FORUM TIMELINES Timelines provides perspectives on HCI history, glancing back at a road that sometimes took unexpected branches and turns. History is not a dry list of events; it is about points of view and differing interpretations. Jonathan Grudin, Editor A Personal History of Modeless Text Editing and Cut/Copy-Paste Larry Tesler Consultant | [email protected] Larry Tesler’s vision of interaction The 1960s a paranoid patient. While working design process has inspired many In 1960, while a student at the on his team, I got to know Alan designers, developers, and researchers. Bronx High School of Science, I Kay, Don Norman, Terry Winograd, His leadership in early graphical user learned a FORTRAN-like language. and David Canfield Smith—all of interface successes led to his receiving I loved its power, but its unintuitive whom became HCI pioneers—and SIGCHI’s Lifetime Practice Award in restrictions frustrated me. I learned a little about cognitive 2011. —Jonathan Grudin In 1961, I entered Stanford psychology. I have been a computer program- as a freshman. In 1962, I made In early 1969, I visited Doug mer for more than 50 years. From usability improvements to a pio- Engelbart’s Augmentation Research the beginning, I was annoyed by neering animation language. Center at SRI in Menlo Park, software that made life harder That project gave me experience California. Engelbart had recently than necessary for users. I got to do with discount usability studies given the first public demonstration something about it as a student at and participatory design [1]. of NLS (oN Line System), a vision- Stanford University and in a variety Soon, word got around that I ary prototype built on a time-shar- of subsequent engineering, user was a pretty good programmer ing system [2]. The groundbreaking experience, and management roles who made software easy to use. event became known as “the moth- at Xerox PARC, Apple, Amazon.com, Professors and grad students alike er of all demos.” Among the innova- and Yahoo! asked me to consult. In 1963, I tions that debuted were the mouse, The best known of my contribu- founded a contract software com- tiled windows, multiple views, tions is cut/copy-paste. I developed pany that was one of only six in the outlining, hypertext, collaborative the pattern over a period of years Palo Alto Yellow Pages. editing, and videoconferencing. in collaboration with a series of During the 1960s, interactive From 1968 to 1970, I sometimes colleagues. But cut/copy-paste was time-sharing began to displace pasted up a quarterly catalog for not a distinct project; it was one of batch, and pointing devices became a local nonprofit. While “cutting a collection of graphical user inter- common on mini-computers. I and pasting” with blades and glue, face (GUI) patterns I called modeless much preferred interactive to I imagined an interactive page text editing. batch, but most interactive pro- makeup system that would simplify I was not the first person to notice grams had modes, which always the process. the ill effects of modes on error tripped me up. I began to analyze During the same period, Pentti rates. Nor was I the first to try to command languages to root out the Kanerva showed me his PDP-10 eliminate the most onerous modes. causes of modes and mode errors. port of Brian Tolliver’s full-screen July + August 2012 But for me, mode reduction became In 1968, I began working at the text editor, TVEDIT. Kanerva had a research endeavor and a business Stanford Artificial Intelligence added a simple error-recovery mission. I helped to develop the Laboratory (SAIL) for Ken Colby, a command called oops. He had also theoretical underpinnings of mode- psychiatrist and cognitive scientist. added a two-step move: The delete less editing and the first products to Colby had developed PARRY, a con- step moved user-specified text interactions affirm the validity of the theories. versational program that simulated to the top of a stack; the retrieve 70 TIMELINES FORUM How Modes Degrade Usability In a 1981 article about Smalltalk [9], I defined a mode step moved the top element of the technical specifications, source stack to a user-designated location. code, and other indented outlines. as “a state of the user interface that lasts for a period of Between steps, the user could do Its regular users found it a good fit time, is not associated with any particular object, and has anything that left the stack intact, for that. But I felt that it would not no role other than to place an interpretation on operator including filing, searching, typing, gain public acceptance as a tool for input.” Three properties of a command language that and moving other text. Although editing common documents such cause mode-related problems are: TVEDIT had modes, it seemed to as letters, memos, and forms. Its me that a two-step move and an command language had numer- • Verbs precede their objects. The most frequently used oops-like error-recovery command ous modes. Outside text-entry commands in many interactive systems involve a verb could help to make a suitably mode, virtually every keystroke and one object. If the language has any consistency at designed editor modeless. and click changed the mode. all, there is prevalent command syntax, which is usually The syntax of the NLS command either prefix or suffix. The distinction is whether the user The 1970s language evolved over time, but it specifies the verb before or after its object. Suffix syntax In 1971, Les Earnest, director of was always prefix, in which the verb has a usability advantage: When the user specifies the the A.I. Lab, asked me to design is specified before its object (see verb last, its object has already been specified, and the and implement a page-makeup Figure 1 and sidebar: “How Modes command can be executed immediately. Systems that language that could number sec- Degrade Usability”). To delete a use prefix syntax must enter a mode to wait for the user tions and generate an index, table paragraph, you told NLS to delete of contents, footnotes, cross- before you told it what paragraph it to specify the object. Keeping track of mode changes can references, and so on. I proposed should delete. distract a user from the task at hand. to make it interactive, but he Clicking the mouse button when • Key meanings are mode-dependent. Languages that wanted a batch system, which the pointer was over something use unmodified letter keys to do anything but enter I admitted would be easier. was called marking. Three fre- those letters as text need at least two modes: text and In a few months of intense work, quently used NLS commands that I created PUB: The Document required marking were: command. If unmodified letter keys are typed in command Compiler [3]. PUB was a markup mode but the user thinks the system is in text mode, language with embedded tags and D(elete) W(ord) <mark unintended and sometimes disastrous results ensue. scripting. It became popular among affected word> <ok> • Mode escapes are inconsistent. Users often get “stuck” graduate students at ARPANET- M(ove) T(ext) <mark source in a mode. An oft-heard question is, “How do I get out of connected universities. text start> <mark source text In 1973, I joined Xerox Palo this mode?” end> <mark destination> <ok> Alto Research Center (PARC) as a member of the PARC Online I(nsert) S(tatement) <mark Office System (POLOS) team but destination> <type text to Then it performed the command. spent some of my time working on insert> <ok> Because move was a single com- Smalltalk with Alan Kay’s Learning mand, both the destination and the Research Group. One reason I was The command-accept action, source had to be visible onscreen interested in working with Kay was symbolized here by <ok>, could be before typing “M.” The same restric- that his invention of overlapping invoked from either the keyboard or tion applied to copy and replace. windows was motivated by a desire the mouse. Text was an arbitrary Features such as collapsible out- to find alternatives to modes. span of text. A statement was usu- lines provided ways to circumvent Most members of the POLOS ally a paragraph. the restriction, but the user had to team had come to Xerox from As the user typed and clicked, learn more syntax and plan ahead. Engelbart’s group at SRI. My pieces of a command line accumu- I believed that competitors would manager, Bill English, had lated in a visible window. The user surpass Xerox in speed of learn- been involved in the design could remove the newest piece of ing and ease of use if we stayed July + August 2012 of the mouse, co-authored the command line from the win- with NLS syntax. Most of my col- Engelbart’s 1968 paper, and man- dow or erase the whole line and leagues were unconcerned. They aged the famous NLS demo. start over. When the user invoked considered NLS intuitive because From its inception, NLS was used <ok>, NLS erased the whole line, of its English-like verb-object mainly to construct and revise preventing further modification. grammar. The syntax allowed a interactions 71 FORUM TIMELINES • Figure 1. NLS prefix in which the verb is specified after syntax, simplified. verb noun mark OK its object (see Figure 2 and sidebar: “How Modes Degrade Usability”). Delete Character I told Rulifson about the error- Move Word mark recovery advantages of suffix syn- tax (see Figures 3 and 4), namely: Insert Statement type Replace Text (2 marks) • If the user made a selection mistake while specifying the object, she could simply select again.