you'll see they're very close. (if not, the contrast on your monitor might be too high)

to post comments

Adobe ventures into open fonts

Posted Aug 8, 2012 17:23 UTC (Wed) by jimparis (guest, #38647) [Link] (1 responses)

> Click on the big pic and you'll see they're very close.

If it looks wrong in the thumbnail but OK in the big picture, this could be a symptom of LWN generating the thumbnail without taking gamma into effect.
For way-too-much info, see http://www.4p8.com/eric.brasseur/gamma.html :)

Adobe ventures into open fonts

Posted Aug 9, 2012 1:18 UTC (Thu) by xorbe (guest, #3165) [Link]

Nah just look at the first two lines. Clearly the non-italics look notably bolder.

Adobe ventures into open fonts

Posted Aug 9, 2012 11:06 UTC (Thu) by alankila (guest, #47141) [Link] (4 responses)

Unfortunately antialiasing has effect of causing excess darkening if gamma-related effects are ignored. So the italics should appear fatter, not lighter, if the problem had something to do with antialiasing.

Adobe ventures into open fonts

Posted Aug 9, 2012 11:28 UTC (Thu) by alankila (guest, #47141) [Link] (3 responses)

And as it happens, I tested this out with webfonts. The results I got on OS X look a whole lot like the image on the article; it turns out that e.g. SourceSansPro-Light.ttf vs. SourceSansPro-LightIt.ttf just have a different weight, which is also apparent in the fact that the glyphs are clearly smaller resulting in more text fitting on screen in the italic version. (Stupid me. Had I thought about this a little more, I wouldn't have realized that this proves that the glyphs aren't same size and therefore also not the same weight.)

On the other hand, if you just tell browser to do "font-style: italic" on the ExtraLight ttf file, the glyph widths remain the same, and the font weight is probably almost the same. If italic is generated from the regular shape by shearing the glyph control points, then the shearing also distorts the weights a bit, but it doesn't bother me. In fact, I can't notice it by eye.

weight vs colour

Posted Aug 10, 2012 10:02 UTC (Fri) by pjm (guest, #2080) [Link] (2 responses)

Regarding shearing changing the weight of the font, I think this should be accompanied by noting that the overall typographic colour theoretically won't change: that is to say, the mathematical area that's black won't be affected by shear, any strokes that become thinner will also become longer.

So a shear-italicized large region of text shouldn't become any darker or lighter when glancing at a page or seen from the corner of one's eye, it should only change how thick the strokes look when directly reading the italicized text.

That's a theoretical argument, and assumes that typographic colour can be measured by a simple mathematical expression (proportion of area), and also ignores the effect of ink bleed on paper, or hinting or gamma issues on screen.

Does anyone know of a better objective measure of typographic colour ?

weight vs colour

Posted Aug 10, 2012 18:09 UTC (Fri) by alankila (guest, #47141) [Link] (1 responses)

That is an enlightened comment.

You are absolutely right that the shear does not actually change the average color of the glyph in its box, even if you would subjectively evaluate the width of the slanted line as thinner than the straight line.

Proper implementation of font blending gets gamma right, even if linux software that does it correctly is very scarce -- in fact nonexistent would be more accurate. As an aside, I was able to get sRGB surface support in the 0.27.2 release of pixman, though, so maybe if I make more noise about this people start to use sRGB surfaces when blending text...

weight vs colour

Posted Aug 16, 2012 22:25 UTC (Thu) by njs (subscriber, #40338) [Link]

How does using sRGB surfaces help? sRGB isn't linear-light... right? Or is it just that when a surface is labeled sRGB then that gives a good excuse to also turn on linear-light alpha composition as well?