Ken Lunde

Wisconsin, the Army and the choice of Japanese

Lunde’s route into East Asian writing began with Russian. In a 2021 retrospective, he recalls his high school offering French and his brother Kern taking a Russian correspondence course. He joined the Army Reserve in September 1982, his senior year. Basic training began on his eighteenth birthday, August 12, 1983, at Fort Leonard Wood, Missouri. Interrogator training at Fort Huachuca, Arizona, followed; he completed forty-seven weeks of Russian at the Defense Language Institute in Monterey in 1984. 3

He entered Wisconsin–Madison in January 1985 intending to major in Russian, then chose linguistics because Russian required six literature courses he did not want. Linguistics required a non-Indo-European language. Seeking a challenge, he chose Japanese over Arabic and took intensive summer courses in 1985 and 1986. 3 His bachelor’s degree followed in 1987, his master’s in 1988, and his doctorate in 1994, with a Japanese minor. His dissertation was entitled Prescriptive Kanji Simplification. 1, 2

Military language study supplied sustained training before university; linguistics made writing systems objects of inquiry; Japanese supplied a community, texts and practical problems. His Japanese expertise was acquired deliberately as a young adult and grew through intensive study and subsequent work.

Edward E. Daub connected Japanese study to publishing. Lunde took his Technical Japanese I and II in 1986–1987. Noticing his attention to kanji distinctions, Daub hired him to help produce Basic Technical Japanese (1990), coauthored with Nobuo Inoue and R. Byron Bird. Lunde persuaded the authors and the University of Wisconsin Press to use Japanese PageMaker and a Japanese LaserWriter with Morisawa fonts, then prepared the camera-ready pages. He says this paid work largely financed graduate school, after Reserve educational benefits had almost covered his bachelor’s and master’s studies. 4

Daub remained more than an employer. Lunde asked him to accompany his doctoral graduation in 1994 and named his younger son Edward after him. They stayed in contact after Lunde moved west; Lunde and his son attended Daub’s memorial in 2016. The teacher who made technical Japanese usable and graduate school financially possible became part of the family’s memory. 4

Entering Adobe through a practical problem

Adobe hired Lunde while he was still a graduate student. In the 2014 Typekit interview, his long-serving manager David Lemon recalled that the company needed someone who understood Japanese character sets and encoding requirements. 5 Lunde’s first day was July 1, 1991; he was twenty-five. He stayed in the same department as Adobe grew and moved its headquarters from Mountain View to San Jose in 1996. 6 The appointment joined an academic specialty to a commercial need: software and fonts intended for Japan had to identify, display and print the right characters reliably.

His first project was Cube, an experimental element library based on Heisei Mincho. Its four design axes controlled width, height, weight and curvature. He presented a proof-of-concept font with about three hundred kanji glyphs to Adobe founders John Warnock and Charles Geschke late in 1991. 5 Demonstrating such a technology and turning it into a usable East Asian product required different scales of production.

Much of the early production work involved Adobe’s Heisei Mincho and Heisei Kaku Gothic fonts, supplementary Japanese glyph packages, and the initial Chinese and Korean character collections. Later milestones included Kozuka Mincho in 1997 and Kozuka Gothic in 2001. Lunde’s twenty-five-year account also records a durable working environment: in 2016 the type department had eleven people, with an average tenure approaching twenty years. Such continuity gave a specialist time to carry knowledge from one font format and generation of products into the next. 6

For an unfamiliar reader, the underlying task can be illustrated with an ordinary Japanese document. The text contains character codes; the font contains drawn shapes, or glyphs. The software must find the appropriate shape, position it, substitute a vertical or regional form when required, and preserve the text’s identity through copying, searching and printing. A beautiful drawing cannot by itself accomplish that chain. Lunde’s position developed around the mappings, rules, files and tests that make it work. Adobe’s published CMap resources document one important bridge: mappings from character codes in an encoding to numbered glyphs in an Adobe character collection. These include older systems such as Shift-JIS and EUC-JP as well as Unicode encodings. 22

His linguistic training helped him diagnose which layer had failed—drawing, character list, encoding or application—and why the difference mattered to regional readers.

Writing the manuals he needed

Lunde’s first book grew from information he had already been distributing. He began a file called JAPAN.INF in 1989; its final 1992 version was available in three Japanese encodings. He approached O’Reilly later that year. Editor Peter Mui and Tim O’Reilly supported a project initially estimated at about 250 pages that became a 470-page book. Lunde received his author’s copy of Understanding Japanese Information Processing on September 15, 1993. His later CJK.INF extended the reference work to Chinese and Korean. The first CJKV Information Processing followed at the end of 1998, with a 1999 imprint; the second edition appeared at the end of 2008, conventionally dated 2009. 8, 10, 11

The expansion was conceptual as well as geographical. The second edition’s contents move from writing systems and romanization through character sets, encoding, conversion, input methods and fonts to dictionaries, machine translation, operating systems, the Web and printing. The “V” includes Vietnamese’s romanized writing and historical uses of Chinese characters and chữ Nôm. Readers are asked to distinguish the language being written from the script used to write it, and both from the bytes used to store it. Code samples, tables and detailed appendices place the books between a linguistic introduction and an implementation reference. Their breadth reflects the practical fact that a document’s failure can pass through several systems before appearing on screen. 9

Publishing the books was itself an exercise in the tools he described. Lunde produced successive manuscripts in Japanese PageMaker, FrameMaker and InDesign. 6 A proposed division of the first CJKV into a main volume and a reference volume became far enough advanced that he maintained two FrameMaker book files before recombining them. 14 The finished first edition ran to 1,101 pages. 10 That scale shows how much tabular, linguistic and systems material the publisher agreed to carry together; it does not establish sales or universal influence.

The blowfish cover emerged through publishing relationships. Lunde proposed the animal and looked through Dover’s pictorial archive with Mui; O’Reilly’s Edie Freedman designed the covers. His analogy connected mishandling Japanese text with the danger of mishandling fugu. The first book’s Japanese translation in 1995 was made by his longtime friend Jack Halpern, known in Japanese as 春遍雀來, and Takeo Suzuki, 鈴木武生. 14

Lunde dates CJKV’s Japanese and Traditional Chinese translations to 2002. 14 O’Reilly Japan’s CJKV日中韓越情報処理, translated by 小松章 and 逆井克己, runs to 1,172 pages. 12 Its title places Japan before China, whereas the Traditional Chinese edition is 中日韓越資訊處理. A Taiwanese bookseller credits its translator as 鄭褚璋 and dates its cataloged edition to September 2003. 46 The available records do not resolve that difference in dates. The change in title ordering is a useful reminder that a reference about common technical problems still reaches readers through particular publishing languages and local expectations.

His audience also extended beyond people buying a large reference book. In May 2005, the Japanese Technology Professionals Association and Hosei University’s California research institute organized an English-language seminar in Burlingame on character encodings, including China’s GB 18030 requirement. Its notice mentioned that the first chapter of his book had been distributed with Acrobat, in English and later Japanese and Chinese. That setting links scholarship to the ordinary business of reaching another market: an engineer, localization specialist or technology company needed enough understanding to make documents and software function for customers abroad. 13

What a font engineer actually builds

The work is easier to understand if three kinds of number are kept apart. A Unicode code point identifies a character in the international repertoire; a glyph ID identifies an entry inside a particular font. 47 An Adobe CID, or character identifier, identifies a glyph in a specified collection. Adobe’s uppercase “CMap” resources map encoded text to CIDs. 22 The lowercase cmap table inside an OpenType font maps character codes to glyph IDs. 47 A font-building tool can use CMap resources to construct that table, but CID and glyph-ID numbers must be interpreted within their respective collections and fonts. 22, 47

Adobe-Japan1 makes the distinction concrete. Its current Supplement 7 specifies 23,060 glyphs, accumulated through successive supplements. These include publishing forms, ruby forms and other typographic alternatives beyond a simple list of encoded Japanese characters. Supplement 4 in 2000 expanded professional publishing coverage; Supplement 5 incorporated work with Apple and Japanese standards in 2002. In April 2019, Supplement 7 added horizontal and vertical forms for the newly named Reiwa era. A political-calendar event thus required a precise, small addition to an established glyph collection. It did not require treating the entire collection as a new Unicode repertoire. 21

An OpenType font packages outlines together with metrics, names, character mappings and layout rules. A font collection can place several font resources in one file while sharing identical tables, including outlines. 47 Language-sensitive substitution then uses the locl feature to select a localized glyph for a character. The application’s language metadata is part of the mechanism: the same code point may display a Japanese, Korean or Chinese form according to the selected language system. This is why a technically capable font can still produce an inappropriate result when an application does not convey the language correctly. 48

Lemon described Lunde’s production work as extensively automated in Perl, with proofing processes to catch errors among tens of thousands of glyphs. Lunde explained that Source Han’s Unicode-based working names helped automate the generation of build resources and made reassignment easier when mapping decisions changed. 5 This organization makes otherwise opaque material inspectable: a colleague can trace a glyph back to its intended character rather than relying on a growing mass of arbitrary labels.

The public Adobe Font Development Kit for OpenType, or AFDKO, supplies tools for building fonts from source data. Source Han repositories expose source files, build instructions, configurations, release notes and issue channels. This gives another view of the occupation. The finished typeface is what a reader encounters, but the engineer manages the reproducible route from a collection of drawings and specifications to usable font files in several configurations. Publishing those materials also lets errors and changes be discussed against a particular version. 23, 19

Ryoko Nishizuka’s Kazuraki presented a deliberately different problem. Released in 2009, its handwritten Japanese forms had proportional dimensions, including vertical ligatures joining two, three or four hiragana. Lunde’s 2012 technical presentation explains why the font used Adobe-Identity-0 rather than forcing its repertoire into Adobe-Japan1. Individual horizontal and vertical metrics were fundamental to the design. His scripts generated mappings and layout features, translating the designer’s intended writing behavior into rules software could execute. The project also exposed applications’ assumptions that Japanese glyphs would occupy uniform squares. 24

Here the collaboration is especially clear. Nishizuka supplied the visual practice; Lunde’s engineering helped retain its unusual spacing and connected forms through a technical system built largely around other expectations. The requirement was to understand an aesthetic decision closely enough to give it operational consequences.

Source Han and the conditions of multinational production

Source Han Sans, released in July 2014, depended on institutions as much as individual expertise. Google approached Adobe in 2011 and contributed financing, requirements, direction and testing in the intended markets. Adobe coordinated the work and supplied design and font-engineering experience. Nishizuka led the design; Iwata in Japan, Sandoll in South Korea and Changzhou Sinotype in China contributed regional work. Adobe distributed the family as Source Han Sans and Google as Noto Sans CJK. Seven weights reached the OpenType limit of 65,535 glyphs per font. Lunde handled character mappings, glyph repertoires and the coalescing of the different sources. 15

The project’s difficulty was partly the uneven distribution of what could be shared. Mark Wilson’s 2017 reporting for Fast Company describes an earlier ambition, attempts to revive it and a lengthy agreement between the companies. Lunde’s comparison of roughly 7,500 Japanese kanji against Korean requirements took about two weeks and found fewer than two hundred needing distinct Korean designs. Chinese regional requirements demanded substantially more new drawing and adjustment. Shared ancestry in the writing system therefore produced neither identical local expectations nor equal workloads. 18

Source Han Serif followed in April 2017, again with regional partners. Google’s release account gives a useful measure of the difference between characters and glyphs: the family covered 43,027 encoded characters while using the maximum 65,535 glyphs. The additional shapes supported regional and typographic requirements. Vertical writing required changes in form, orientation and placement; Hangul required its own coverage and composition behavior. The families were released for free use under open licenses, allowing developers to bundle them and manufacturers to embed them. Their availability depended on financed production, even though users did not pay a separate license fee for each installation. 16, 17

The initial result also had limits that became visible through particular readers. The 2014 announcement’s discussion acknowledged that its Traditional Chinese forms followed Taiwan’s Ministry of Education standard, with coverage for additional Hong Kong characters; this did not yet amount to a distinct Hong Kong regional design. 15 Version 2 in 2018 added Hong Kong as a fifth locale. Lunde’s technical account credits Arphic Technology in Taiwan with about 1,400 new Hong Kong glyphs and roughly 750 adjustments. Nishizuka also revised Bopomofo with specialists, and the engineering handled tone-mark placement and related layout. Language selection in applications had to catch up as well: InDesign CC 2019 added Hong Kong tagging. 44

That history gives “pan-CJK” a precise meaning. A common family can offer compatible design across regions while carrying separate forms and layout behavior within it. Regional expertise enters both at initial production and through later corrections. The family still does not contain every historical Han character in Unicode. Its coverage is a selected, engineered repertoire constrained by the intended uses and the font format, rather than a promise to render the entire encoded writing tradition. 44, 28

Adjacent projects reveal more everyday use cases. Masataka Hattori’s Source Han Code JP combined Japanese glyphs with Latin forms from Paul D. Hunt’s Source Code Pro. Hattori wanted Latin text to sit more comfortably with Japanese in programming: a two-thirds-width Latin design improved size and weight relationships while keeping predictable widths for indentation. Lunde translated the 2015 announcement. Source Han Mono later extended a related approach across five locales, seven weights and two styles, with source and build instructions published in its repository. The division of credits matters: Hattori identified and developed the coding use case, Hunt supplied the Latin design, and Lunde’s involvement belongs within that collaborative engineering and dissemination history. 43, 20

Testing, mistakes and the less visible products

Adobe Blank, introduced in 2013, shows how far this work can move from designing visible letters. Adobe researcher Joel Brandt needed a font whose mapped characters would be invisible and have zero advance width. Lunde built one covering the permitted Unicode values while balancing the size of its mappings against the number of blank glyphs. It allowed a test element’s width to reveal whether the desired font had loaded or the blank font was still being used. This served font-selection previews and testing. Its usefulness lay in deliberately suppressing visible shape, so that an otherwise difficult software condition became measurable. 40

Large visible fonts bring a different kind of uncertainty. In June 2015 Lunde publicly reported that he had inadvertently reverted 212 vertical kana in Source Han Sans to earlier forms. The mistake had passed through two releases. A user’s report prompted a critical update; Lunde thanked the reporter, apologized and rebuilt the affected material over a weekend. AFDKO tools let him copy the required outline data from an earlier font rather than repeat a much longer build process. Google’s corresponding Noto fonts received the correction too. This is unusually concrete evidence of maintenance: source control, a report, diagnosis, a shortcut that preserves the needed data, and a new release. 42

Standards themselves also require checking. In a 2016 article, Lunde compared the Taiwanese Ministry of Education’s rendering of 湱 with several other standards and a dictionary source. He concluded that a stroke connection in one rendering was an error, rather than a regional form to reproduce uncritically. His examples also included duplicated Korean mappings that could share a glyph. The practical lesson is that compliance requires comparison and judgment: an authoritative chart is an important source, but a published drawing can contain a human mistake. The engineer must document why a discrepancy should be preserved or corrected. 41

The repeated pattern is organized skepticism. A glyph is checked against its code and source, one standard against another, a release against its predecessor, and a reported failure against a reproducible build. Automation makes the volume manageable; the records allow another person to inspect the decision. Neither removes the need for readers and collaborators who recognize something wrong.

Characters, variants and committee work

The standards side of Lunde’s career addresses a problem that a font cannot resolve on its own: deciding what text should be representable independently of any particular typeface. Unicode’s Han unification identifies shared abstract characters across the writing traditions that use them. It leaves regional glyph forms to fonts and layout. A code point therefore does not tell software whether the surrounding text is Japanese or Chinese, and a shared character need not have one compulsory shape. This distinction is central to the work: encoding establishes textual identity; typography supplies a form appropriate to the document and reader. 28

The Ideographic Variation Database, or IVD, provides another mechanism when a particular variant needs to be retained in plain text. An ideographic variation sequence combines a base character with a variation selector, associated with a registered collection. It narrows the intended set of glyph forms without specifying every detail of a font’s drawing. Unicode Technical Standard #37, currently edited by Lunde, sets out stable collection mappings, registration and public review. Its acknowledgments also identify earlier work by Shinichi Hiura and Eric Muller. The mechanism belongs to a continuing collective history of reconciling textual interchange with meaningful variation. 29

Unihan is a different resource: a database of information about Han characters. It includes readings, dictionary references, radical-and-stroke information, variants, source references and mappings. These properties let a character be found and compared through several kinds of evidence. Lunde edits Unicode Standard Annex #38, which documents the database; the edition checked for this profile is dated September 1, 2026. 30 He has also published technical notes on unusual Han structures and the history of Unihan properties, both revised on September 16, 2026. These are continuing works of documentation with their own revision histories. 34, 35

Within Unicode, he chairs the CJK & Unihan Working Group, whose vice-chair is Eiso Chan, 陳永聰. The group brings specialists together, develops recommendations and reports to the Unicode Technical Committee, or UTC. Its recommendations are nonbinding; the UTC makes the final decisions. This places Lunde in a coordinating and editorial role rather than giving him personal authority to encode a character. It also separates the expert discussion from the formal decision that follows it. 27

IRG contributors span China, Japan, South Korea, Taiwan, Hong Kong, Macao and Vietnam, alongside scholarly and technical organizations such as the SAT Buddhist-text project. Its site lists Lunde as convenor, Chen Zhuang (陈壮) as chief editor, Yi Bai (白易) as unification-document editor, and Henry Chan (陳輝恒) as online-review-tool manager. These divided responsibilities make the work a collective operation. 25

A proposed character must be supported by evidence, not simply by a request for a new drawing. The IRG’s procedures call for documented use in running text, source references, structural information such as radicals and strokes, and a font that reviewers can examine. Submissions are checked against already encoded characters and other proposals. Reviewers comment; submitters respond; questionable items can be deferred or withdrawn. The online tool and registered documents preserve the discussion. A code in a temporary review font’s private-use area is a way to show the candidate during this process, not an international encoding already granted to it. 26

Institutional history also matters to the current description. ISO/IEC’s character-coding subcommittee confirmed Lunde as the new IRG convenor in June 2024, thanking Lu Qin for twenty years of service. 38 On June 26, 2026, it converted the IRG into SC2 Advisory Group 1, to give technical advice on ISO/IEC 10646, and appointed Lunde for a one-year 2026–2027 term subject to Unicode’s endorsement. The resolution says participants represent their expert opinions rather than national-body interests, and provides for existing liaison organizations to participate. 39 The IRG site records this conversion beneath its older WG2 heading. 25

Repertoire review and font coverage have separate schedules. The IRG site records its 2021 working set becoming Extension J in Unicode 17.0 in 2025, while the 2024 set remains under review. On October 3, 2026, it lists the next meeting for October 12–16 in Tokyo. 25

Leaving Adobe and taking on inherited work

Lunde’s Adobe employment ended on October 18, 2019. He described the departure as abrupt; his public accounts do not establish the circumstances sufficiently to reconstruct a specific dispute. His last two months carried the title Type Architect. 7 The chronology after Adobe is clearer: he worked as an Apple contractor from January 16, 2020, through July 30, 2021, and became an employee on August 2, 2021. 8 The university profile continues to list Apple as his employer. 2 His current employment supplies a concrete occupational setting, but public sources do not describe his Apple projects, pay or the financial arrangements for his standards work.

John H. Jenkins, a longtime Unicode contributor and close friend of Lunde for more than twenty years, became his colleague at Apple in 2020. Jenkins died on February 28, 2023. Lunde’s annual account describes assuming maintenance of Unihan and additional editorial duties, learning the tools and assets Jenkins had used, and restoring the database’s operating continuity by June. He deferred most noncritical changes to a later release while handling necessary additions and corrections. Behind the release schedule was the loss of a person who held a particular body of knowledge. 32

The same account makes the recurrent labor visible. Preparing the CJK working group’s recommendations could follow a three-hour meeting with an entire weekend of report work. Lunde attended the 2023 UTC meetings at Google in Sunnyvale, Adobe in San Jose and Microsoft in Redmond. Standards were being maintained through documents and meetings hosted by competing technology companies, with expertise crossing their institutional boundaries. 32

In 2025 another frequent collaborator, Andrew West, died. Lunde wrote that they had met in person only once, in 2019, despite a long online working relationship. He retained a character West had submitted as the IRG site’s banner in his memory. That detail conveys an ordinary feature of this specialist world: colleagues can share years of exacting work through documents and correspondence while rarely occupying the same room. 33

Unicode’s 2018 Bulldog Award had recognized Lunde’s CJKV work, fonts, variation-database contributions, editorial work and willingness to help younger contributors. The award is the institution’s own assessment, but its range fits the documented career. Much of his contribution takes the form of making accumulated knowledge usable by other people: in a book, a build system, a database, a correction or a meeting report. 31

Family, hobbies and a life in California

Adobe was also where Lunde met his wife, Hitomi. His 2019 retrospective says they had known one another as colleagues for approximately five years before marrying. Their daughter Ruby was born in March 2000 and grew up able to speak English and Japanese with her two sets of grandparents. Both sons were born before their parents joined Adobe; by 2019 Lunde was also a grandfather. 7 These details situate the transpacific language work inside an English- and Japanese-speaking household.

Marksmanship is another sustained part of his life. He founded Adobe’s Marksmanship League in October 2000; it normally met monthly, with corporate support toward range fees, targets and ammunition. Its two-hundredth event took place in May 2019. 7 The activity joined a personal interest to a workplace social network over nearly nineteen years.

Vernon Delano Lunde, who died in 2022, had been a pilot, hunter, marksman and fly-fisher. He and Jeanne had three children: Kern, Kitrin and Ken. His parents retired from Mount Horeb to Hot Springs, South Dakota, in 2004. Ken described annual or twice-yearly drives of about 1,400 miles from San Jose, taking hunting equipment and wine or liquor, and sitting on the deck with his father over drinks and stories. An inherited Randall knife carried Vernon’s name. These details preserve both physical distance from Wisconsin and repeated return through family visits; they also show interests that do not arise from the font profession. 36

A 2022 interview with Tesla Owners of Silicon Valley supplies a lighter account of household life. The family had three differently colored Model 3s, named Baby Pearl, Baby Ruby and Baby Sapphire; the last belonged to a son living in Milpitas. Ruby teased him about giving a car her name. Dog walks helped him observe colors before choosing his own car, and overnight charging became a garage routine. His reasons for purchasing included performance, technology, cost and whether the charging network made the Wyoming route to South Dakota practical. He had rejected an earlier, more expensive option. The source is an owners’ club interview, but its details place a conspicuously technical interest among family jokes, errands and long-distance visits. 37

These accounts should not be stretched into an explanation of every professional choice. They do, however, prevent the career from collapsing into a list of standards and font releases. Language study, a teacher, a wife met at work, children using two languages, aging parents far away and recurring hobbies belong to the circumstances in which the specialist practice developed.

A continuing public role

Lunde continues to explain standards outside committee meetings. The Character Information Technology Promotion Council’s fifteenth-anniversary program in Tokyo in December 2025 listed his English-language keynote, My Brief History of Standardization. The wider program brought together government, linguistic research and the font industry: talks on administrative character standards and character identity, and a panel involving Japanese type companies. That setting places his work among the people who must reconcile scholarly distinctions, public administration and everyday software. It is one part of a continuing Japanese professional world, alongside his corporate work in California. 45

The career’s shape becomes clearer when these worlds are kept together. Reserve benefits and paid textbook production made university study workable. Adobe supplied a sustained setting for large-scale font development. Google and regional partners made the Source Han undertaking financially and technically possible. Publishers and translators carried his reference work to different readers. Unicode, ISO/IEC and the IRG provided procedures through which private expertise could become a shared, reviewable resource. Friends and collaborators held knowledge that had to be carried forward after their deaths.

Lunde’s distinctive practice lies in moving between these scales without losing the individual object under examination: a kanji in a textbook, a vertical kana reverted in a release, a regional glyph, a dictionary reference or a candidate character awaiting evidence. The broader results depend on repeated small acts of comparison, translation, construction and correction. The public record through October 3, 2026, shows that practice continuing through Apple employment, standards coordination and Unihan documentation. Its continuity also depends on people and records: a colleague’s tools must be understood, a regional expert’s judgment retained, and a disputed character made reviewable by the next participant.