TUGboat, Volume 36 (2015), No.

2 95

Digital typography with Hermann Zapf

Peter Karow
Hermann Zapf was involved in many of the demands
for digital typefaces which came into existence from
1972 through 1997. These issues included formats,
variations, interpolation, rasterizing, hinting, and
grayscaling.
Within modern text composition, digital text is
a special part that should proceed without manual
assistance and human layout. Up to now, the mile-
stones were these: kerning, optical scaling, paragraph
composition (H-program), and chapter composition
(chapter fit).
Figure 1: The first version of Ikarus-Format as
started in 1972 (left). Contouring used to make
Digital typefaces Outline and Inline versions, shadowing and contouring
In 1972, creative typesetting professionals produced used to make Relief and Drop Shadow (right).
variations manually using their photographic experi-
ence and a phototypesetting machine to make fonts
with contours and shadows. Naturally, shadowing
as well as contouring were my next features, which I
started on 26 February 1973. This date might be re-
garded as the birth of “digital typefaces”: new forms
were generated automatically (see Figure 1).
My programs crashed quite often and generated
substandard looking results. But anyhow, I got it fly-
ing further and further. That influenced my decision
to call the software Ikarus [6] in 1974.
The first time I met Hermann Zapf was in the
summer of 1975. I had just completed the basic parts
of the Ikarus program which could already generate
“digital typefaces” in the form of contoured, slanted Figure 2: Typical desktop in 1972 with digitizer
and shadowed versions based on hand-digitized out- tablet, “mouse”, keyboard (below), direct VDT,
lines. alphanumeric terminal, electronic digitizer.
Zapf was amazed to see what I could present
to him (see Figures 1 and 2). He decided to include
me with a presentation of Ikarus in his general pre- Never did I hear such arguments from Hermann
sentation which he had to give at the next ATypI Zapf; he was already very familiar with computers
conference in the fall of 1975 in Warsaw. and always very eager and demanding to see the next
It began a close relationship between us. This innovation.
had a very decisive influence on the future develop- Today, the property “digital” is not only ac-
ment of our company URW Software & Type GmbH. cepted but also embraced. It serves as an additional
These days, everyone regards all fonts on com- characteristic which doesn’t interfere with the old
puters as digital fonts since they are stored in digital properties and holds an extremely high significance
formats such as OpenType. In the early seventies, regarding a font. It allows and creates new and
we had long discussions with famous designers. They important functions which did not exist before.
argued that pure mapping from analogue to digital With the invention of the Digiset by Rudolf Hell
is not changing the basic quality of a typeface (old in 1965, typefaces were digitized for the first time.
properties), namely its type, appearance, effect, ex- No additional ideas were put into place other than
pression and congeniality. Therefore they asserted using them 1:1 for typesetting on the Digiset, scaled
that the typefaces only had digital images and were linearly and displayed at resolutions between 1,000
therefore still analogue. “Digital” at that time was and 2,400 dpi. Hermann Zapf was engaged by Hell to
regarded as a pseudo-property. consult on typesetting and to design new typefaces.
96 TUGboat, Volume 36 (2015), No. 2

Figure 3: Kerning can be regarded as a power that repulses characters the nearer
they come to each other, and that attracts characters the farther they get from each
other (left). Kerning can be used also to calculate character positions for overlapping
and touching of text.

In 1985 during the yearly ATypI conference (this wrote about it [1]. Hermann Zapf urged me to “do
time sponsored by Hell in Kiel) Hermann Zapf dis- something”.
cussed with me the problem that too many people In 1991 at URW, we adopted the following ap-
were talking about digital typefaces and unfortu- proach for text fonts. First, the smaller the type
nately not knowing what they are really all about. size:
So, I decided to write a book with the title “Digital 1) the wider the composition,
Formats for Typefaces” which he kindly corrected as
2) the thicker the strokes,
a co-editor. Finally, we could present it in 1986 dur-
ing the next ATypI conference in New York. Later I 3) the broader the characters, especially
changed the title: it became “Digital Typefaces” [3]. the lowercase.
And second, the larger the type size in titles:
Digital text
1) the more compact the composition,
Being pushed by Hermann Zapf, I started automatic
2) the thinner the strokes, especially the hairlines,
kerning [4, 7] in 1980 together with Margret Al-
brecht. We wanted to save money because the gen- 3) the narrower the characters, especially
eration of kerning tables along with left and right the lowercase.
side bearings took a lot of time in our typeface pro- Simplified, one applies the rule that space and
duction. As in other cases of artificial intelligence, stroke width of light fonts (text fonts) are reduced
we had to go through several approaches throughout or enlarged by 7% on the average if the point size is
the years until 1987. Finally, we mixed programmed enlarged or reduced by a factor of 2 [8] (see Figure 4).
ideas and heuristic parameters gained by process- For bold fonts the opposite is true.
ing a lot of existing kerning tables manufactured To my knowledge, optical scaling was not em-
by different companies. We expanded kerning to ployed for bold fonts in the past because they weren’t
get overlapping and blending of characters in tightly (and still aren’t) used very often, and if so, they were
composed words [9] (see Figure 3). cut just for these special cases in certain point sizes
In hot metal printing, optical scaling was as a special effort.
usual [2]. In any case one had to cut the point Paragraph composition was our next project
sizes individually, so it was a matter of knowledge, in 1990 and it became the favorite of both of us.
but not of money. This changed when phototype- We called the program H-engine. Developed
setting came up and the possibility of linear scaling with and named after Hermann Zapf [5, 12] — it uses
came into existence. Optical scaling didn’t play a a justification per paragraph system — as described
role at the beginning of DTP, however, a lot of people by Donald E. Knuth [10, 11], along with “kerning
TUGboat, Volume 36 (2015), No. 2 97

cal demerits, which are obtained from a function of

the actual line lengths, given line lengths, given line
widths and tolerances of the layout parameters.
If hyphenation is turned on, words are replaced
by syllables. The H-engine has to follow a lot of
exceptions and to provide solutions for them, e.g.
ligature substitution, consecutive hyphens and good
or bad locations for hyphenation within a word. This
level of text/typographic detail promotes a better
fit and contributes to the reader’s comfort (see Fig-
ure 5).
A comparison between the H-engine and today’s
typical composition tools demonstrates the superi-
ority of the H-engine (see Figure 6). In 1995, the
“ H-engine” was implemented in InDesign by Adobe.
This has been a big step in digital text composition.
I believe that not too many people learned that
much about the “digital side” of Hermann Zapf as
Don Knuth and me. I learned a lot from him and
thank him.

References
[1] André, A., Vatton, I.: “Dynamic Optical Scal-
ing and Variable-Sized Characters”, Electronic
Publishing 7(4): 231–250, 1994
[2] Johnson, B., “Optical Scaling”, Master’s Thesis
for RIT, Rochester, New York, 1994
[3] Karow, P., “Digital Typefaces”, Springer Verlag,
Figure 4: Different point sizes, which have been
Heidelberg, 1994
generated at the same size for the purpose of
comparison. [4] Karow, P., “Font Technology”, Springer Verlag,
Heidelberg, 1994
[5] Karow, P., “ H-program”, brochure, URW Verlag,
Hamburg, 1992
on the fly” and expanding/condensing of characters
in order to obtain margin lines for a column that [6] Karow, P., “Ikarus”, German brochure, URW
are optically straight (optical margins), and achieve Verlag, Hamburg, 1975
typeset spaces among words within lines of text that [7] Karow, P., “Kernus”, brochure, URW Verlag,
are fairly constant in order to avoid rivers and creeks. Hamburg, 1993
Rivers run vertically through poorly spaced [8] Karow, P., “Optical Scaling”, brochure, URW
words in consecutive lines of text when the spaces Verlag, Hamburg, 1991
between the words have the same space or a greater [9] Karow, P., “Set theory for characters”, brochure,
space than the distance between the baselines of the URW Verlag, Hamburg, 1987
text. In contrast, a creek is a less severe form where [10] Knuth, D. E., Plass, M. F., “Breaking para-
the spaces between words are accidentally too wide graphs into lines”, Software—Practice & Expe-
within one line. rience, 11/11), 1119–1184, Nov. 1982
The basic feature of the H-engine, which was
[11] Knuth, D. E., “Mathematical Typography”,
programmed by Margret Albrecht, is to regard all
Stanford University, Cal., February 1978
lines of a given paragraph at once — making the jus-
tification per paragraph. At first, all words or [12] Zapf, H., Alphabet Stories — A Chronicle of
syllables are distributed to the lines together in a Technical Developments, Mergenthaler Edition
manner where each line gets a line length nearest to Linotype GmbH, Bad Homburg, 2007
its given individually parametrized width (as default
there is usually column width). The following opti-  Peter Karow
Hamburg, Germany
mization is controlled by minimizing the typographi-
98 TUGboat, Volume 36 (2015), No. 2

Figure 5: Comparison between H-engine and standard composition, without hyphenation.

TUGboat, Volume 36 (2015), No. 2 99

Figure 6: Comparison between H-engine and standard composition, narrow columns,

with and without hyphenation.