Text: Burton R. Pollin, “Programming the Book,” Dictionary of Names and Titles in Poe's Collected Works, (1968), xxxvii-xl (This material is protected by copyright)

[page xxxvii:]

PROGRAMMING THE BOOK

Gary Berlind

As a result of his prior experiences with computer-produced publications, Dr. Pollin is the possessor of considerable knowledge about the capabilities and limitations of computing machinery. He was therefore able to make his programming needs known to me as efficiently and clearly as could be desired by any programmer. Thus, I was free to concentrate my full effort upon the design and implementation of an efficient data-processing system for the production of this index.

Input consisted of approximately 7,000 keypunched cards as described by Dr. Pollin in his introduction. It should be noted that all index entries had been collected and sorted manually, so that my system was required only to provide a quick and inexpensive method of formatting and “typesetting” the pages from which the present volume has been photo-offset.

The Flowchart on the next page represents the actual system as developed.

The cards originally existed in two separate collections, “name cards” and “title cards.” Via a card-to-disk program the data was transferred to two distinct direct-access files, “CARDS 1” and “CARDS 2.” All further processing was done by using these files, thereby eliminating cumbersome card handling.

The main program in the system was the Formatting Program which, using CARDS 1 and CARDS 2 as input, produced the six indices. Input parameters to this program included specifications for page width and page length, in order that each index could be produced in an appropriate format. A “selectivity module” was also incorporated, making possible the selection of index items for the four selective indices. [page xxxix:]

Because the IBM 029 keypunch has no provision for the punching of lower-case alphabetic characters, the Formatting Program was responsible for the conversion of the input text to upper- and lower-case according to the following convention: all alphabetic characters were automatically translated to lower-case unless preceded by a dollar sign ($). This provided for the capitalization of single letters; two other conventions were used to allow for the capitalization of entire words. The first, used primarily for single words, required the placement of a double dollar sign ($$) before the word. The second, of value when extended sequences of words were to be printed in upper-case, required that such strings be enclosed by the symbols < and >. Extensive use of the latter convention was made in the keypunching of book titles.

Other functions performed by the Formatting Program included the elimination of extra blanks in the input, the “blocking,” within index entries, of page references by volume, and the generation of guide-letters at the top of each page.

Six passes with the Formatting Program were made, thereby producing indices one through six. These were not printed immediately, but for reasons that will soon become evident were placed in direct-access storage files. Via a proof-listing program, printed versions of the indices were produced (COPY 1 ...COPY 6). These computer-generated page-proofs were nearly identical in appearance to the pages of this book, the primary difference being that each listing contained line-numbers running down the left margin of the page. Dr. Pollin and his aides then proof-read the indices in a form closely approximating that which would ultimately be produced by the system.

Corrections, made on the printed proof-listings, were transferred to the indices themselves via an on-line editing program [page xl:]operating through the computer's system console typewriter. The program permitted the real-time retrieval of any line of any index, and the modification of any part of that line. With some practice Dr. Pollin acquired a proficiency at the editing keyboard, and before long all six indices were declared to be ready for final printing. A final-printing program, which added page numbers and headings (while suppressing line numbering) produced the final printout.

The statistics reported on page were produced by a specially written program that serially scanned through indices one and two and kept count of the various kinds of index entries in those files.

All programs used in the system were written in OS/360 Assembler Language. Processing was done on an IBM 360 Model 30 currently operating at the New York University Heights Academic Computing Facility. The computer has a high-speed memory comprising approximately 65,000 characters. External storage devices consist of three 2311 disk-drives and a 2321 data-cell of which we used only a small portion of its 400,000,000 character capacity when we once needed extra storage space.

The card-to-disk program processed cards at card-reader speed, which is approximately 1000 cards/minute. The Formatting Program generated indices one through six in approximately 10, 8, 4, 3, 3, and 4 minutes, respectively. The proof-listing and final-printing programs operated at printer speed, which with the extended 120-character print chain is about 240 lines/minute. It took close to 6 minutes for the statistics program to glean its information from indices one and two.