The roots and history of what is became the Computing Science Research Center, before October 2005 locally called Center 1127, go well back into the past. At that point, it was dissolved, with the members going into several groups. For many years, though, it was among the most stable organizations in Bell Labs; even the numerical name remained moderately constant, though some variations are noted below. Management and population, of course, have been coming and going over the years, and likewise our dominant and peripheral research interests have varied, but overall retained a consistency of theme.
Center 127, as a specifically Computing Science research organization (and as it was originally numbered), began as a split-off, in 1960, from a more general Mathematics Research organization (12) within Bell Labs as the increasing importance of computers became evident to the company. Our numerical names were a digit shorter then: Research was just Area 1!
A variety of places, including Lucent's history site, have information about the company's work in CS-related areas. Probably the most comprehensive is in the multi-volume A History of Engineering and Science in the Bell System, especially the volume subtitled Communications Sciences (1984; ISBN 0-932764-08-1).
The relevant section of this book was initially prepared by B. D. (Bernie) Holbrook and W. S. (Stan) Brown, department heads during the 1960s and 1970s. A draft version of the chapter was also made available as CS Tech. Report #99. CSTR 99 has been revived (so far without its figures), and is available in either PDF or PostScript formats; an experimental version browsable in HTML is around too.
It is comprehensive about the history through the 1970s; the book version is more expansive on somewhat later things, especially about Unix, through the efforts of Doug McIlroy.
Doug McIlroy presented a talk on Bell Labs' computing history from his point of view; an audio version of the talk is available. It is just over an hour, and is occasionally hard to make out, but it is fascinating. Again, it distinguishes only in passing our own center's contributions. One of the interesting parts is near the end (about 56:00 into the presentation, where he mentions some of the programs that really surprised and delighted him by their freshness of approach: These included BLODI (Vyssotsky and Lochbaum), an early dataflow language for specifying signal-processing operations; Eqn (Kernighan and Cherry), and he later mentioned the more general Software Tools approach of which this was a precursor (Kernighan and Plauger); Typo (Cherry and Morris), which found likely typographical errors in a paper using only trigraph statistics; Struct (Baker) that rewrote spaghetti Fortran into well-structure Ratfor; Grampp's work on (non-publicized) programs and scripts for automatically detecting security flaws in typical Unix setups; and Ken Church's work on aligning parallel texts in different languages.
The general histories of the contributions of Bell Labs to the beginning and the flowering of the company's CS research appropriately do not discuss organizational structure, but it is interesting for historians of our long-lived organization to peek into this. In these notes I use the term "director" for the person who led the center and ignore the title-inflation of recent years.
It appears that John R. Pierce was originally named to head the new center 127 when it was created, but this appointment was reversed very rapidly, and the first actual director was Henry Pollak during 1960-1962; he also headed the existing Math center. Pollak was succeeded in turn by Edward E. David, whose background was in acoustics. In 1965 David was promoted to Executive Director of a new division 13, and we moved into it; thus 127's numerical name was changed to 137 under David as both director and ED; this separated us organizationally from the Math Center, which remained in 12 under Pierce. Shortly after, still in 1965, Thomas H. Crowley was promoted to became our director. He remained until 1967, when he moved to Whippany (with visits to Kwajalein Atoll) to work on the anti-missile project, then known as NIKE-X, later as Safeguard.
At Crowley's promotion, Samuel P. Morgan moved from the Math Center to become 137's director. A bit later (in 1972, following Pierce's retirement) a reorganization changed the name back to 127. At this point a structure was reached that remained reasonably stable for quite some time, with the Math/Statistics Center, the Acoustics and Behavioral Research group, and the Computing Sciences Research centers under division 12.
Morgan remained as director until 1982, when he stepped down in favor of A. G. (Sandy) Fraser. Sam thus can claim the title of longest-serving director, as well as presiding over an especially productive period for the center as a whole. His own productive research career continued for many years after leaving management, through 50+ years of employment.
Fraser stayed as director for several years (until about 1988) when he was promoted to ED, and then left to head up the AT&T Research Labs in the 1995 Trivestiture. Fraser's replacement was Al Aho. Aho remained in the position for a couple of years, then left to work at Bellcore, though he confessed at the time that his ultimate destination was academia; this seems to have been achieved, through a route more convoluted than he had planned.
After Aho left, a "gang of 5" Department Heads collectively managed the center for some months until Ravi Sethi emerged as the obvious choice for Director, which he then became.
Sethi in turn was promoted to Executive Director, and after a brief interregnum, Bill Coughran emerged to replace him in 1995.
Coughran, in turn, was selected (1998) to head the sadly ill-fated Bell Labs Silicon Valley research lab.
For a period after Coughran's move to California, Sethi was again nominal director, although he had to share his time by serving as Executive Director for us and other centers.
By this time, Al Aho, having left Bellcore and achieving academia by becoming head of the CS department at Columbia, was induced to return to Bell Labs and then (in what amounted to a job-swap with Sethi) became Director by 2000.
In 2002, Al returned to Columbia, this time aiming to teach and write (and not be their department head, he's happy to aver). Wim Sweldens, amid general relief in not enduring another interregnum, became Director (in the old language) and entitled Vice President in the new; he moved sideways (and up?) in the terminal reorganization.
The Center has (seemingly from the start) contributed in a variety of areas. Historically, the first three seem to be the longest-lived; they are also so general that much CS research can be jammed into them one way or another.
However, it turns out that all of these were very much intertwined, and it's somewhat problematical to classify either the projects or the people. Many of us have simultaneously or successively worked in several of the areas. For example, is the work of Aho, Sethi, Ullman on parsing part of "languages" or algorithms? At what point does data networking shade into systems?
At the same time, there were some areas in which not much effort was spent: although there were connections and publications on databases, this area tended to be investigated from the algorithmic point of view, not in creating database systems. Some venues were avoided, e.g. artificial intelligence (unless you count the Thompson/Condon work on chess, which they don't consider AI).
In the early 1960s, the local systems research centered around the internally developed series of BESYS operating systems for the IBM 704 through 7094 machines. The work on BESYS was evidently scattered around Bell Labs, and it predates the center's formation. In 1964, E. E. David, together with Vic Vyssotsky, seem to have formed the main local force behind the collaboration of Bell Labs, MIT, and General Electric in forming the Multics project. This led to the replacement of several of the large IBM systems at the Murray Hill and Whippany computer centers by GE gear; Holmdel and the new Indian Hill lab remained committed to IBM. This Multics work, even at Bell Labs, was distributed. Perhaps the most prominent local contribution to Multics as such was the EPL (Early PL/I) compiler by Dogu McIlroy and Bob Morris, although there was in the later 1960s much other Multics-oriented work; an adaptation of BCPL to Multics led more or less directly to the B and C languages. Similarly, Multics experience also was influential on Unix.
During the 1960s, the service facility providing general computation resources for Murray Hill remained organizationally under the control of research management. The support required significant resources: the mainframes were large, and needed operators who arranged to read cards, tend printers, put output in boxes. There was also significant programming support and counseling of users, particularly in providing advice about numeric computation.
In 1969, two crises occurred: it was decided that Multics would not live up to its promises and would be dropped, and also that direct support of general computation for Murray Hill (and likewise other locations) didn't really belong in the research organization (127/137) where it had been centered, and thus a separate Murray Hill and Whippany Comp Center was created as a service organization. Morgan has retained the organization charts from his years. Two of them are reproduced here.
Morgan continued as co-director of the two groups until the end of 1970.
As of summer of 1969, the center's technical population numbered only 32, with 3 assistants. Many of us then were still from the original Math Center; others had joined from other parts of the Labs after the center's creation, but still others were fresh hires. This population was what remained after the separation of Comp Center support from 137 plus various movements within the company and and to the outside.
The departure of Multics was certainly one of the sparks that ignited the late-60s and later work on Unix; this work continued well into the 1980s, and is documented elsewhere; there is a collection of papers about this. After the 1980s, systems work was concentrated mainly on the Plan 9 system and its spinoff Inferno.
In the language area during the earlier 1960s, research contributions include McIlroy's work on macro-processors for assembly languages (BEFAP). Later work in the center would encompass the B, C, then C++ languages, and later (in a post-trivestiture ingraft) SML/NJ. In parallel with the Unix effort during the 1970s was influential work on "little languages," software tools for text processing, and computer typesetting.
The algorithmic tradition continues, most visibly in recent years in widely used tools and techniques for verification of program behavior.
Computational geometry has turned out to be important for wireless networks in a variety of settings, as well as in the numerical realm.
During the 1970s and after, a small but influential group contributed to CPU architecture, beginning with Fraser's experiments with C-oriented machines, and culminating in the Crisp and Hobbit microprocessors, by the 1980s incorporated into AT&T computers. Also in the 1980s, a series of innovative bit-map terminals were designed, again yielding products. This work combined aspects of distributed systems, graphics interfaces, and networking.
Telephony as such was also done; one group built our own software for TPC, a POTS-type switch based on the Dimension PBX hardware. Its design influenced the architecture of the 5E ESS system. Later, Pathstar, an Inferno-based phone switch using IP technology became a Lucent product.
There are many more that could be detailed; a partial list includes design and analysis of digital and analog circuits (CAD tools), optical character recognition and classification, graphics rendering, even computer chess.
The help of Sam Morgan (especially) and Doug McIlroy has contributed immeasurably to this account in particular, but also to the life of the center over the years.
Copyright © 2003 Lucent Technologies Inc. All rights reserved.