Today, essentially all of the American Physical Society’s publishing operations are electronic. The publication of record for a paper is now its posting on the web, not its appearance in print on a journal page. All communications among editors, referees, and authors are electronic, usually via the web. It must be conceded though that there are still situations in which paper is definitely to be preferred.
The transformation of physics publishing, which still continues, as more products and services are offered to the scientific community and to the general public, began in the middle of the last century. Sam Goudsmit’s creation of Physical Review Letters in 1958 as a publication separate from the Physical Review and its Letters to the Editor section was a significant event and not just in the way in which research was reported to the scientific community  . PRL also represented a change in publishing technology.
Traditionally, journal pages were prepared by skilled craftsmen who worked at expensive, hot-metal or monotype print shops. PRL pages were created using standard IBM typewriters and special hanger-key devices that allowed the insertion of the many extra characters associated with typical physics articles. Subscripts and superscripts were inserted by rolling the typewriter platen. Then the single-column galleys were pasted onto boards together with appropriately sized figures and tables to create the well-known two-column format of the journal. The most obvious difference from the previous technology was that columns of print were no longer right justified. For many years the final step in preparing the weekly issues for the printer occurred every Wednesday morning, when PRL Editor George Trigg, surrounded by members of the production department, would carefully examine every page and send people scurrying off to fix whatever infelicities he discovered.
By the late 1960s the move away from conventional printing into typewriter composition continued with the composition of Physical Review A by a commercial company under the watchful eye of Editor Kenneth Metzner. At the same time Physical Review C, under Editor James Roesser and Assistant Editor Margaret Judd, began running on a night shift using the PRL facilities. (My involvement with these efforts began when Goudsmit hired me in mid-1969 as Editor of Physical Review B. By mid-1971 I became responsible for most editorial-office operations.)
In parallel with these projects, but at a much more fundamental level, tools were developed to enable the journal editors to manage the explosive growth of submissions. In the early 1960s referees were selected from the journal’s editorial boards, from physicists at Brookhaven National Laboratory (where the editorial offices were then located), and from the annually published APS membership directory. The membership directory list was updated by hand when referees moved from one institution to another. By the end of the decade, every week editors would receive inches-thick lists of referees sorted by journal and by name, printed on 11 inches×14 inches computer paper. Data for the underlying databases were prepared by Janet Watson (still working at the editorial office) on IBM punch cards and carried daily to Brookhaven’s big CDC 7600 computers for processing. Use of those data made possible the (almost) automatic generation of forms, mailing labels, and other materials needed to send papers for review. To complete the editorial handling process, the only major task left was correspondence with authors.
Two specially designed keyboards were built in-house during the early 1970s, to be used in conjunction with a small HP 2116A computer and a tape drive. That system allowed manuscript titles to be entered and stored in final publication form, including displayed math and alternative fonts, such as foreign characters, which could be set as subscripts or superscripts. Those capabilities enabled the composition of all the semiannual Physical Review indexes. Correspondence, which included manuscript titles in a very readable format, also became possible with the use of a Teletype Model 37 KSR teletypewriter with an expanded character set. At this time the APS Executive Committee raised a cautionary note—that although the APS was willing to support staff members interested in computer composition, it was not the intention of the APS to become a leader in major developments. In other words: Don’t expect significant increases in staffing or in funding. That view had changed radically by 1976.
During the 1970s Kenneth Thompson and Brian Ritchie of Bell Labs were developing UNIX, a system that could handle not only the database requirements for APS publications but also provide tools for computer-assisted photocomposition. Both B. Chalmers Frazer, the journals’ Managing Editor, and Joseph Burton, the APS Treasurer, who was at the time also Director of the Physics Research Lab at Bell Labs, were enthusiastic in their support for incorporating UNIX into APS operations. Because of the support of Frazer and Burton, the development of comprehensive programs and procedures for handling all editorial actions could proceed, as could exploration of the technical and economic feasibilities of producing the journals via computer-assisted photocomposition. The APS Council authorized the project in 1976. Once the Council acted, APS became a major player in the development and incorporation of computers into scientific publishing. They were, and still are, exciting times.
The first completely photocomposed paper, written by P. L. Taylor of Case Western Reserve, appeared in Physical Review B in the 1 April 1977 issue. The first completely photocomposed issue of the journal appeared a year later—Physical Review B, 1 April 1978; over 250 papers on 2000 pages were successfully photocomposed in that year. After evaluation by an external panel, a full-scale production test was initiated in October using an American Institute of Physics production group under the supervision of Margaret Judd, the APS Editorial Systems Manager. Once that test was successfully concluded, the Council, in 1980, approved the production of all APS journals by means of UNIX-based computer-assisted photocomposition. The transition from the APS production operation to AIP’s Woodbury production facility was completed by October 1981. A year later 34,000 pages were produced. Because of the requirements of exceedingly tight schedules and high reliability of hardware and software, the conversion to the new composition system of PRL, which remained under direct APS control, was not implemented until the 4 March 1984 issue.
Major expansions of electronic activities within the APS Editorial Office began in the 1980s, as did communications among scientists. Study of the feasibility of the submission of manuscripts in machine readable form by authors began in 1980. The replacement of UNIX’s troff by Donald Knuth’s TEX in 1989 radically altered the mode of submission of manuscripts to the Editorial Office.
Electronic communications via Bitnet were introduced in 1986 and were running at about 120 messages per month. By 1991 nearly 15,000 referee reports were received via Bitnet or Fax. The preservation of the input tapes from previous photocomposition projects provided impetus for the creation of PROLA and the subsequent move of all activities to the World Wide Web. The distance traveled from IBM punch cards and an HP 2116 A to the present capability is quite astonishing.