Optimized Orthography Facility
The Optimized Orthography Facility or OOF is a hardware system that generated and processed Shinasthana characters, later emulated in software. The encoding it used to address character sets became an industry standard for processing Shinasthana characters. Due to Shinasthana's non-alphabetic nature, early computers were unable to store or display Shinasthana characters. OOF was developed by Themiclesian computer developer and manufacturer DPS first as an add-on to its existing System/1200 mainframes and then as its own line of products. It was conceptualized as early as 1947, but the first working prototype was not produced until 1963, appearing on the market in 1965. OOF generated Shinasthana characters via a look-up table which stored characters as raster information in an 8-by-8 matrix.
Demand
While DPC (then called Tunning Machinery Company until 1950) had begun its computer development with input from UBMC of the Organized States in 1947, all computer input and output were in assembly language or, if in prose, in the Sylvan alphabet. Literacy in Tyrannian of prospective Themiclesian clients expanded in the previous decades, but DPC felt that allowing computers at least to store data in Shinasthana characters would expand that clientele immensely. DPC management deduced this from the still-predominant use of Shinasthana in accounting and communication, and two key functions of early computers was on-line inventory control and accounting.
Development
DPC began development of OOF in 1953, though this proved to be an intermittent and challenging process. One out of many was the scarcity of memory; since Shinasthana was written with pictosemantic glyphs rather than a phonetic alphabet, storing the shapes of any adequate number of glyphs in raster information required obscene amounts of memory that was completely impractical during the 50s. A minimal set of 5,000 characters would consume 16,000 words of main memory, in an era when a powerful computer typically had around 8,000 words; even with the luxury of doubling the memory capacity, which involved installing additional addressing circuitry, it left no memory for program code and data. Between 1954 and 1964, DPC engineers tried to manouevre around this problem by analyzing Shinasthana characters into smaller components and then piecing them together as required; however, this did not result in a decisive reduction in storage consumption. Furthermore, this procedure reduced picture quality, which under the already-minimalist 8-by-8 matrix was considered at the very edge of legibility. Other schemes, such as marking nouns with dots and verbs with crosses, achieved even less success.
Memory prices began to drop after the late 50s as automated machinery began to magnetic core replace manual work in stages of manufacture. Magnetic core memory, as a solid state, non-volatile medium, had the advantages of being faster than Drum memory and more stable than Williams tube memory. Improvements in manufacturing process, particularly with the first memory manufacturing plant set up in Themiclesia in 1959, gradually lowered the "memory barrier" that OOF developers faced for years. By 1960, each byte costed only $5 (OSD, nominal), down from approximately $92 per byte per month (OSD, nominal 1953), since Williams tube memory was prone to burn-out and on average only lasted one month. In 1963, this further dropped to $3 per byte, which meant the entire bank of 5,000 characters, each defined by a 64-bit raster, totalling 40,000 bytes, could be stored with $120,000 worth of memory.
Since work had begun on the to-be-released System/1200 series of mainframe computers without the OOF team's input, they quickly decided to develop an add-on component of the System/1200 line instead of a standalone system that they previously envisioned. In retrospect, they realized this was a wise choice, since a standalone OOF would cost considerably more than the System/1200 but enjoy none of its peripheral support.
Hardware
Other than the rasterization of the set of Shinasthana characters, the centrepiece of OOF was the addressing system that included a 64-bit data path from the character storage and permitted single-cycle retrievals of all 64 bits of data that defined any one character. This was crucial for performance under the System/1200 architecture, because the OOF's cycle was half the speed of the System/1200 but double the data; this meant that with three cycles on the System/1200 CPU, a complete character could be requested and received by the CPu and sent off for display on a terminal. At a slower rate or less optimal synchronization, OOF's performance as perceived by each terminal-user would fare much worse.
After each fetch command, the data from OOF was transmitted in parallel into two special registers accessed by the System/1200 CPU, called OOF0 and OOF1; after an Shinasthana character input was received on the terminal, the CPU then transmitted data from these two registers to the relevant channel controller, and it to the terminal on which the character was entered. The terminal controller translated the OOF data into the terminal's display protocol, since the terminals themselves had no logic to decode OOF data. To enable the controller to decode this information, an add-on was also required; this was the Mini-OOF (DPC 1273).
Software
To use OOF, the mainframe ran a program called the Shinasthana Orthography Digital Encoding Software (pronounced "so-deck"). The role of this program is analogous to the input methods of modern computers. While each terminal in the System/1200 line came with a keyboard, it was clearly impractical to include a key for every Shinasthana character; therefore, a way to identify the correct character so that OOF could retrieve it had to be devised. This scheme came to be known as the Bi-phasal Input Gamut, or, more concisely BIG-OOF.
BIG OOF
DPC engineers took inspiration from the ancient practice of splitting the Shinasthana character phonetically into two halves, the first defining the initial consonant, and the second half the remainder. However, since there were more variations in the second half than the keyboard could accommodate, the first segment of the second half, representing the syllable onglide, was encoded with a separate key press. Initial consonants appeared on the left of the keyboard, and prefix consonants were accessed with modifier keys; the onglide appeared on the right of the keyboard. The final element, which encoded the syllable nucleus, coda, and tone, was spread across the entire keyboard; hence, the first two strokes could be entered in either order, but the third stroke could only be entered after the first two. The rule of final element was that an unshifted character accessed non-nasal finals, and shifted presses accessed nasal finals according to their traditional complementary arrangement (-d with -n, -g/0 with -ng, and -b with -m). The level tone was on the home row, the rising tone on the Q row, the falling on the Z row, and unvoiced stops (which had no tone distinction) on the number row.
BIG OOF, based on the traditional phonological analysis of Shinasthana character, proved quite popular with data entry clerks and analysts alike; it had the advantage of being phonetically intuitive and relatively economical with the number of strokes required; all possible phonetic variations in a given syllable could be reproduced within three main strokes and two modifier storkes. On the other hand, since it encoded for phonology, and Shinasthana has no shortage of homophones, each typed character may need to be "switched" with a homophone with one or more key presses until the desired one appeared. This was a major shortcoming, as homophones are not displayed before they appeared on the screen, and in some cases clerks may need to switch up to twenty characters to find the one desired; however, the most common ones were pre-programmed to appear early in the switching process. Another issue is that ignorance of phonological composition of a character prevents the any input.
Interpretation
As the OOF team did not work on the System/1200 terminal development, it did not came with logic to interpret key presses based on BIG OOF correspondence; therefore, the Mini-OOF was responsible for intercepting and interpreting BIG OOF encoding. The input string was sent as a 18-bit quarter-word in parallel, with the first five bits corresponding to the intial, the second five bits the onglide, and the last eight bits encoded for the final. Mini-OOF proceeds to translate this quarter-word into a 12-bit address that OOF uses to retrieve its character raster data.
Conversely, when OOF replies with character information (which is 64 bits long), it is mapped such that the first bit corresponds to the dot on the upper left, scanning downwards and then processing to the next column on the right. This data is then sent to the terminal in four pulses of 16 bits each, which allows for sufficient time for the CRT scanner to reset to original position (since the CRT scanned the 320 vertical lines in four interlaced phases of 80 lines each) and other technical issues. In order to allow the characters to persist on the screen, the terminals had POI circuitry enabled, which allowed the terminal's internal logic to reach optical information on the screen (temporarily retained by the phosphor) and then reproduce them line-by-line, thus eliminating the need for screen memory (which would have been prohibitively costly) and constant character requests from OOF.
Sales
OOF as an add-on eventually came to support 87% of all System/1200 installations, and most of the exceptions were reported from abroad. The success of OOF and System/1200 went hand-in-hand, as described by many media analysts at the time. While OS manufacturer still held a technological edge over DPC in terms of pure computational power, the DPC has, by OOF, opened a completely pristine market in which it faced no competition for the next two decades. OOF would, by 1970, become a standard piece of computing equipment in almost every business's data analysis or statistics department. It also found a warm reception in the publishing and typesetting industries, since before then most Shinasthana publications were edited and typeset entirely by hand, a laborious and unhealthy process, since the types were cast from lead, and ink contained carcenogenic compounds.
Pricing
OOF itself did not result in a large number of invention patents but a considerable number of improvement patents. That improvement patents lasted for only twelve years in Themiclesia meant that DPC had to sell as many OOFs as possible, before the technology became publicly available. The DPC's strategy, in combination with System/1200 cales, required the OOF to be sold at a profit margin not exceeding 14.5%, and sometimes even less if a large order of System/1200 was simultaneously placed. Alone, OOF costed about $421,000, and each Mini-OOF, $96,000.
Foreign distribution
During development, DPC had distribution channels in the OS, Menghe (Republic), Dayashina, and several Casaterran states. Of these, Menghe and Dayashina appeared as the markets most likely to purchase OOF on a widespread basis, since both had orthographical traditions based on Mengja, which OOF generates. Yet by the time it hit the market, the Republic of Menghe had fallen to the Democratic People's Republic, which was strongly FSR-leaning; the government announcing its policy not to recognize the DPRM only weeks before the product launched, the DPC found it impossible to market this product to the unrecognized government and focused instead on the Dayashinese market. It did not, however, prohibit redistributors from selling the system to Menghean procurers, though such instances are understood to be very infrequent. With Allied endorsement, OOF saw considerable sales in Dayashina; estimates place total installations in Dayashina at around 1,100 by the time OOF was withdrawn from marketing. In Dayashina, OOF was called "National Text Generator" (国文生成器).
Derivative products
1500 Shinasthana Typing System
OOF found itself re-implemented as a single-user editing and typesetting system (the DPC 1500). This was meant to be a complete hardware and software suite for small-scale publishing houses or one-man operations. DPC believed that the availability of such a system at a modest price would trigger a boom in the publishing industry, since it knew that a large number of universities could not publish as many journals as they wished due to the oligopolistic market dominated large publishers. However, almost as a complete surprise, the UBMC of OS released the Typemetric Typesetter in 1971, only a few months after the 1500's debut, and UBMC's system cost only $26,500, whereas the 1500 was around $172,000. Granted, the UBMC's machine was mechanical and could not produce Shinasthana characters, but the DPC failed to appreciate that most journals could also be published in Tyrannian, though Shinasthana was ostensibly preferable, except at such a large price disparity.
1530 Random Access Shinasthana Storage
This was a hard disc drive system (patented by UBMC in 1952, expired 1969) specifically optimized for storing OOF data. The grounds for its development was that DPC's terminals produced picture in such a way that created a lag that required the terminal to store incoming data before it could be demodulated and scanned. The 1530 RASS, announced in 1971, stored OOF data in a staggered format so that each 16-bit pulse of data arrived in sync with the terminal's refreshing rate, and four pulses representing four different characters were delivered across the 64-bit width of the System/1200 channel path. In this manner, the monitor produced a quarter of each of four consecutive characters at once, without losing data or reducing the amount of information drawn per refresh cycle. The 1530 was superseded by the 1532 in 1977, whose maximal configuration increased storage capacity from the 1531's 23 million to 150 million Shinasthana characters. The 1531 sold for $432,000, and the 1532 for $489,000.
1571 Shinasthana Word Processor
The 1571 was released in 1978 at the very beginning of the age of microcomputers as a standalone system that generated, displayed, and manipulated Shinasthana text. This system was used by small businesses to process accounts and prepare documents and correspondence. It included a logic unit the size of a minibar, a monochrome monitor, a dot-matrix printer, and a keyboard legendary for its touch and response. However, this keyboard was also the subject of many a noise complain from neighbours of its users, since the keyboard was roughly as noisy as an electric typewriter of the age. The 1571 saw the adoption of flip-flop memory over traditional magnetic core memory; flip-flop memory reduced access times by around 90% and were much more compact than core implementations of the character set; however, the logic needed to drive the memory was still considerable.
1572 Shinasthana Word Processor Facility
The 1572 was a hardware add-on, via parallel port, to the increasingly popular UBMC Personal Computer line. While PCs were not initially as popular in Themiclesia as in the OS, flourishing commercial activity between the two states necessitated many OS businesses with existing UBMC equipment to process Shinasthana text. The 1572 was developed in conjunction with UBMC to address this need, and UBMC offered this product as an optional expansion to its PC line. Ironically, due to the popularity of the PC, the 1572 easily outsold the 1571 as early as 1983. The 1572 is the final piece of hardware dedicated to OOF functionality that DPC released. Since the advent of Static RAM, the proliferation of floppy disc storage, and the increasing accessibility of ROM chip based character sets, dedicated hardware became increasingly uncompetitive; by 1981, the entire character set and software could be put on a single floppy disc with room to spare, and given sufficient read/write speed and processor clock rate, there was no difference in performance compared to a bulky and expensive hardware system.
Assessment
As one of many conceptually original products that DPC developed and marketed in the 60s, the OOF and its derivatives was considered a commercial and social success, establishing a widely accepted standard for encoding Shinasthana text. Though the specific communication protocol changed in response to hardware advances, and this matter eventually disappeared as microcomputers gained the computational capacity to generate text completely in software, the open-ended encoding scheme that OOF's developers (intentionally or not) used allowed much larger character sets to be included while retaining compatibility with older systems with minimal issues.