“I can’t wake up one morning and say, ‘Screw the letter B,’” type designer Matthew Carter told me last year when I interviewed him for the Economist, just after he had received a MacArthur Foundation “genius” fellowship. Carter, arguably the leading living creator and adapter of fonts in the Western world, was talking about the limits of pushing legibility and readability.

I thought of his comment when a recent furor erupted over the new “house” font for Android 4.0 (Ice Cream Sandwich), called Roboto. Roboto is a bespoke sans-serif font, created by a Google employee and used throughout Android’s user interface (UI) as part of the larger user experience (UX) overhaul. The intent is to make Android more intuitive, cohesive, and fluid, and work better on a variety of screen sizes, especially tablets.

Roboto was almost immediately branded a Frankenfont, a multi-headed hydra, and many other names by font purists and tyros alike, because of what seems to be a borrowing of identifiable features of several well-known fonts, including Helvetica. Stephen Coles at Typographica singled out characters he felt quite similar in form from Helvetica, Myriad, Universe, FF DIN, and Ronnia.

I was swept up in this as well. I glanced at the font, looked at various comparisons, and thought: What a shame that the opportunity to create something new and distinctive was lost. Roboto seemed to draw largely from the same well that Helvetica came from. Which was an odd choice, given that Apple had opted first for Helvetica for its iOS devices, and later (in iOS 4 for Retina Display devices) for Helvetica Neue, a set of improvements on the original.

But the longer I looked at Roboto, the less it seemed to me as nearly derivative, despite commonalities with other fonts. The designer, Christian Robertson, wasn't working in a vacuum. His design, directed by UX chief Matias Duarte, has to react to the constraints and abilities of Android hardware—at all the various screen sizes it will be available—and expand on the ways in which the previous system font, Droid Sans (created by Ascender's Steve Matteson), met UI and developer needs.

Carter said last year, "All industrial designers, and I consider myself one, work within constraints. Architects have to build roofs that keep the rain out and so on. It's particularly severe in the case of type designers, because what we work with had its form essentially frozen way before there was even typography. The Latin alphabet hasn't changed in a very long time," said Carter. (Carter declined to comment on Roboto in particular, but gave me permission to quote generally from last year's interview.)

Duarte echoed this in an interview conducted a few weeks ago. He said, about constraints around developing interfaces and fonts for new media, that "The important thing is each of the new technologies creates new boundaries for new types of expression. There are new tradeoffs. For everything that is lost, there are new possibilities."

The Feel of a Hand in an Iron Glove

Roboto is a sans serif—more technically a grotesk face with straight sides. Duarte has a neat essay on Google+ in which he sketches out the history of major type styles and defines Roboto's position within it. It's a good read and not necessary to repeat here at the same length.

Google supplied me with the full family (so far) of 16 faces to examine: a regular and oblique (the sans serif name for a slanted type that's not drawn differently, as with italics) of Light, Thin, Condensed, Bold Condensed, Regular, Medium, Bold, and Black. This warms the cockles of my typographer's heart, because with many different weights of a typeface, you can use differentiation to signify importance or meaning without having to rely solely on placement, size, or other faces. (The sign of a bad design is typically the use of many different sizes and faces. Find a great design, and you'll find remarkable restraint. The exceptions, which are legion, break that rule and prove it at the same time.)

The versions the firm supplied have hinting, or cues applied to the mathematical outline of each symbol or glyph that improve the conversion of the curve into a bitmap. It's unclear how much hinting is used by Android's font rasterizer, as Robertson noted in a comment on the Frankenfont blog post at Typographica that Roboto won't look at good in "older Windows browsers" because of a lack of certain kinds of TrueType hinting. Rasterization can be a CPU time sink, although TrueType (as opposed to PostScript) was designed to optimize that rendering.

What you notice first is that the uppercase is much more compact than the Helveticas. Helvetica tries to explore the full roundness of capital letters, with more than a suggestion of a circle. Roboto is ovoid, and trimmer around the middle. Are the flatter verticals in the C, D, O, G, and Q, and rounded corners supposed to suggest the proportions of a mobile phone? That's entirely too literal a reading, I'm sure.

Some of the bloodymindedness of Helvetica is gone, too. The G in Helvetica that reminds me of Peter Griffin's face from The Family Guy is no Kirk Douglas in Roboto, where it has a pert little chin instead of that giant block. The Q's violent diagonal slash in Helvetica is just little stroke akimbo in Roboto.

The lowercase also appears more condensed in the regular weight compared to the same weight of the Helveticas—but there's a trick. I was comparing the fonts continuously side by side, and something bothered me. Then I realized: they have nearly the identical average metrics when set in lines of copy rather than looked at overlaid on one another. That is, for a given length of upper-and-lowercase text at the same point size, Roboto occupies almost exactly the same horizontal space as Helvetica Neue.

The reason is the additional spacing around the letters. It is slight, but it adds up, and the face is designed to have a little openness when viewing on screen. But that openness can't equate to a repetitive blandness. A typeface may not produce an even rhythm or the eye finds nothing to grasp onto, and the face may appear legible but be unreadable.

As Duarte notes in a Google+ post about the font,

“One of the potential drawbacks of a grotesk font is that the structured evenness of the type can make it more difficult to read. We started by softening up the lower case letters, and then experimented with opening up some of the glyphs to get a more diverse rhythm. We found that by adding a little more diversity to the lower case the font become more readable.

The designers did this by varying the angles in the lowercase at which strokes end on curved letters, not on the purely vertical strokes. This may seem subtle, but examine a few fonts close up, and you'll see these differentiations immediately. Helvetica, for instance, squares off horizontal all the terminal ends of vertical curves in a, c, e, s, and so on. The horizontal curves end in perpendicular squared ends in the t, f, r, and the little tail on the a.

Robertson writes about this in a comment added to the Typographica post by Stephen Coles, cited earlier:

“It has been the hard and fast rule for sans serif types that the a, c, e, g and s must agree as to their angle of exit. Interestingly, this is not the case for serif types, and certainly isn't true for any kind of handwriting. It is common for the lower case ‘e’ to be more open than the 'a' for example. If there is a single story 'g' it will often remain open, or even curve back the other way (up until it forms a two story g).

That's what makes Roboto stand out. I don't find it entirely successful, but as Gypsy Rose Lee is asserted to have said in the eponymous musical about her, "You've gotta have a gimmick." Roboto isn't a humanist san serif, like Optima (a font I adore, by Hermann Zapf), with tapered thicknesses in straight strokes. But it still manages to reference handwriting, and to have the homunculi in our brains pull the right levers, even though it's below the level of perception for the non-typophiliac.

This lets Roboto have the evenness and spacing needed for onscreen rasterization, while preserving a tiny bit of the feel of the hand that makes a typeface seem created by human beings, not automatons. Duarte said in our talk that Roboto tries to preserve the physical feel of a hand writing letters. It's there; subtle, but there. It's why I like the font after living with it. I worry that as fewer people write well or write at all by hand, that that sense of the motion of a stroke disappears entirely.

Genuine Artificial Personality

Roboto has to establish a new personality for Android, one that's a distinct break with the past as Google puts all its efforts behind the unified single-platform-fits-all 4.0. Droid Sans was distinctive, but perhaps too playful and not as suited towards the more extensive and elaborate use of type in Android 4.0. A new font signals from the top that the experience will be different. (Whether that experience is better or worse is a different matter.)

For fonts designed for screen reading, "there's always a contradictory set of requirements," type guru John Berry said in an interview. John is a friend, colleague, and mentor, and the former editor of influential type journal U&lc. He spent the last several years, until recently, in Microsoft's font group. "One is you want it to be completely plain, generic, get out of the way; and the other is you want it to be distinctive. And they are directly in conflict with one another."

Roboto pricks at your sense of the familiar at first, but then, like a person you see passing in a crowd that you believe is a friend, and then on fully facing realize is a stranger, the font asserts its own identity. Duarte describes picking up an Android 4.0 phone and seeing Roboto as: "There he is, that old friend—that new friend, really—without having such a strong character that it really hampers the ability to communicate." It's a tricky balance to achieve.

This is what made Apple's choice of Helvetica, and later Helvetica Neue, particularly odd for iOS: it is one of the best-known faces in the world, and produces an implicit recognition that has nothing to do with Apple nor the device. The choice of using an off-the-rack font can't be pecuniary, because development costs are relatively cheap, whether the type family is designed by the ubiquitous Matteson of Ascender (who has had his hands all over screen-oriented fonts in recent years) or an in-house staffer.

That's relative to all the rest of the costs that go into an operating system, or even just the massive time sink of the user-interface design component. For a perfection freak like Steve Jobs, the fact that he didn't demand a perfect font for the task defies my limited understanding of him. Maybe he thought Helvetica was perfect. He's wrong, but maybe he thought that. (The existence and use of Helvetica Neue in later devices is the refutation.)

This reminds me of a story my design teacher Alvin Eisenman told in the 1980s, when I was studying graphic design as an undergrad at Yale. Alvin said he and other designers were approached in the 1950s by Reader's Digest to develop a new face for the magazine. (Alvin was responsible for training oceans of designers, including many influential type designers and typographers.)

He couldn't specify new kinds of paper or ink, and the design had to be conservative in the consumption of ink. Any tiny cost decision in production was multiplied by a factor of tens of millions of copies. But the magazine was willing to have large quantities of test type, cut in metal for machine setting, to get the right fit. Google has clearly chosen the Reader's Digest route; Apple tied its star to all of the connotations that arise from Helvetica. (Apple once also did terrible things to ITC Garamond.)

In Your Hands

Android 4.0 has to run on a variety of device resolutions, from the low 100s of ppi to well over 300 ppi. It needed a face that holds up at the lowest density, but also looks terrific the more pixels you throw at it in the same visual territory. The face has to almost have hidden richness, so that it is bland and readable at low density, and interesting (but not too much so) at higher density.

Further, Duarte noted, and you can see when you compare Android 2.x with 4.0, that the decision was made to use type rather than other elements, like symbols and icons. Images don't resize well unless they're vector art, which requires more time and effort to make work at varying sizes, and more computational power to render. Type is a simpler problem, already optimized, and which can be just as meaningful when small or large.

The first natively installed 4.0 phone, the designed-for-Google Galaxy Nexus, finally shipped December 14th, but 4.0 updates for older devices and other new hardware built for 4.0 may not appear until months into 2012. Those with some moxie can download and install Ice Cream Sandwich on existing hardware, too.

The proof will be in the device. All my talk in this article doesn't bring you much closer to knowing how Roboto on an Android phone, ereader, or tablet will hold up. The best type disappears as it fulfills its purpose. Google had a change to signal, and Duarte said, "We wanted it to be something designers could talk about." Roboto has surely achieved that goal.

(Thanks to Grant Paul for Android 4.0 screen captures!)

Tweet

On Twitter: Boing Boing & Glenn Fleishman.
Read more in Typography & Books.

Apple is a cipher, and its reasons for making changes often a mystery. A couple of days ago, the company updated its iBooks software for iOS devices to version 1.5, and added a de-skeuomorphizing full-screen mode (making the page similar to a Kindle display), a night-time reading color theme, and nicer covers for free, public-domain books. The release notes mention four new fonts, all superb choices, but avoid the fact that three less-loved fonts were removed.

iBooks shipped for the iPad in 2010 with five font choices: Baskerville, Cochin, Palatino, Times New Roman, and Verdana. When the small-screen version for the iPod touch and iPhone appeared, so did Georgia in iBooks 1.1. Few of these choices made sense as screen-reading fonts, even when Apple's densified its small screens with "retina" displays with four times the pixels in the same area.

Yes, I'm a font snob. And if you go down into my basement, you'll find a shelf full of monographs on Hermann Zapf, Jan Tschichold, and other others. But I'm not a snob about only choosing fonts from particular designers. Rather, choosing the right font for the right task. Apple seemingly tasked an intern working on a degree in graphic design for offset printing to pick the random assortment in iBooks. It's not that they are bad; on the contrary. They are mostly maladroit. Faces read on a screen need to have the right proportions and nature to work within the constraints and particulars of that medium.

Cochin (adapted and expanded upon by Matthew Carter) is too decorative for this purpose. It has a beautiful and slightly eccentric italic that I love, and have used on projects in print, but which is illegible at otherwise readable sizes on screen. Times New Roman (attributed to Stanley Morison) is crabbed, and meant to work on cheap paper at small sizes. It's only a modern standby because of the historical accident of Apple choosing it for early LaserWriter printers. Verdana (Carter) is a solid Web font, but wider than appropriate for portrait views, and not intended for this sort of reading. Baskerville (a classic face) was absurd on screen: it's a subtle collection of thicks, thins, and curves that don't read on a display.

I've always liked reading type in Georgia, however, which, like Verdana, was designed by Carter for Microsoft as part of the first Web-native screen font set. (Carter started with bitmaps and then drew outlines for both faces. He then worked closely with an expert in hinting, the art of fitting curves to bitmaps, to ensure a pixel-perfect fit.) Palatino (Hermann Zapf) is also acceptable in this version; it was also an early LaserWriter font, so it brings back some happy memories. It's regular enough to work.

The release of iBooks 1.5 offers an interesting swap out. My three least favorite fonts for reading on screen were removed: Baskerville, Cochin, and Verdana have been erased from the list. Only the dread Times New Roman remains alongside Georgia and Palatino. Added into the mix are four other faces: Athelas, Charter, Iowan, and Seravek. Only one of these I was familiar with. (Charter is from Carter, so he lost Verdana and Cochin which puts him down only one, if you're keeping score. I kid, as of the four faces in past and current iBooks, he might receive royalties only on Cochin, and then potentially just as a one-time payment.)

Since you're reading Boing Boing, I don't have to tell you that Athelas is named after what the common folk in Middle Earth called "kingsfoil," a healing herb when crushed and attended to by a true king of Númenor. Anyway. It's a gorgeous and relatively recently designed face, the winner of a couple of significant awards in 2006 and 2008, and holds up well onscreen, despite its elegance in print.¹

Charter is 25 years old, and one of the early non-Adobe faces that was designed to work on relatively low-resolution laser printers. The 300 ppi density of a laser printer is coincidentally close to the current highest screen densities on smartphones from many makers. Perhaps not a coincidence. Charter "sits big on the body," as we snobs like to say, which means that its x height (the vertical dimension from baseline to the top of a lowercase x) is quite close to the full capital height. This makes a face seem larger at any given numeric size (measured in the archaic unit of points, 72 to an inch) than comparable fonts that have more balance between capitals and lowercase. Bitstream (a type foundry co-founded by Carter) donated the face in 1992 to the X Consortium.

Iowan, a 1990 Bitstream foundry face designed by a sign painter and letterer, has never been on my radar. My friend John D. Berry explains in a 2001 essay perhaps why that's so. Iowan was released at a time when type sophistication was on the rise in the desktop-publishing world, and the font wasn't fully fleshed out until 2000 with old-style (also called upper-and-lowercase) figures, and other doodads that print designers like to create harmonious designs. It seems an odd choice for a screen face, but I have to say it works. It's also big on the body, and has both thick-enough strokes and enough visual interest (the slanting strokes on the tops of the lowercase serifs) to make it easy to read over long passages.

I like Seravek, the only sans serif added, because it's quirky. It has that nifty uplift on the lower-case L, which provides a little extra horizontal space than a traditional straight vertical, useful in screen reading. It has something in common with Gill Sans, although shed of the thins and thicks and super-quirks in Gill. It's new enough, released in 2007, to work in print and on high-resolution displays as well.

Of course, the ultimate solution to fonts in ebook readers and ereading software is to allow embeddable and downloadable fonts. Let readers choose the fonts they want to use from the sets of free options from Google, Microsoft, and others, and let publishers include as an option the typefaces that they believe best suit a book's design.

Licensing is part of the problem. If a print designer distributes a PDF with embedded fonts, most font licenses (for for-fee fonts) encompass this use, because the font is an integral part of the PDF. The EPUB format documents used in most ebook readers and apps (except Amazon, which uses MOBI, and is soon moving to some HTML5-like solution) is an XML specification, and is more akin to a Web page.

Many services now offer live Web fonts referenced into a Web page using a combination of JavaScript and CSS. One could expect this to happen with ebooks, as well, allowing the return of more sophisticated typography to this medium. As someone who has worked professionally through at least three revolutions in type and typesetting (optical, DTP, and Web/ebook), I wait impatiently each time for technology to catch up with the book arts. (Correction: iBooks does allow fonts to be packaged and referenced, but it's a little convoluted, and licensing remains an issue in any case.)

At least in making these font trades in iBooks 1.5, Apple has somehow empowered some group within the company to make more appropriate decisions regarding type. If we're lucky, that power will spread further, and we can regain a richer typographic history in modern clothes.

Posted Dec. 11, 2011

Tweet

On Twitter: Boing Boing & Glenn Fleishman.
Read more in Typography, Gadgets, & Books.