The Raster Tragedy at Low-Resolution Revisited:
Opportunities and Challenges beyond “Delta-Hinting”

Prima facie it would seem that turning outline fonts into pixels is a straightforward if not trivial problem: The outlines are blessed by the designer, scaling the outlines is mathematically exact, and “turning on” interior pixels follows strict rules. In theory, this doesn’t sound like it requires any rocket science.

In practice, however, we are still rendering fonts on the wrong side of the Nyquist limit, regardless of the anti-aliasing methods discussed in , , and . Therefore, “hinting” doesn’t go away—we merely get to use it for more interesting purposes than “pixel-popping,” such as the opportunities illustrated in , , and .

Given the resolutions of 96 to 120 DPI on today’s desktop or laptop screens, I can not single out a combination of rendering method and “hinting” strategy (cf eg the table in ) that, simultaneously,

But “hinting” can become more flexible to adapt to layout scenarios (cf ) and respect end-user preferences (cf ).

It may come as a surprise that the type designer’s intent is not readily encoded in the outline font format—certainly not in the TrueType format. TrueType outlines are partitioned lists of control points along with flags making them “on-” or “off-curve” points. But that’s just about it: there are no explicit concepts of stems, crossbars, or serifs, let alone concepts like “positioning crossbars at the ‘visual center’ between the baseline and the cap height.”

Without this information, a particular “hinting” strategy may or may not reflect the type designer’s intent. For instance, while I think my strategy to position the crossbar reflects Bodoni’s intent (cf ), I can’t prove his consent. The best I can do is to separate the strategy’s implementation from its use by encapsulating it in its own TrueType function. If need should be, this empowers the font to change the “hinting” strategy without “re-hinting” the characters (cf ).

With all of the above in mind, most of the “Raster Wars” I read about in the blogosphere become supremely futile fights in cyberspace. Really! What’s the point? Some people like broccoli, some people don’t—however healthy it may be. People’s tastes vary, be it in cuisine or in typography. But if done properly, “hinting” can cater to the varying tastes in font rendering (cf and ). I’ll call this making peace by “hinting.”

As screen resolutions (DPI) increase, raster tragedies become “less tragic.” At about 200 DPI, “typical” vertical strokes no longer require trade-offs between positioning and rendering contrast (cf ) even with text sizes as small as 8 pt. This helps character proportions and inter-character spacing (cf ), particularly when combined with fractional pixel positioning (cf ).

Current smart phones may have resolutions of around 300 DPI, but by the above logic this is not enough for all fonts: For pronounced stroke design contrasts (cf ), horizontal strokes are a lot thinner than vertical strokes which requires a commensurately higher DPI. At the same time these phones often use the small type sizes from the ink-and-paper phone book era to cram more information onto their 3½″ screens. This may defeat the advantage of 300 DPI over 200 DPI.

Whether or not you need “hinting” at 300 DPI, despite all of today’s anti-aliasing, depends on your typophile standards. L’appétit vient en mangeant: the more resolution you have, the more you want your standards satisfied on-screen. Instead of going away, 200 or 300 DPI can extend advanced “hinting” to more sophisticated opportunities like Optical Scaling and other aspects of Micro-Typography.

Back in 1990, when the first scalable font formats appeared on the market to render text in “black-and-white” on low resolution screens, “hinting” was a necessary evil. To turn scattered pixels into coherent—if pixilated—characters and make text somewhat readable, you had to use some form of “hinting.” Today I see “hinting” as an opportunity to get on-screen text rendering as close to the art of printing as the available screen technologies allow.

previous chapter, ↑ top, start page