Computer-Mediated Communication Magazine / Volume 1, Number 3 / July 1, 1994 / Page 5


PLATO: The Emergence of On-Line Community

by David R. Woolley (drwool@well.sf.ca.us)

The PLATO system was designed for Computer-Based Education. But for many people, PLATO's most enduring legacy is the on-line community spawned by its communication features.

PLATO originated in the early 1960's at the Urbana campus of the University of Illinois. Professor Don Bitzer became interested in using computers for teaching, and with some colleagues founded the Computer-based Education Research Laboratory (CERL). Bitzer, an electrical engineer, collaborated with a few other engineers to design the PLATO hardware. To write the software, he collected a staff of creative eccentrics ranging from university professors to high school students, few of whom had any computer background. Together they built a system that was at least a decade ahead of its time in many ways.

PLATO is a timesharing system. (It was, in fact, one of the first timesharing systems to be operated in public.) Both courseware authors and their students use the same high-resolution graphics display terminals, which are connected to a central mainframe. A special-purpose programming language called TUTOR is used to write educational software.

Throughout the 1960's, PLATO remained a small system, supporting only a single classroom of terminals. About 1972, PLATO began a transition to a new generation of mainframes that would eventually support up to one thousand users simultaneously.

PLATO Notes: Original Development

In the summer of 1973, Paul Tenczar asked me to write a program that would let PLATO users report system bugs on-line. Tenczar was the head of the system software staff, and I was a 17-year old university student and junior system programmer. I had been with CERL for about a year, learning the ropes and doing minor programming tasks at minimum wage.

We already had a way for users to report bugs, but it was just an open text file called "notes". Anyone could edit the file and add a comment to the end. After investigating a problem, a system programmer would insert a response (typically something like "+++Fixed - RWB").

This was simple enough, but there were problems. For one thing, only one person could edit the file at a time. For another, there was no security at all. It was impossible to know for sure who had written a note. Most people signed or at least initialed their comments, but there was nothing to enforce this. And occasionally some joker would think it was fun to delete the entire file.

It was just such an incident that prompted Tenczar to ask me to develop a replacement. His idea was a simple refinement of the method we had been using: a user would type a problem report into a special-purpose program, which would automatically tag it with the date and the user's ID and store it safely in a tamper-proof file. The same program would allow convenient viewing of the stored notes. Each would appear on a split screen, with the user's note on the top half and the system staff's response below.

It occurred to me that half a screen might not be enough space for some notes. And that some problems might require back-and-forth conversation between a user and the system staff. A limit of one response per note wouldn't permit much dialog.

I came up with a design that allowed up to 63 responses per note, and displayed each response by itself on a separate screen. Responses were chained together in sequence after a note, so that each note could become the starting point of an ongoing conversation. This is what John Quarterman calls a star-structured conferencing system, and PLATO Notes was apparently the first of its kind.

My first prototype kept all notes in one file. Upon entry you would see an index of the most recent notes, listing each note's number, date, title, and number of responses. You could then select a note to read, or page back through the index to find older notes.

As I showed this to other members of the system staff, we began to talk about other ways that this program might be used beyond just problem reports. We thought it would be nice to have a separate area where new users could ask questions and get help from more experienced users, and another area where the system staff could announce new PLATO features. So I added a top-level menu to let people choose among three notesfiles: System Announcements, Help Notes, and General Notes.

Notes was released on August 7, 1973. It was named after the text file it replaced, so that people accustomed to typing "notes" would be taken to the right place.

Every note or response appeared on its own screen. Since PLATO was designed for education, its architecture was biased toward carefully crafted full-screen displays. It was easy to place text or graphics at specific locations on the screen, but nearly impossible to scroll text. For Notes, this was both an advantage and a drawback. One nice feature was that the note title, date, time, and author's name always appeared in the same place. After using Notes for a while, your eye "knew" exactly where to look for these things.

On the down side, each posting was limited to 20 lines of text so as to fit on one screen. The only way to overcome this was to write a series of responses, but that allowed other responders to slip in and disrupt the flow. Still, the 20-line limit had the virtue of encouraging brevity.

Most options for reading notes required only a single keypress. While reading a response, for example, one keypress could perform any of these functions (among others):

There were too many options to list them all on every screen. Most prompts were quite minimal, but a Help key was universally available. It would display a complete list of the options available at any point.

Notes quickly became an indispensable part of the landscape. It appeared just as PLATO was beginning a phenomenal growth spurt made possible by the new mainframe. Although PLATO had been evolving for over a decade by this time, to the new flood of users coming on-line, PLATO without Notes was hard to imagine.

The PLATO Architecture

PLATO is designed to be extremely responsive to keys. Every keypress is processed individually by the central mainframe, but the response (or "echo") is usually so fast as to appear instantaneous. An echo time of 100 milliseconds is excellent; anything over 250 is considered unacceptable.

This is vital, especially because displays do not appear instantaneously. Originally, all PLATO terminals communicated at 1200 bps. At that speed, a long posting in Notes might take up to 10 seconds to fill the screen. But a single keypress aborts the display and moves on if the first line or two of a note doesn't spark your interest.

The ability to abort pending display output is crucial. Even now that faster connections are possible, connecting through a network that does not permit aborting output makes PLATO feel maddeningly sluggish.

Talkomatic and "Term-Talk"

Any competent PLATO programmer can quickly hack together a simple chat program that lets two users exchange typed one-line messages. PLATO's architecture makes this trivial. A few such programs existed on PLATO before 1973, but they did not get much use, probably because the user community was quite small and most terminals were still in a single building.

In the fall of 1973, Doug Brown designed a program that let several users chat as a group. He wrote a simple prototype to demonstrate the concept and called it Talkomatic.

The real magic of Talkomatic was that it transmitted characters instantly as they were typed, instead of waiting for a complete line of text. The screen was divided into several horizontal windows, with one participant in each. This let all the participants type at once without their messages becoming a confusing jumble. Seeing messages appear literally as they were typed made the conversation feel much more alive than in line-by-line chat programs.

I worked with Doug to expand Talkomatic to support multiple channels and add other features. Each channel supported up to five active participants and any number of monitors, who could watch but couldn't type anything. (One drawback to the Talkomatic approach is that the size of the screen limits the number of participants in a channel.)

Empty channels were open to anyone, but any active participant in a channel could choose to "protect" it. This prevented anyone from monitoring the channel, and the participants could then decide who else to admit.

Talkomatic was an instant hit. Soon it was logging over 40 hours of use per day. It was not officially part of the PLATO system software, and in fact it was used mostly for what administrators would consider frivolous purposes. There was no way to contact a specific person to let them know you wanted to talk, so it was more like a virtual water cooler than a telephone substitute. People would hang out in a channel and chat or flirt with whoever dropped by.

But Talkomatic was so appealing that it inspired the system staff to create an officially supported chat feature. It became known as "term-talk" because it could be accessed from anywhere on PLATO by pressing the TERM key and typing "talk". The TERM key was originally meant to provide hypertext-like branching to term definitions. In practice, it was rarely used for terms, but it was handy for instant access to features like "talk".

A "term-talk" conversation was limited to two people, but had its own advantages: you could page a specific person, and you could use it without exiting from whatever else you are doing. A person receiving a page would see a flashing message at the bottom of the screen identifying the pager, and could use "term-talk" to accept. The bottom two lines of the screen then became a miniature Talkomatic. An unwanted page could be rejected with "term-busy", or simply ignored until the pager gave up.

A feature was later added to "term-talk" that allowed the participants to switch to "monitor" mode, in which one person could actually view the other's screen. The person being monitored was free to move about the system normally, editing files, running programs, etc. This was extremely useful for remote consulting: someone who needed help could literally show an on-line consultant what they were trying to do while maintaining a conversation at the bottom of the screen. To ensure privacy, monitor mode could be initiated only by the person whose screen was to be monitored.

Personal Notes

With Notes and "term-talk" in place, it began to seem natural to use PLATO as a means of communication. What it obviously lacked was a way to send private mail. Kim Mast tackled the job, and in August of 1974, Personal Notes was released.

Personal Notes was similar to Notes in many ways: each note appeared on a separate screen, and options such as moving to the next or previous note, deleting a note, or responding, were available as single key presses. There was no index of notes, however. Entering Personal Notes took you immediately to the first note you had not yet read. From there, you could move forward or backward through your notes.

Kim and I worked together to integrate all of PLATO's communication features into a seamless package. For example, while viewing a note, you could:

All of these options were available in both Notes and Personal Notes, and the same keys were used in both.

Notes Categories

The success of Notes led to overcrowding. There were only two notesfiles which users could write in, and they were used for practically everything. It became a chore to wade through the volume of notes written every day, and people began to ask for a way to filter out notes they weren't interested in.

My solution was this: the system staff would define a list of categories, such as "bug reports", "suggestions", "events", "jokes", etc. Anyone writing a note would assign it to one of these categories. Users could select which categories they wanted to see when reading notes, and their selections would be stored permanently as part of their user records.

In early 1975 I created a version of Notes that supported categories, and released it to a limited group of users for testing and comments.

Suggestions from users were vital to PLATO's evolution, and Notes was no exception. Since I had written Notes originally, it was my turf, and I made most of the decisions about what features were implemented. But I had the benefit of lots of ideas from users as well as from the rest of the system staff. Often a suggestion would strike me immediately as great idea, and if it was not too difficult, it might be implemented and released within a day or two. Not all ideas were implemented, by any means. But sometimes I would argue against a proposed change, only to be convinced of its merit by cogent arguments or by the sheer number of people voicing support for it. Thus, Notes was shaped largely by a consensus of the entire PLATO community.

The notes categories concept was well received at first, but it got bogged down in controversy over features and never made it to general release. A particularly contentious issue was how notes should be presented. One faction wanted to see all notes in chronological order, with the categories serving only as a filter to skip unwanted notes. Others wanted categories to serve an organizing function, as well: all the notes from one category would be shown, then all the notes from the next category, and so on.

Strange as it seems now, I held out against organizing notes by category. I was used to reading notes about many different subjects all jumbled together, and just wanted to be able to see all the new notes listed together in one place. But support for more organization grew, and I began to see that I was in a losing battle.

In the meantime, though, other problems became apparent. First, I realized that as the volume of notes increased, there would be technical problems with keeping everything in one large file. Second, it wasn't clear how many categories would be needed. I had designed in a limit of 60, which seemed like a vast increase over the 3 we had been living with. But if we ever needed more it would be very difficult to increase the limit. After months of wrangling, my concept of notes categories seemed fatally flawed. I really didn't know where to go with it.

About this time, a few people began to ask for private notesfiles. We had all seen how useful Notes was for discussing development of PLATO itself. Couldn't the concept be extended to allow any small group of people working on a project to communicate among themselves? In fact, a group in Chicago that was using PLATO to develop medical courseware wrote a clone of Notes for their own use.

Suddenly the future clicked into focus. I abandoned the categories project and began to implement Group Notes.

Group Notes

Group Notes was a generalization of the original Notes. Now there could be an unlimited number of notesfiles, and users would be able to create private notesfiles for use by their own work groups. Group notesfiles would serve the same purposes for which notes categories were designed with none of the inherent problems. The 60-category limit vanished. Distributing notes across many files solved the technical problems of dealing with large volumes of information. The burden of managing notesfiles would be distributed, as well; no longer would the system staff have to oversee everything. And, yes, notes would be organized by subject, as so many people had insisted. Group Notes is one of those ideas that, with hindsight, seems glaringly obvious.

Group Notes was released in January, 1976, and thereafter use of Notes skyrocketed. Soon there were public notesfiles for subjects like books, movies, religion, music, and science fiction, as well as many private notesfiles for work groups.

The internal structure of notesfiles still had not changed much since 1973, and it was beginning to show its weaknesses. In particular, it made it difficult to implement a sorely needed option to read all responses written since a certain date and time. So I rewrote Notes almost from scratch, and converted all notesfiles to a new internal structure in July, 1976. Although it has been modified many times since, this version forms the core of the Notes software still in use today.

Access Lists

Access lists are the key to Group Notes. A person who creates a notesfile is automatically registered as a "director" of the file. A director can edit the access list to specify who else can access the notesfile and with what privileges. Access can be specified for individual user IDs or for entire work groups, and any level of access can be granted to the general public (anyone not specifically listed).

There are six access levels:

Read/write is the most common type of access. It permits both writing new notes and responding to existing notes.

Read/respond permits responding, but not writing new notes.

Write-only access permits a user to write new notes, but not read or respond. It is sometimes used as a blanket access level for the public, providing a way for someone to request access to a private notesfile. It is also useful for collecting comments from the public about some issue, while maintaining the privacy of each person's remarks.

Generally, anyone who can read a notesfile can also view its access list, although the director can choose to prohibit this.

Reading By Date

Notes offers a way to read all notes and responses written since a certain date and time. This feature is designed so that you can sequence through all new postings using a single key. For every note with new responses, the base note is displayed first to provide context. A keypress then skips to the first new response. Pressing the same key repeatedly sequences through the rest of the response chain, and then skips to the next note with new responses.

In 1978, John Matheny implemented the Notes Sequencer, a great boon to habitual notes readers. The Sequencer lets you create a personal list of the notesfiles you read regularly, and automatically keeps track of the last time you read each one. Using the Sequencer, you can quickly scan all the notesfiles in your list for new postings with a minimum of key presses.

Deleting Notes

Someone who has second thoughts after posting a note or response can delete it or edit it, as long as no responses have been added after it. This restriction is meant to avoid garbling the thread of a conversation. Deleting a response from the middle of a chain can make the following responses seem nonsensical. But an author who desperately wants to delete a posting anyway can appeal to the notesfile director, who can delete any posting without restriction.

A director can delete a response from the middle of a chain without disturbing subsequent responses. However, if a director deletes a base note, all responses disappear with it. Directors frequently use this power to clean out a notesfile, removing old notes that are no longer of interest.

Anonymity

The idea of anonymity in Notes was controversial when first proposed, but the issue was resolved by leaving it to the discretion of each notesfile director. If a director chooses to allow anonymity, then anyone posting a note or response in that notesfile is given the option of making it anonymous.

An anonymous note is truly anonymous. Not even the notesfile director nor the system staff can determine who posted it, because the user ID is not saved anywhere. The word "anonymous" appears in the header where the user ID would normally be.

PLATO Notes avoids some of the problems that have plagued experiments with anonymity in other conferencing systems. It is not possible to masquerade as someone else, because Notes does not allow the use of pseudonyms. The only identification that can appear in the header is the author's actual user ID or the word "anonymous". The fact that anonymity is the choice of each user is important, too. Someone could post an anonymous note saying "I'm David Woolley and I kick my dog every morning," but everyone reading it knows that the author specifically chose to make this note anonymous, so the identity claimed in the text is not to be taken seriously.

Most notesfile directors do not permit anonymity, but it is very useful in some situations. Anonymity can be abused, but a notesfile director can delete offensive postings. The version of Notes now used on NovaNET even allows a director to review anonymous postings before they become publicly visible.

Director Messages

Another privilege that notesfile directors have is to flag their postings with a "director message", a single line of text which appears above the standard header. Directors often use the message to flag official postings, such as statements about policy or notices that an inappropriate note was deleted. The director can specify what the message should say, but a single message has to suffice for all situations since there is only one director message per notesfile. Typical messages range from the serious ("OFFICIAL MESSAGE") to the humorous ("Not Operating With A Full Deck").

A director can toggle the message on or off for any posting, even those written by other people. For example, in a notesfile used to report problems, a director might set the message to "FIXED" and use it to flag problems that have been resolved.

"Term-Comments"

One of the ways that Notes supports PLATO's educational purposes is through a feature called "term-comments". While running a program, a user can press TERM and type "comments", and then type a note to the program author. Such comments are collected in a notesfile that the author has associated with the program. Each note is tagged with a header indicating the exact point in the program where the comment was made, so if a note reports that "entorpy is misspelled on this page", the author knows exactly where to look for the error.

Linked Notes

Around 1975, Control Data Corporation set up its own PLATO system in Minneapolis and began turning PLATO into a product. By 1985, over 100 PLATO systems were operating at sites around the world, about 60% of them running full-time. Some of them were linked together with dedicated lines so that files and notes could be exchanged easily. Both Group Notes and Personal Notes were modified to support intersystem links in 1978.

A notesfile can be linked between any number of systems. From a user's viewpoint, a linked notesfile is exactly like any other, except that the notice "Linked Notesfile" appears on the index display, and in the headers of some postings a system identifier appears after the author's user ID.

When a note or response is posted in a linked notesfile, it appears immediately in the local copy of the file, and is put in a queue to be broadcast to all systems which share that notesfile. The Notes software does its best to keep the file identical on all systems, although it can't guarantee that responses in a given chain appear in exactly the same order everywhere. There can be a delay of several minutes to an hour before a response is posted on linked systems (or even longer if one of the systems is down for an extended period).

Star Structured vs. Tree Structured Conferencing

Although Notes has evolved in many ways over the years, one thing that has never changed is the star structure of its notesfiles. One or two PLATO users wrote experimental versions of Notes using tree-structured (or "threaded") notesfiles, but most people who tried them found them hard to use and the idea did not catch on.

My own feeling is that a star structure is much more conducive to ongoing discussion. Human conversation is inherently disorganized, and a tree structure attempts to impose too much discipline. Conversations often tend to fragment and dissipate quickly in a tree. Some people seem at home with a tree structure, but in my experience more people find it rather foreign and overly complex.

With a star structure, each base note and its chain of responses resembles a conversation that we might have with a group of people gathered around a table. The conversation might drift or develop multiple threads, but if that becomes a problem, it is easily dealt with by simply starting new base notes to carry on divergent threads. A notesfile director can suggest this, but often the participants do it themselves.

Multiplayer Games

There are myriad games on PLATO. Some are for single players, but the most popular ones involve two or more players at separate terminals.

Games were certainly not a priority when PLATO was designed, but it turns out that its architecture supports multiplayer games superbly. The crucial features are:

Rick Blomme wrote PLATO's first two-player game in the late 1960's, a simple version of MIT's Spacewar. Possibly the most popular game in PLATO history is Avatar, one of several dungeons'n'dragons games. Empire, a multiplayer game based on Star Trek, is another favorite. Other multiplayer games range from Airfight (a precursor to Microsoft Flight Simulator), to Wordwar (a spelling and speed-typing game) and card games such as contract bridge.

Most games were written by unpaid programmers. The only reward they could hope for was the prestige of having written a popular game. Some game authors now receive royalties, but it amounts to only a few cents per hour of use, often split between a number of co-authors. A number of games that originated on PLATO have been recreated commercially as video arcade or personal computer games.

The On-Line Community

The sense of an on-line community began to emerge on PLATO in 1973-74, as Notes, Talkomatic, "term-talk", and Personal Notes were introduced in quick succession. People met and got acquainted in Talkomatic, and carried on romances via "term-talk" and Personal Notes. The release of Group Notes in 1976 gave the community fertile new ground for growth, but by that time it was already well established. The community had been building its own additions to the software infrastructure in the form of multiplayer games and alternative on-line communications. One such program was Pad, an on-line bulletin board where people could post graffiti or random musings. Another was Newsreport, a lighthearted on-line newspaper published periodically by Bruce Parello, aka The Red Sweater.

With the abundance of special interest notesfiles made possible by Group Notes, many on-line personalities developed. One of the best known was Dr. Graper (actually a student at the University of Delaware named David J. Graper). He began posting wild, surrealistic stories in a public notesfile where they were not exactly appropriate, but they were so hilariously entertaining that people clamored for more, and eventually someone created a notesfile called Grapenotes as a platform for his ravings.

The early PLATO community was concentrated in Illinois and consisted mostly of people in academia: educators turned instructional designers, and students hired as programmers. Later it grew to include more people from business, government, and the military as Control Data marketed PLATO as a general-purpose tool for training. It also grew geographically, spreading across the United States and around the world. The building that housed CERL became something of a Mecca to the far-flung PLATO community. Many people traveled to Urbana to see the lab and meet those of us who worked there. It was odd to meet people face to face after getting to know them on-line. My images of people based on their postings in Notes sometimes turned out to diverge wildly from reality.

The growing PLATO community also developed all of the problems that are now well known in on-line communities, such as flaming, men impersonating women as a prank, etc. Free speech was the general rule, but there were a few much-discussed incidents in which political postings in notesfiles were officially quashed for fear of jeopardizing PLATO's funding. Nobody on PLATO had ever experienced an on-line community before, so there was a lot of fumbling in the dark as social norms were established.

Over the years, PLATO has affected many lives in profound ways. So many real-life marriages have resulted from on-line encounters that such stories no longer seem remarkable.

Usage Statistics

The CERL PLATO system logged 10 million hours of use between September, 1978 and May, 1985 (a period for which the most complete statistics are available).

About 3.35 million of those hours (over one third) were spent in Notes. About 3.3 million messages were posted. By the end of this period there were about 2000 notesfiles.

No figures are available for time spent in Personal Notes, "term-talk", or Talkomatic. But some numbers are known for games. Avatar alone accounted for about 600,000 hours, and Empire claimed another 300,000 or so. All told, games probably accounted for about 20% of PLATO usage during this period.

Few statistics are available for the many Control Data systems, but none were as large as the CERL system. An educated guess is that CERL accounted for about 25% of all PLATO usage worldwide.

The numbers are incomplete, but it is probable that people interacting with other people represented at least half of all PLATO usage. This is remarkable considering that the designers of PLATO never envisioned that communication between people would play more than an incidental role.

The PLATO Diaspora

Control Data ran into serious trouble in the late 1980's, and sold or closed many of its businesses. At the same time, microcomputers were becoming a more cost-effective platform for education than PLATO with its mainframe-based architecture, and many of the Control Data systems were shut down.

Today the PLATO name is owned by Minneapolis-based TRO, Inc., but TRO no longer runs any mainframe PLATO systems. Control Data's PLATO has been renamed CYBIS. Control Data Systems supports about a dozen CYBIS systems at university and government sites. There might still be some former Control Data customers running PLATO on their own.

IMSATT Corporation, a Control Data spin-off based in Falls Church, Virginia, has recently begun offering a CYBIS-based service called Homer. It is targeted at home users and is available over the Internet.

At the University of Illinois, where it all began, PLATO has been renamed NovaNET. The U of I system racks up about 1.5 million hours of use per year, and is now operated by a private company, University Communications, Inc., of Tucson, Arizona.

The CYBIS systems still use the original PLATO Notes software. On NovaNET, a team headed by Dale Sinder rewrote Notes in 1991. Among the new features are multi-page notes and better search capabilities. But all of the key features of PLATO Notes, including the star structure of its notesfiles, have been kept.

Personal Notes has also been replaced on NovaNET. The new version uses a star structure to provide a level of organization that was never possible before. Each user's mailbox now looks and works much like a group notesfile, with the user as its director and write-only access for everyone else. The new Personal Notes also sends and receives Internet e-mail..

PLATO's Progeny

As an educational/multimedia system, PLATO has many offspring. Its most successful direct descendant is TenCORE, a DOS-based authoring system. Macromedia's Authorware, an authoring system for the Macintosh and Windows, is also firmly rooted in PLATO.

As a communication system, PLATO has numerous other descendants. Many people who experienced the on-line PLATO community were inspired to replicate it on other platforms.

Lotus Notes is the best-known example. It was developed by Ray Ozzie, Tim Halvorsen, and Len Kawell, all of whom had worked at CERL in the late 1970's. It would be an exaggeration to call Lotus Notes a clone of PLATO Notes, because Ozzie expanded the concept to include powerful capabilities that were never contemplated for PLATO. But many of its basic features were modeled after PLATO Notes.

Here are a few other descendants of PLATO Notes:

Computer conferencing is just now hitting the big time, not only with Lotus Notes, but with large consumer-oriented services like America Online and Prodigy, and more sure to follow.

Among the on-line services I have seen, the WELL has best succeeded in building a community comparable to PLATO's. Ironically, the WELL has its roots with EIES and Confer; as far as I know, its founders were unaware of PLATO.

But the WELL is an intentional community. PLATO was an accidental one which emerged spontaneously in an environment that had been created for other purposes. In 1970 few suspected that a human community could grow and thrive within the electronic circuitry of a computer. PLATO demonstrated that this is not only possible, but inevitable.


Acknowledgments

Many thanks to Al Avner, a veritable fount of statistics. Additional information was provided by Rick Blomme, Jim Bowery, Rich Braun, Greg Corson, Brian Dear, Sherwin Gooch, Mark Goodrich, Rob Kolstad, Dave LePage, Kim Mast, John Matheny, Dale Sinder, Joe Sneddon, Dan Tripp, and John S. Quarterman's book, The Matrix. Thanks also to John Quarterman for encouraging me to write this article.

Copyright © 1994 by David R. Woolley

David R. Woolley is a consultant and software designer in Minneapolis.

An earlier version of this article appeared in the January 1994 issue of the Matrix News.


This Issue / Index