SPECIAL PURPOSE COMPUTER FOR HIGH-SPEED PAGE COMPOSITION

Transcription

1 A SPECIAL PURPOSE COMPUTER FOR HIGH-SPEED PAGE COMPOSITION Constantine J. Makris Mergenthaler Linotype Company, Division of Eltra Corporation Brooklyn, N ew York INTRODUCTION For several years now, computerized page composition has attracted the interest of workers in the printing and data processing fields. Considerable work has been conducted in the past in both fields by various concerns with the primary goal to speed up the laborious work of page makeup and thus reduce the page makeup CQst. Our studies in the page composition area have shown that page makeup can be implemented by either one of two methods-namely, page composition employing the stored program flexibility of a general-purpose computer or page composition executed by a computer specifically designed for that purpose. We adapted both methods, for each one in itself presents attractive features depending upon the application and the nature of the matter which is to be processed. This paper discusses only one phase of the Linotron System; specifically, the typography program of page formatting as executed by a special-purpose computer called the "Page Formatting Computer." During the system concept period, the philosophy 137 of this system design was guided by three primary objectives: 1. High-page composition machine throughput. 2. Flexibility of machine to handle a large variety of tabular and text formats. 3. Machine simplicity to insure smooth man-machine interface. The first two sections of this paper present the Linotron System and its organization in the page typography program and a functional synopsis of the characteristics of the phototypesetter. The succeeding five sections discuss the organization of input material to the page typography program and present various definitions and problems encountered in the process of page formatting. The last section presents the internal organization of the page formatting computer and describes and brings into focus the major characteristics of the system. LINOTRON SYSTEM The Linotron (a Mergenthaler-CBS Laboratories development) is a high-performance phototypesetter

2 138 PROCEEDINGS-FALL JOINT COMPUTER CONFERENCE, 1966 with x-y plotting features that produces high-quality lexical-graphical matter under the control of a properly programmed magnetic tape. The formatted tape which drives the Linotron phototypesetter is the product of the typography program performed by either a general-purpose or a special-purpose computer. This discussion is focused on the manner with which the Page Formatting Computer (PFC) is organized and tied in with the overall page composition and typography system. Figure 1 shows a simplified diagram of the Linotron System. The basic task of the PFC in the overall Linotron System is to receive from magnetic tape edited lexical-graphical matter. Then the PFC by means of its computational capability properly composes and sets this matter on the page according to prescribed page formats of good typography. The system outputs this formatted information on a magnetic tape medium with all the necessary instructions to drive the Linotron Phototypesetter. THE LINOTRON PHOTOCOMPOSER Figure 2 shows the block diagram of the Linotron Photocomposer and the flow of data through the system. It is conveniently subdivided into seven large functional subsystems: 1. Magnetic tape recorder 2. Tape control and buffer storage 3. Control logic 4. Character generator 5. Cathode ray tube display system 6. Positioning servo 7. Recording camera Magnetic Tape Recorder The magnetic tape read transport reads the control tape which is the precipitation of the page formatting program. On this tape are recorded codes of the characters to be typeset with the manner with which they will appear on the copy. In addition to the character codes, character positioning and system control codes have been properly recorded to insure proper functional operation of the system. Data can be furnished to the system in either 6, 7, or 8 level codes at 556 bpi. Control Logic This is the section of the machine which reads the information from the input core buffer and sorts the character codes from the positioning and/or control codes. It routes the character codes and character size codes into the character generator block; and the positioning codes into the positioning regis- PROOF READ EDIT INSERTION PROGRAM TYPOGRAPHY PROGRAM PAGE FORMATTING COMPUTER LlNOTRON PHOTOTYPESETTER FINAL COpy Figure 1. Simplified block diagram of Linotron system.

3 A COMPUTER FOR HIGH-SPEED PAGE COMPOSITION 139 IN PUT MAG. CHARACTER CATHODE RECORDING OU.TPU T TAPE GENERATOR ~ RAY TUBE - CAMERA - READER ~,, ~ l, r,.. TAPE CONTROL - CONTROL - POSITIONING & BUFFER. LOGIC.. SERVO - - Figure 2. I Block diagram of Linotron Phototypesetter. ters. It also generates control signals from the servo block, cathode ray tube assembly and recording camera assembly upon the recognition of control codes. Character Generator The function of the character generator is to receive the character and character point size code and in return generate a high-quality video signal of the character. The character is selected from an assembly of 1024 characters whose typographical configuration is stored in four film plates (grids). The heart of the character generator is the Linotron tube-a single-envelope, high-quality tube which transforms the light image of characters focused on its photocathode into a character video signal of a given size. ~ The generation of the character video signal is accomplished by optically focusing the entire complement of a character grid (256 characters) into the photosensitive cathode of the tube. The resulting electron image beam from the photo cathode is imaged onto an aperture plate by means of magnetic lenses and then wobbles back and forth at the aperture plate. A small aperture of inches is associated with each of the 256 character images at the aperture plate. As the images are wobbled back and forth, the aperture's scanning of the image is accomplished as in the image-dissector video camera tube. The electrons which pass through the apertures are all trapped except for the aperture that scans the selected character. This is accomplished by means of a bar matrix (16 vertical X 16 horizontal) that lines up with each aperture and is biased with negative voltage. The control logic places positive voltages on the selected x-y bars that line up with the desired aperture. The electron beam in the form of pulses is amplified by means of the multiplier section of the tube thus producing a video signal. Cathode Ray Tube The cathode ray tube assembly consists of a high-resolution, high-intensity CRT and deflection system. The electron beam of the tube which produces a spot of less than on the screen sweeps into synchronization with the scan rate of the Linotron tube. The character video signal is utilized for unblanking the Z axis of the tube thus the output screen exhibits an accurate reproduction of the original character. Size of the characters can be changed by changing the amplitude of the sweep. This is controlled automatically by the control logic when a point-size change has been called for by the tape. Positioning Servo The precise character positioning on the screen face of the CRT is accomplished with the aid of the positioning servo, which makes use of a second CRT, the spot position of which is accurately referenced to a set of precision-ruled gratings. Initially, the horizontal and vertical coordinates of character position are obtained from the control logic and used to position the spot close to the correct character position. Then the servo circuitry brings the spot precisely to the desired position-the precision gratings

4 140 PROCEEDINGS-FALL JOINT COMPUTER CONFERENCE, 1966 being used as a reference-and keeps it there until the character is printed. The CRT display, which is actually producing the character, is slaved to the positioning servo tube. Its beam accurately follows the beam motion of the servo tube. Thus, the positioning of the character is accurately controlled by the precision servo. Recording Camera The full composed page displayed by the CRT is photographed onto the film or paper by the recording camera. The film of the camera advances frame by frame (page length) as instructed by the logic control block when it recognized film advance code. FORMAT ANALYSIS AND DEFINITIONS OF PAGE FORMATTING A study and anal:'sis of a large variety of page formats was a prerequisite to the design of an efficient page-formatting system. This study disclosed the following fundamental page typesetting characteristics: 1. The formatting program of any new page format does not necessarily have to be rewritten from scratch. 2. The page makeup program consists of a number of distinct typesetting operations, such as line justification, column justification, etc. Each typesetting operation required in the page makeup can be split into a number of small independent routines, each one of which is programmed to accomplish just a fragment of the overall program of the typeset operation. 3. The program of any page makeup can be accomplished by the proper collation of independent typeset microroutines. 4. Each page format contained its own format parameters which have to be inputted and stored as program parameters. 5. The editing process could be minimized to just a simple data item labeling. The following page formatting definitions had to be developed as format analysis was taking place: 1. Format: The unique spatial layout of the page items (such as columns, fields, paragraphs, etc.) within the page, and the specific sequencing of the typesetting routines within the item in the page as well as the specific sequencing of the routines between one item on the page and another item on the same page. 2. Typesetting Routines: Format-independent, self -sustained routines which implement a defined fraction of a typeset operation program. 3. Format Program: The program which sets the complete page format by means of selecting and sequencing the proper typesetting routines. 4. Format Parameters (Program Parameters): The spatial parameters of the format (such as, X, Y marginal dimensions, line measures, etc.) and typographical parameters of the format data (such as typeface, point size, etc.) 5. Edit Instructions (Page Item Identifiers): Labels inserted before page items which identify copy blocks of identical spatial, typographical and typesetting characteristics within the format. Figure 3 shows two pages of a representative text format. CLASSIFICATION OF INPUT DATA The source data input to the PFC has been classified as (1) edited lexical-graphical input data and (2) page format parameter data. Appropriate format parameters had to be defined for each format on hand which had to be applied to the input and stored as program parameters. The format parameters have been divided into two categories: (1) the typographical parameters of the format that prescribe the style of type and size for each page item and (2) the format dimensional parameters. Since the format parameters had to be changed occasionally at the command of the typographer or page format designer, they had to be generated separately and stored in a separate tape-the parameter library tape. It is noted that the bulk of the typographical format parameters consists of the character widths for each typeface at different point sizes. For example, for 1024 characters that the Linotron Phototypesetter handles simultaneously and for 8-point sizes, there are 8192 character widths with a byte of 3 BCD digits maximum. Assuming 2 BCD digit storage per core locations in a character width look-up

5 A COMPUTER FOR HIGH-SPEED PAGE COMPOSITION 141 ~44. Introduction I I Cha..1)ter 3 ~SUPPORT I 0---1~.II~se=c:fitii;on;Tl.lGiiE~NiiEi'RAiiLLII Combat, combat support, and combat service support units are provided to the forward infantry, mechanized, or armored brigades and battalions as required to assist in the accomplishment of the mission. These units ma be organic, attached, in support of, or under operation control of the brigade or battalion. For the purpose of this chapter only, those units normally assisting the ~CONTENTSI General Tactical Air Support Artillery Support Chemical Support....4l) Engineer Support Ground Transport Support General This section generally covers organic normal supporting units of mechanized infantry and armored brigades. Nonorganic combat support units available to brigades in the support role include tactical air support; Army aviation; and artillery, chemical, engineer, and ground transportation units. An appropriate number of mechanized infantry battalions and tank battalions are attached to the brigade headquarters according to the operation plan. 47. Tactical Air Support a. General. The flexibility and long-range striking power of tactical air makes it an important means of destroying the enemy. Superiority in the air, or at least relative freedom of action, is a predominant factor in securing success in desert operations. Tactical air power has three general missions: gaining Section II. COMBAT SUPPORT brigade or smaller units in combat will be covered. 45. Other Specific References For a detailed discussion of combat support of infantry and armor elements, see the FM's of the 7- and 17 -series. A detailed description of combat service support functions is contained in chapter 3 (secs. III, VI, and VIII) and appendix II, FM air superiority, interdicting the battle area, and providing close support. These are inherent in joint air-ground operations and apply equally to desert operations. Since desert areas produce little~.llj~~~""'''''1iiio military forc~-""'" position and then position after dark. This causes the enemy to concentrate his efforts on a false position while friendly units will have moved to the primary position. 2 Retrograde operations are characterized by detailed centralized planning and decentralized execution. Communications and control become increasingly difficult. Subsordinate commanders must have detailed knowledge of the overall plan so that they may properly conduct independent actions when communication with higher 01' adjacent units is lost. Retrograde planning for desert operations is influenced by desert terrain and its effect on the mobility of the force concerned. Lack of obstacles and barriers dictates a speedy, organized withdrawal. b. Dispersion in the desert is greater during daylight hours than at night. After dark, dispersion is less than in daylight hours. This causes the enemy to concentrate his efforts 1 Typical of these fortress.type defenses were the coastal harbors or Tobruk, Bardia, and Mersa Matruh in North Africa during WO"ld War II. EI Alarnein, due to its right flank being on the sea and anchored on the left by an impassable (to vehicles) salt marsh (the Qattarra depression), was also considered a fortress. 15 Figure 3. Representative text format.

6 142 PROCEEDINGS-FALL JOINT COMPUTER CONFERENCE, 1966 table, there is a 16,384 of maximum core requirements plus time consumed by the program to address and retrieve the character widths. To avoid this core storage and time requirement, the character width parameters have been separated from the parameter library tape and a flexible circuit logic block has been designed which would accommodate the character width parameters with fast access and minimum cost. ORGANIZATION OF INPUT EDITED DATA The edit program of the system is the program where raw data together with edited instructions are generated directly from the keyboard. The purpose of this program is to provide properly edited, unformatted data on paper tape or magnetic tape. This data is utilized as input to the formatting program. The edit instructions have been made simple in order to allow some flexibility to the keyboarding. A decimal number from 1 to 99 inserted before the data of a page item (copy block) serves as the edit instruction. Copy blocks which demonstrate fixed formatting characteristics such as quad left data, quad right data, main text justified copy, headings, etc., are labeled with the assigned item identifier instruction number. Table 1 shows a list of page copy blocks that have been assigned to an item identifier edit instruction number. ORGANIZATION OF FORMAT PARAMETERS The typographical page parameters fall into three categories: (1) type face of copy block, (2) point size of characters, and (3) character grid number where the type face is located. The dimensional parameters of the page format have been subdivided into the horizontal and vertical parameters as follows: Horizontal 1. The Xon group: These are the absolute horizontal vectors that define the horizontal location of various copy blocks on the page with the left side of the page as reference, such as absolute margin of column, indentations, etc. 2. The LLn group: These are relative dimensions that define the line length of copy blocks useful in line justification operation. 3. The Smin and Smax group: These parameters define the maximum and minimum range of values that Table 1. Assignment of Copy Blocks to Item Identifiers for Format Item Identifier Copy Blocks Numbers Chapter number (single column) 01 Chapter title (single column) 02 Section number and title (single column) 03 Contents (title of table of contents) 04 Chapter number (double column) 05 Chapter title (double column) 06 Section number and title (double column) 07 Side head (quad left) 09 Side head (quad right) 10 &~~rt 11 Text (with shorter line length than body text and having same right-hand margin) 12 Text (with shorter line length than body text and centered to text body) 13 Text of lower point size (with same line length as body text) 14 Text of lower point size (with shorter line length than body text and centered to text body) 15 Notes (with same line length as body text) 16 Notes (with shorter line length than body text and centered to text body) 17 Notes (with shorter line length than body text and having same right-hand margin) 18 Table of contents (single column leadered table) 19 Footnote 20 Line plot (single column) 21 Line plot (double column) 22 Underscore 23 Beginning of caption 24 End of caption 25 For Tabular Formats Fixed fields and/or unjustified line copy (line that does not overset) Justified line copy (line that may overset) Constant page heading Running heading Column heading or subheading Notes Running heading (index) the word space can receive in the line justification procedure. 4. The ~x group: These are incremental fixed values selected for a particular format to position on page copy blocks with respect to previously set data. Vertical 1. The Yon group: These are absolute dimensional parameters that reference and position items from the top edge of the page. 2. The Y LA group: These are the incremental parameters that define the line advance of text copy.

7 A COMPUTER FOR HIGH-SPEED PAGE COMPOSITION The Y LP group: These are parameter values that define the various incremental separations between the end and the beginning of a copy block or vice versa, such as end-of -text paragraph to the following heading. 4. The Yo or ~Y group: These values define the limits and tolerance of the vertical data settings in column justification operations. For tabular formats where the fields of data and digits per field of data are fixed, parameters besides dimensional ones have been derived-such as, number of fields, number of digits per field, number of lines, number of columns, number of groups of lines. HYPHENATION In setting a justified line of type of small line measure, hyphenation is unavoidable; hence, the system had to be equipped with a hyphenator algorithm that could hyphenate the over set word and thus help the program justify the line. The hyphenator algorithm that has been developed by Mergenthaler is based on the eight-window principle developed by Lockheed Co. It consists of a logic mechanism that recognizes and samples the structure of the word as the word flows through a field of 11 windows (a 6 bit X 11 position shift register). Each character of the word under investigation is shifted from one window to another at the rate of the applied clock. The contents of each window are decoded and the information is supplied in a parallel form into a cluster of hyphenation logic rules. The logic of each hyphenation rule consists of a static gating. When the contents of the windows have the proper character combination, which can satisfy either one of the rules, then a hyphen can be inserted by the control section of the Hyphenator between the characters that surround the checkpoint. The Hyphenator has demonstrated high accuracy -over 97 % -with acceptable hyphen efficiency on representative vocabularies. It has been built in a modular form with a data transfer rate of 0 to 100,000 characters per second, suitable for interfacing high-speed computers. MAJOR CHARACTERISTICS OF THE PFC The basic characteristics of the page format processor are as follows: 1. The capability to compose a large variety of tabular and text page formats by receiving edited source data on magnetic tape. 2. The ability to format and process pages by means of a wired-in program at a considerably higher throughput rate than the conventional stored program computers of medium size and cost. 3. The ability to change the program of the machine conveniently by merely changing either one or two printed circuit matrix cards depending upon the format program complexity. INTERNAL ORGANIZATION OF THE PFC SYSTEM General The Page Formatting Computer that can prepare formatted tape to drive the Linotron Photocomposer is basically a special-purpose computer with a wired-in rather than stored program. It can be viewed, however, as a general-purpose page formatting machine in that the program of the Processor can be changed readily with different page programs. This is possible by plugging into the computer the desired format program control boards which actually are matrix logic cards. The simplicity and flexibility with which the format program of the machine can be changed is attributable to the internal organization of the system. The control of the CPU of the system consists of a library of independent typesetting routines whose program has been specifically designed to handle microtypeset operations. The proper selection and sequence of these microtypeset routines by the format program control can compose a variety of tabular and text formats. The formatting program is accomplished at relatively high speeds due to the parallel operation of the arithmetic computations and the simultaneous execution of the routines program. The machine is equipped with three core memory units: 1. Input Memory (buffer) 8K X bits 2. Output Memory (page memory) 16K X 6 bits 3. Working Memory (scratchpad and parameter storage) 2048 X 17 bits

8 144 PROCEEDINGS-FALL JOINT COMPUTER CONFERENCE, 1966 The number system with which the system performs arithmetic operations is in BCD. Five BCD digits parallel arithmetic is performed in two arithmetic sections that operate simultaneously. The internal coding system of the computer is based on 6-level BCD code. The system, however, can accept input data recorded on either 6-level or 8-level code, in BCD ASCI or extended BCD coding system. The computer is equipped with a cluster of address tracking, index and utility registers. The machine control section of the processor responds to nearly 160 command instructions which can be addressed by the format program control independently. The basic machine cycle (MC) of the system is 5 microseconds and instruction can be executed from about 0.2 MC to 10 MC depending upon the nature of the routine called for by the instructions. Figure 4 presents the system's internal organization. It is. subdivided into three units: (1) processor storage unit, (2) central processor unit, and (3) input/ output unit. Processor Storage Unit This unit includes the input memory (1M) and the output memory (OM) of the system. The unformatted data is stored in the 1M. The CPU operates on the data stored in the 1M and forms lines. When the CPU has formed a line of type in the 1M, it transfers it to the OM. The OM is the page memory of the system. In this memory, the data of a complete page is formatted and assembled. In certain formats-such as double column formats-the last item on the page, as recognized by the CPU in the 1M, may cause repositioning of all page items already stored semi-formatted in the OM. The OM, therefore, stores the data of the page until the last page item is read from the 1M which mayor may not overset the page. Central Processing Unit The various functional blocks of the central processing unit have been designed and developed with the modular concept in mind. These blocks have in- CENTRAL PROCESSING UNIT I/O CONTROL BUS LIBRARY OF ROUTINES. CONTROL OF ROUTINES PROGRAM Figure 4. Page' formatting computer system diagram.

9 A COMPUTER FOR HIGH-SPEED PAGE COMPOSITION 145 dividual controls of their own, thus making them independent, self-sufficient and asynchronous. Under the control of the routines, the blocks can be timeshared to perform the routine program in a synchronous manner with the rate that the routine itself dictates. The CPU is divided into the following sections: 1. The working memory section 2. The arithmetic section 3. The address, index and utility registers 4. The input instruction and data decoding section 5. The character width section 6. The library of control program of typesetting routines 7. The format program control 8. The hyphenator The Working Memory Section The working core memory of the system is basically the scratchpad storage of the CPU and the format parameter storage. During the page formatting program, the results of temporary computations are stored in this memory. Words up to 17 bits long can be retrieved or stored in this memory practically from all the registers as well as from the accumulators of the arithmetic section in either parallel or sequential mode. This is made possible by means of the. input output address and data channel multiplexer that can communicate with the rest of the blocks. The typographical and dimensional parameters of the format are also stored by the parameter library in a predetermined area of this memory. The Arithmetic Section The arithmetic section of the system has been divided into two units-the horizontal arithmetic unit equipped with two accumulators that can perform parallel addition and subtraction of 5 BCD digits in 1.5 microsecs.; and the vertical accumulator unit equipped with three parallel accumulators of the same speed. The separation of this section into two arithmetic units and the reinforcement of each unit with several accumulators insures fast and parallel computations of the horizontal and vertical arithmetic equations involved in setting the data on the x-y plane of the page. Comparison decision modules and auxiliary registers-namely, Sn, en, Ln and Y n registers-have been incorporated in this section to speed up, facilitate and simplify the routines program. All accumulators can store their results in either the scratchpad or input and output memory through the I/O channel multiplexing. Address, Index, Utility Registers Addressing, tracking and indexing of data in the scratchpad input and output memory is implemented by means. of the respective registers. The A-B registers are primarily utilized by routines that form the line in the input memory. The 10 register is utilized by routines to help sort out graphic data, footnotes or any type of headings from the input data and subsequently store this data temporarily in a separate area of the input memory. Registers P and M are primarily utilized by routines to index in the scratchpad memory the addresses of lines and various items that lie semi-formatted in the output area as well as the temporary storage area of the input and scratchpad memories. Registers C and D are associated with indexing data address of items already stored in the OM. Input Instruction and Data Decoding This is the section that decodes the edit instructions and data characters when the machine is in the read input data mode. It sorts out the input edit instructions (Item Identifiers or Typographical Control Codes) from the printed data codes and keeps up to date the format program control by means of control signals on the state of instruction and data the program is working on. Certain routines receive control lines from this section. This block provides command control signals to the character width block and P z, F, G storage registers when input character codes are to receive a width value. It also provides command control signals to the address matrix of the working memory to switch the memory to the appropriate core area where the parameters of the particular copy block are stored. Character Width Section The function of this block is to receive a 6-bit character code from the input/ output channel multiplexer of the input memory and yield as an output the typographical width of this character in parallel BCD form. The character width that is being outputted by the

10 146 PROCEEDINGS-FALL JOINT COMPUTER CONFERENCE, 1966 block is a function of the three typographical parameters which are received from the parameter section of the working memory or directly from the input; namely, the typeface, the point size of the face, and the Linotron grid number. The width information is accessed at its output in less than 300 nanoseconds thus speeding up a typographical command so repetitive in printed matter. The character width information is routed via the program of the specific routine into the Horizontal Arithmetic Unit. This block is rather flexible since it holds up to 16 different type faces which can be substituted externally by a different one when the job calls for it. The Hyphenator The duty pf this section is to receive the word which oversets the line and output the same word with all the grammatically legitimate hyphens that it can find. The overset word is transferred from the 1M into the Hyphenator by means of the I/O channel multiplexer and from the Hyphenator into the scratch pad area of the working memory. When the program finds that a word is split within the justification zone, then the line is set in the (OM) with the last word hyphenated. When the routines are instructed to perform, they control the operation of each individual module within the CPU and thus determine and synchronize the direction and flow of data within the machine. They output clocked signals to the computer blocks that they control by having access to all clocks that the machine cycle and clock distributor provide. Interfacing control signals are received by this section from various blocks of the CPU for certain routines whose program calls for it. Upon the completion of their program, the routines turn themselves off and the majority report the end of the routines program to the Format Program control which initially activated them. Figure 5 shows the flow chart of a typical program for such a routine. The Format Program Control The format program control is the controller of the central processor unit in that it controls the Control of Typesetting Routines Program (Routine Library) Like any other control section of the computer, this is the instruction execution section of the CPU. It consists of 160 independent routines subdivided into the following four distinct sets: 1. Routines associated with the horizontal computation and set up of text body copy blocks in the page memory. 2. Routines associated with the setting and temporary storing in the input memory copy blocks that may cause a column or a page to overset, such as headings, footnotes, nonmandatory graphics and captions as well as updating and storing page-derived running heading copy blocks. 3. Routines that are associated with the vertical setting of copy blocks within the columns of the page. 4. Routines associated with the setting of table data, generating and setting page number and setting initial parameters within the computer blocks. Figure 5. Flow chart of routine scanning graphics data in 1M.

11 A COMPUTER FOR HIGH-SPEED PAGE COMPOSITION 147 various computations and operations which produce a properly composed typographical page. Upon the provision of edit instructions and data by the decoding section, the format program control selects and sequences the appropriate typesetting routines. As a program controller, it supervises the state of the routines and the state of the page being formatted. It follows prescribed page composition rules on graphics insertion, footnote formatting, running heading updating, etc. Following is a list of some of the page make-up rules which it obeys and executes: 1. A double-column nonmandatory graphic* referred to on a page will be set by the program at the top of the next page. 2. A single-column nonmandatory graphic referred to on a page will be set by the program either on the same page or on succeeding pages. The program will never place a graphic in a column preceding the graphic reference in the text. 3. A single-column nonmandatory graphic that is referred to in the first column and the program discovers that this graphic cannot fit in that column will be placed at the top of the second column. The program will continue processing the text in the first column until the valid column depth is reached. 4. When a single-column nonmandatory graphic appears in the second column of the page and the program discovers that it cannot fit it in that column, it will transfer this graphic to the next page on the top of the first column. The program will continue to process the data in the second column until the valid column depth is reached. 5. Any single-column mandatory graphic t will be placed by the program right after the reference point in the text. If the graphic does not fit in the first column, the length of the first column will be left short. The program will transfer this graphic at the top of the second column and the depth of the second column will be made equal to the depth of the first. 6. Any single-column mandatory graphic that is called for in the second column of the text and the program cannot fit into this column will be transferred to the top of the first column of the next page. The program will readjust the depth of the second column to equal the depth of the first column. * Nonmandatory graphic is the graphic which is not necessarily set on the same printed page. t Mandatory graphic is a graphic that is placed right after the point of text reference. 7. Any double-column mandatory graphic that appears in the first column for the first time will cause the program to split the already processed column into two parts. If the split occurs between a heading, the heading will be part of the second column and the second column will be longer than the first. If the split occurs between text, the program will justify the second column to the depth of the first column. The program will continue processing the data following this graphic in column one. Exceptions to the rules are as follows: (1) A minimum depth will be required of the first column in order to be able to justify the second column to the depth of the first. (2) If a split occurs between a heading and the heading is the first line of the column, then the column will not be split. The second column will contain no text and the graphic will be placed after the text in the first column. 8. If a double-column mandatory graphic is called for in the second column, the material preceding it will be arranged by the program into two justified columns above the graphic. If the graphic does not fit, the program will end the page and the graphic will be placed on the top of the next page. If in the process of splitting the column, the program senses that the column is split immediately after a heading, the heading will be placed in the second column and the first part of the column will be justified to the second part of the column. 9. If a graphic-text format contains double-column mandatory graphics and a single-column nonmandatory graphic, then the single-column nonmandatory graphic will be treated in the program as a doublecolumn nonmandatory graphic. Rule 1 will then apply. 10. If a footnote reference appears in the text of the first column being processed, the footnote information will be placed by the program at the bottom of the first column. If the depth of the footnote data is such that the program cannot fit it into the first column, then the program will place this footnote at the bottom of the second column. 11. If a footnote reference appears in the text of the second column being processed, the program will place this footnote at the bottom of the second column. If the depth of this footnote is such that the program cannot fit it in the second column, it will place it in the following page. The footnote reference (superscript) will be an asterisk instead of a numeric in such cases.

12 148 PROCEEDINGS-FALL JOINT COMPUTER CONFERENCE, If a single-line heading appears at the end of the column being processed and the program cannot fit this heading with one line of the text following, then the program will transfer the heading to the top of the second column. 13. If a new-page-derived running heading is detected, the program will replace the wording of the old-page running heading which is temporarily stored with the new one. 14. Any double column heading that appears in the first column will cause the program to split the already processed column into two parts and the program will proceed as in Rule 7. The Format program control section can house up to six different format programs which are plain matrix boards that can be plugged into the machine. The Input/Output Unit The Input/Output Unit is equipped with three tape stations and a page display monitor-the Formatscope. The first and second tape stations can accept tape drives reading either IBM 729 tape formats or IBM 360 tape formats depending upon what tape drive is chosen by the input requirements. Both tape stations can accept tape drives with speeds up to ips reading tapes with packing density up to 800 bpi. Thus the system accepts data recorded in either 6 or 8-level code. The code converter block reduces the 8-level code to 6-level BCD code, the internal coding of the system. Either one of the two input stations can be utilized for input data loading or parameter loading. The third tape drive is utilized as output when the formatted page stored in the output memory is to be loaded on a magnetic tape. The F ormatscope is a video display output device which, as an x-y plotter, upon command can display on a video screen the data of a complete formatted page. Each character is shown as a dot. It serves as an on-line diagnostic device for the operation of the machine proper.