During the past fifty years, I have been privileged to see the digital revolution courtside, as well as play the game myself. It’s a journey from physical media – punched cards, paper tape, and typewritten documents – to digital media, high-speed computing (megaflops, gigaflops, teraflops, and petaflops), high-speed broadband, powerful consumer devices, and globe-spanning AI data centers.
I wrote my first program – FORTRAN, naturally – in 1974, and the visceral thrill of seeing mathematical equations dance and move inside the machine, an IBM S/360 Model 50, hooked me for life. (See A Feeling for the Code.) Fifty years is a long time, and I am almost as old as one can be and have three computer science degrees. (See You Might Be A Computing Old Timer If …)
Motivated by the recent port of my blog (See HPCDan Ports Reed’s Ruminations (Or Don Quixote Rides Again)), I have been reflecting on the prodigious changes in how we create, process, and disseminate information. Herewith, are a few ruminations.
My Purdue Thesis Journey: An Analog Beginning
When I began graduate school at Purdue in the late 1970s, the computer science department still retained a professional typist on staff, and paper, rather than digital content, dominated. Armed with a trusty IBM Selectric typewriter and specialized typeballs for mathematics characters, the typist prepared paper drafts for the department’s faculty.
Graduate students, however, were on their own, both for intellectual content creation and its tangible output as a Ph.D. dissertation. Being “on your own” is a research litmus test; can you think creatively about new, interesting, and unsolved problems, using extant and, where necessary, newly created techniques. (See Research: When There Are No Anwers in the Back of the Book.)
To formalize thesis content, most graduate students hired typists to convert a combination of handwritten text and badly typed manuscripts into formal thesis drafts. It was a niche business for professional typists who could decipher semi-legible handwriting on arcane topics, and handle the idiosyncrasies of poor students anxious to graduate.
Armed with a unique, original document, graduate student supplicants then sought the blessing and imprimatur of the graduate school’s thesis office. There, a small staff carefully measured the margins of each page and verified the style of the figures, tables, and references.
It was an enervating ordeal, generating fear that almost rivaled that of the Ph.D. qualifying examination and the public dissertation defense. Many a student stumbled crestfallen from the graduate thesis office, denied entrance to Valhalla for failing to honor the sacred thesis format rules. Their penalty – retyping some or all of the thesis – before they could again seek approval. Should the thesis committee demand revisions, even more iterations might be required.
By comparison to these earlier generations of students, I was blessed. I still faced the dreaded thesis format check, but I was armed with better tools.
Newer Document Technologies Emerge
The Purdue University Computing Center (PUCC) housed a CDC 6500 and CDC 6600, which were the campus teaching and research workhorses via the Purdue Remote Online Console System (PROCSY). Meanwhile, to support academic computer science research, the department had just acquired a VAX 11/780 running an early release of the UC-Berkeley variant of Bell Labs Unix, which supported both nroff (line printers) and troff (typesetters).
As part of Unix, troff, written by Joe Ossanna and Brian Kernighan, was a command line tool where formatting directives were embedded in the document text file(s). It was (and is) a tool for geeks, written by geeks, and user friendliness was notably not a design goal. All these years later, it is still the core Linux tool for formatting manual (man) pages.
As an aside, I still have the July/August 1978 issue of the Bell System Technical Journal, which was dedicated to Unix. In the early days of Unix, this was the definitive reference.
Multicomputer Networks, troff, and the Versatec
In 1981, I began writing parts of my thesis draft in longhand, a practice I continued for many years before becoming feeling fully comfortable composing at the keyboard. Using a dumb terminal connected to the VAX, I then typed and formatted the draft using troff, which let me print as many drafts as needed. By using a set of macros whose output had been blessed by the graduate school thesis format gods, thesis format approval was straightforward. Nevertheless, I still faced challenges.
My thesis was about the then crazy idea of building supercomputers from large numbers of interconnected microprocessors. It’s worth remembering that at that time, the 16-bit Intel 80286/80287 was not yet released. A supercomputer built from very slow microprocessors was going to need lots of nodes. Not surprisingly, my thesis contained a substantial number of interconnection network diagrams, along with scalability analysis plots.
Alas, troff itself did not include tools for plotting data or drawing diagrams. A later tool, pic, could generate troff commands to draw simple line figures, but it was not available at the time. This forced me to draw the complex figures by hand, on a drafting board, in India ink. There was no room for error – one mistake or drip – and the figure was ruined. My officemates became accustomed to the cycles of quiet, intense drawing, followed by a loud expletive, the wadding of despoiled paper, and a frustrated toss at the trashcan.
The process may have seemed asymptotic, but it ultimately converged. Armed with hand drawn figures, the complex text, and thesis committee approval, it was at last time to print the final copy. For this, I relied on the department’s Versatec electrostatic printer. (This predated the availability of affordable laser printers.)
The Versatec was a finicky beast, prone to smearing when printing long documents. Starting around midnight, when I was unlikely to be interrupted, I would print 5-6 pages, break the printer down, clean it, then print another 5-6 pages. Sometime around sunrise, I finished printing the draft, separated the fanfold paper, and inserted my hand drawn figures at the appropriate locations. I was tired and frustrated, but relieved; the end was in sight!
After dozing in my office until the local stores opened, I walked down to one of the local photocopy shops, clutching my newly printed draft. Oblivious to the beautiful spring day, disheveled and sleep deprived, I placed the draft on the counter and asked for ten copies. I also made it unmistakenly clear that divine retribution – the Old Testament wrath of God – would be visited on the recipient, his or her family, and generations of their descendants should any harm come to the original.
As I recall, the recipient nodded nonchalantly. Nonplussed, only then did it penetrate my tired brain that I was the very embodiment of a recognized and well-understood stereotype – an anxious and agitated graduate student, one who had been paroled but still languishing in a holding cell.
From Paper and Hot Lead to Electrons
As I completed my Ph.D., thesis, though advocating for radical change in how we built supercomputers, I failed to realize that I was astride a bridge between two worlds, the old analog one of paper, typewriters, and bespoke computers, and the new one of consumer technologies and digital creation and dissemination.
One world was ending, and the early dawn of a new one was breaking. Had I been more prescient, I would have seen the obvious signs in my own life. I was drawing interconnection networks for massively parallel processors using India ink and a drafting board, yet I was also creating Mead and Conway VLSI circuits using graph paper and colored pencils.
How did it all happen? Let’s start with printing.
Modern Linotype typesetting machines are lineal descendants of Johannes Gutenberg’s original, 15th century movable type printing press, albeit faster and larger. With a brief technology refresher, Gutenberg would have readily understood their design and function. They were state-of-the-art up until the 1980s, when they were upended by digital technologies, and the concomitant shift to phototypesetting and offset presses.
In the brief interregnum of the late 1970s and early 1980s, that polymath, Don Knuth, who was unhappy with the quality of his phototypeset Art of Computer Programming, developed the TeX typesetting system and the Metafont description language for raster font specification. Soon, words and phrases such as kerning and Bézier curve entered the vernacular of an entirely new community.
TeX was soon followed by Leslie Lamport’s LaTeX, an easy-to-use set of macros – a markup language – for document preparation atop TeX. Easy-to-use is, of course, a relative term. Although LaTeX is now the de facto standard for scientific writing, it never gained the wide adoption of WYSIWYG tools like Microsoft Word.
Roughly concurrently, Adobe developed the PostScript page description language, which was licensed and used by Apple on the Apple LaserWriter, a peripheral for the Apple Macintosh.
Suddenly, the pieces of Doug Engelbart’s “Mother of All Demos” – WYSIWIG document preparation and digital printing – became consumer technologies. The digital printing revolution was now well underway, followed quickly by the now ubiquitous portable document format (PDF).
When NCSA Mosaic was released in 1993, the final puzzle piece quickly fell into place. Born digital materials could be processed and shared globally, and printing as we knew it for the past century was no more. In so doing, we unleashed a global digital village, one whose societal effects are not yet fully understood. (See Contemplative Reflection and Instantaneous Communication)
In retrospect, it all happened quickly – in a little more than a single decade. That’s how revolutions occur – slowly, then all at once. (See Reluctant Revolutionaries, the Trolley Paradox, and Ender’s Game.)
My Illinois Transition and the Digital Revolution
With a freshly minted Ph.D., I was delighted to join the computer science department (now the Siebel School of Computing and Data Science) at the University of Illinois as a young assistant professor in 1984, after a year at UNC-Chapel Hill. Not only was Illinois one of the birthplaces of supercomputing, (See Stored Program Computing: Ideas and Individuals Do Change the World), it had (and has) a rich set of computing infrastructure. In 1984, these included a set of DEC VAX 11/780 and 11/750 computers, as well as a set of AT&T 3B20 and 3B2 computers. I was well acquainted with the VAX, having used it extensively as a Purdue graduate student.
The 3B20s were originally designed to support the 5ESS telephone switching system, with multiple redundancies for reliable telephone switching. Among their useful idiosyncrasies was a Delco car battery as a backup power supply. One snowy evening, my colleague and collaborator, Roy Campbell, slipped into the computer room in the Digital Computer Laboratory and (temporarily) borrowed the Delco backup battery to jumpstart his car. (The breakup of Ma Bell had just occurred, though the profound effects on telephony and networks were yet to be felt.)
As valuable as a warm battery might be on a cold Illinois winter evening, I was more interested in the 3B2s. They had a graphical interface, allowing me to finally use a GUI and pic to draw my interconnection network diagrams in the comfort of my faculty office!
That 3B2 transition was short-lived. Soon I had a SUN 3/50 diskless workstation with a black-and-white raster screen in my office, and a laser printer next door. (I had hair too, but that’s a story for another day.)
I wrote my first book with Richard Fujimoto using troff, combining his insights on node design with mine on interconnection networks. At the urging of my first graduate students, I abandoned troff and began writing papers using LaTeX and previewing them digitally.
With the founding of the National Center for Supercomputing Applications (NCSA) and Cray X/MP access, and NCSA and its partners driving NSFNet for remote access, I and many others were soon connected to the world. Just a few years later, the NCSA Mosaic web browser, developed by Illinois undergraduate Marc Andreessen and NCSA staff member Eric Bina, launched the dot.com boom.
The web was simple then, with static links, and I created my first website using just a text editor and raw html. (See www-pablo.cs.uiuc.edu sometime in the 1990s.) on the Internet Archive. Cascading style sheets (CSS) were just a proposal, and Netscape, founded by Jim Clark, Marc Andreessen, and a group of Illinois people, was just imagining what became JavaScript.
In partnership with NCSA, I was fortunate to develop some of the first World Wide Web performance analysis and web analytics tools. All this was at a time when the NCSA web server was the busiest in the world. Charlie Catlett and I also identified network protocol processing bottlenecks in the NCSA web server cluster, leading to part of what we now call round robin DNS.
And that crazy idea of building massive computing capability using lots of microprocessors? It turned out to be a good one, both at NCSA, where we deployed the first terascale Linux clusters for national use, and around the world. (See Exascale Software: Just a Few Orders of Magnitude) Today’s supercomputers, the cloud revolution, and our world-spanning AI infrastructure all rest on large numbers of interconnected microprocessors. (Thank you Charlie Catlett and Larry Smarr.)
Coda
From my time as young graduate student in the 1970s/1980s, as a professor and then head of computer science at Illinois (the 1980s/1990s), through leading NCSA during the construction of large-scale clusters for research computing (the 2000s), launching the Renaissance Computing Institute (RENCI) to address societal issues (the 2000s), and building energy efficient data centers at Microsoft (the 2000s), to global science and technology policy created by extraordinary computing advances (the 1990s/now), I have been fortunate to have a courtside seat to the digital revolution – documents, the web, supercomputing, and the cloud. (See Dan@Microsoft, End of an Era, and NCSA@30: Reflections on the Revolution.)
I am now watching the next great revolution – AI everywhere – with unabashed interest. (See Mediations on AI: History, Hype, Myth, Reality, and Futures)
Please leave a comment …