Canadian Book Publishing Development Program Cultural Industries Division Arts and Culture Branch Department of Communications 365 Laurier Street West Ottawa, Ontario
The Coach House Press, 401 (rear) Huron St., Toronto M5S 2G5 The Porcupine's Quill, 68 Main Street, Erin, Ont. N0B 1T0 Penumbra Press, Moonbeam, Ontario
Prepared by David Slocombe, Consultant
April 30, 1983 
This report presents a proposal for a computer facility for small Canadian publishers, to assist in making them competitive with larger firms and foreign publishers in the Canadian market.
A Pilot Project was carried out to determine an optimal configuration of equipment and software and to explore office and production systems which use this configuration. We believe that the Pilot Project was highly successful and therefore wish to seek funding for making a permanent installation for all three of the participant firms.
Everyone knows that the computer industry is a growth industry, that sales of computer equipment of all types and sizes are expanding ever more rapidly, so that pundits talk of the computer "revolution", the dawn of the "information age", and so on. What no one seems to notice is the effect of all this on the firm trying to keep up with the opportunities afforded by the new technology.
The truth is that no one, not even consultants who spend full time on it, can have a clear idea of the best way to take advantage of computer technology -- because every month the situation is different. For example, in the course of negotiating with salesmen for the best quotation for a computer for my three clients, I have had to resort to bargaining for a  promise that, no matter what we have on the quotation today for submission with this report, the equipment actually purchased will be the best the vendor has to offer for the money in the month we actually take delivery. And we know that almost certainly what will be available then will be very much better than what we can specify today.
A case in point is the "disk drives" that are used for mass storage of data and programs: what is sometimes called "secondary storage". The quotations attached to this report say "two 20-megabyte winchester disk drives: $5,500", and it is true that these are minimally adequate for our purposes. But that negotiated price of $5,500 was chosen to cover the expected price of the new, much faster and larger (and cheaper!) disk drives which the salesman expects to be selling by July or August. That $5,500 is too small a price for the current crop of drives available today, so the salesman is banking on having newer technology to give us by late summer at the latest or his profit will be severely reduced.
Thus it can be seen that decisions about computer purchases have to be made in the context of not a single moment in time but a whole range of moments. One is forced to invest in futures with every decision to purchase or one is almost certainly wasting money. This is a very serious complication.
By the same token, it is important to assume from the start that the computer equipment one buys today will be replaced by much more effective (yet at the same time cheaper) equipment  within a few years. The option of running today's equipment into the ground before replacing it is simply not open, because long before that equipment wears out two things will happen:
It behoves the purchaser of computers, therefore, to worry a great deal today about what happens when he or she goes out to buy the next computer after this one. Will it be possible to get the software which will have been developed for today's computer transferred easily to the next computer? Is it reasonable to expect that this software will be easily adaptable to the more sophisticated capabilities of the newer computer? Will the effort expended to learn to use today's computer system have to be repeated when the next "generation" of machine comes along?
Such concerns have been the bread-and-butter concerns of Vice-Presidents for Information Services and the like at our larger corporations for decades now. But as computers get much cheaper and become accessible to small businessmen it is likely that these worries are being glossed over in the rush to "get on with it". Yet, the expenditures which smaller businesses are making on computers are often similar to those which the bigger firms make in proportion to the the scale of the business. Indeed, the risk which the small business takes in making these decisions is actually much greater, because there is less "fat"  to sustain them if things don't go according to plan and profits suffer.
In preparing recommendations for computer purchases for The Coach House Press, The Porcupine's Quill, and Penumbra Press, I have chosen to adopt a form of conservatism. I have tried to ensure that the "hassle" caused by acquiring computer equipment -- the sizable number of manhours which the principals and their staffs will devote to making the new equipment an effective co-worker in their enterprises -- will be well-compensated by the benefits flowing directly from that equipment. I have kept in mind that my clients are in the business of publishing Canadian literature and that if computerization, however "marvelous", however "cost-effective", distracts them from their task of producing books and journals and placing them in the hands of Canadians, then the fancy technology will not be living up to its promise.
When we began to formulate plans for technological upgrading of the three presses, our goals were necessarily vague. It would have been useless to have very concrete dreams when we had at the time no clear idea what was feasible for firms so small. We knew only that the manuscript preparation and editing tasks were time-consuming and costly and error-prone, and that turning those manuscripts into type was a clumsy as well as expensive process. We felt -- as every publisher must feel -- that sales could be  assisted by a better use of the information derived from previous sales, and that each firm needed much tighter control over its operations without, naturally, wanting to expend much human labour usually reserved for editorial work on realizing that control.
Over the years, CHP has learned much about applying computer technology to these problems, but the results have generally been only partly "practical" because the equipment we have had to work with was so primitive. It was a natural approach, therefore, to consider what we might do in terms of what CHP had done imperfectly, to consider what was missing. This led us to an initial "configuration" for our computer systems, which in turn suggested further good things we might hope to achieve if each press possessed such a configuration.
A series of re-evaluations took place as we attempted to optimize this configuration so that the cost/benefit ratio (actually, the "hassle"/benefit ratio) would be as low as possible. Then CHP engaged in a nine-month Pilot Project to gain experience with a configuration close to what we had envisioned. Again, re-evaluations took place as experience accumulated and, towards the end, as we began serious discussions with various suppliers of computer equipment and software and found out just what we could hope to obtain for a cost that did not seem outrageous. 
Without question the most crucial decision made for this proposal was to adopt the UNIX Operating System as the standard operating system for whatever computer is purchased. On this decision depend virtually all the other choices in this report.
An operating system is the computer program -- actually a whole suite of computer programs -- which controls the overall operation of a computer, which manages the various resources the computer uses to do its tasks, and talks to the computer user ("interfaces with the user") about what to do. Usually it also provides the services most commonly desired of a computer by means of ready-made programs so that the user is not forced to make (or buy) special programs to do these things.
Without an operating system a computer is a useless, dead thing that simply lies there inert and flabby. It is the operating system that is the life breathed into the golem to make it useful. Once the operating system is "running" in a computer, one can tell the computer to perform useful tasks; that is, one can tell the operating system to "run" programs which do useful things. It is the job of the operating system to find the text of the program, place it in the computer's working memory, set it running when the resources it requires are available (and set other programs running instead whenever the resources are not available), handle any exceptional situations which arise while the program is running, prevent one program from affecting the operation of other programs which may be around at the same time,  and finally to ask the user for something new to do when current tasks are completed.
The first operating system which made any impact upon the world was the one designed in the early '60s to control the then-new IBM computer called the System 360. It is reported to have taken five thousand man-years of skilled work to create, by programmers and engineers scattered all over the world. Although virtually all large IBM computers as well as the computers of some other American manufacturers (and all large Soviet computers!) still run this operating system or a direct descendant of it, there was and is a great deal wrong with its design. Hindsight is a wonderful thing in engineering!
In particular, "OS" (as it is called) is far from "user-friendly". At the time it was designed, computers were monstrously expensive and people were cheap. Computers were thought of as (and were) giants dispensing certain otherwise-unattainable gifts, giants tended by servants who had been initiated into the mysteries required to keep the gifts flowing smoothly and quickly. "User-friendliness" was far from anyone's mind then. Given the computing climate, it would not have seemed too unreasonable if large "IBM machines" had consumed the occasional programmer or operator along with their daily diet of punched cards.
Times change. The capabilities of the million-dollar air-conditioned monsters of the '60s can be had for a few thousands now, packaged up in human-sized, attractive boxes designed to fit  the decor of office or home. And the roles of machine and "user" (no longer "operator") have changed. The computer, like the car, is expected to be reliable, polite, approachable, and -- above all -- usable for small, mundane tasks (instead of only grandiose ones) with a minimum of trouble for its owner.
These small but powerful machines have been on the market for only a few years. Until recently they had only the most primitive of operating systems to control them, systems reminiscent of the "control programs" which preceded IBM's OS in the '50s and not much more "user-friendly". Now, an even more powerful new generation of small machines is just reaching the market -- the "16-bit microcomputers" -- and it would be sinful not to supply these with sophisticated operating systems.
Although there are a number of such operating systems available, there is one -- called UNIX -- which will, many industry-watchers believe, come to dominate the market as a de facto standard. It was developed originally by computer scientists at Bell Laboratories (which is the owner of the UNIX trademark) and enhanced by hundreds of faculty members and graduate students in computer-science departments at universities all over North America. Recently, Bell Labs have made the licence for UNIX available to computer manufacturers, firms specializing in computer software, and "end users" (you and me) for a price that is so low that it would be insane for anyone to try to create an operating system of this scale and sophistication from scratch. Instead, it is easier and cheaper  to get a UNIX licence and enhance that operating system. A great many firms in the computer industry are doing just that.
Besides cost, there is another reason why UNIX is becoming so common. It is (relatively) "portable": that is, although it was designed on one particular type of computer, it can fairly easily be persuaded to do its magic on any brand of computer of sufficiently recent design to be interesting to anyone. It will not run on "8-bit" computers of the sort which started the personal computer revolution -- they are just not powerful enough and they do not have certain features essential to sophisticated operating systems. But it can be made to run on all but the most primitive of the "16-bit" machines which will be coming into common use in the next year or so.
Already, at least 40 manufacturers of computers have "ported" UNIX to one or more of their products. These include many recently started companies as well as IBM and DEC and NCR and Amdahl.
For Canadian publishers contemplating the purchase of a computer, the broad availability of computers all "running UNIX" is of fundamental importance. In the past, when a firm got its first computer and built its office systems, it had made a commitment to the manufacturer of that particular computer which was virtually eternal. Large firms and government departments which acquired IBM computers running OS years ago, and have developed sophisticated "systems" around them since, cannot decide to switch to a different manufacturer for their next  computer (except for Amdahl, because Amdahl makes computers which look just like IBM computers). The cost of conversion from OS to a different manufacturer's operating system would be enormous, frequently beyond the resources of even very large organizations.
As a result, when large, established computer manufacturers talk about the importance to their business of "customer loyalty" they are really referring to "customer slavery". At least until recently, there has been a kind of universal "structural" monopoly in the computer industry. Once an IBM customer (or DEC or NCR ...) always an IBM customer.
But the new, smaller computers, made by a variety of manufacturers, are likely to share the same operating system, and therefore a firm buying its "next" computer will not be penalized very much if it finds that a different manufacturer currently has the best new machine for the money. That operating system will be UNIX.
Last but not least, the UNIX operating system is the best operating system available. It was designed by computer scientists with wide experience of other systems, and designed for their own personal use as their day-to-day working environment. By what was perhaps an accident, it was designed so that improvements could be added to it in small increments by the users themselves without any need for committee meetings and approvals, because the "improvements" affected only those users who wanted to be affected. A process of natural selection has worked to cause bad ideas to disappear from UNIX while good ideas  gradually spread until they were universally adopted.
There are some detractors who maintain that UNIX is not user-friendly because the "standard" user-interfaces (known as the Bourne shell and the C shell) are designed to be convenient to sophisticated computer users rather than just anybody off the street. However, UNIX user-interfaces are just ordinary programs like any other and can be replaced for specific purposes or a particular class of users by special programs of local design. This has been done by many firms, and some such programs are marketed separately from UNIX as additional software.
The quantity of ready-to-use software that comes with UNIX is enormous. An idea of the quantity involved can perhaps given by the following: it would take roughly 80 of the little "diskettes" which are used to store programs or data on Apple or Radio Shack computers or the usual word-processors to hold the programs which UNIX users expect to be "on-line" and available at all times while they are working. It is not uncommon to price programs like many of these at $1,000 or more each. Yet most manufacturers are making the whole of UNIX available for about this price. No wonder UNIX is being supplied with most of the new small computers! 
There are two common ways to go about making a major investment decision: carrying out a strategic planning study, and setting up a pilot project. Actually these two approaches are seldom used alone: a pilot project can hardly be set up without some planning, and strategic planning nearly always involves using (unasked) the results of other firms' experiences as if these others had done the pilot project for you.
There are advantages and disadvantages to putting either one of the approaches in the ascendancy. Pilot projects are risky in the sense that the outcome is unknown and considerable expense is necessary. On the other hand, strategic planning must concentrate on the most obvious and widespread advantages offered by a new technology when frequently the greatest gains to be made lie hidden in unique peculiarities of the situation. These gains cannot be discovered in advance without experimentation.
N Dean Meyer pointed out in a recent article (Computerworld OA, Feb. 13, 1983) that what is really needed when a firm is contemplating a new technology is a track record -- a history of achievement and a cadre of people experienced in applying it. He continues:
Implementing pilots before a [strategic] plan risks technical chaos. However, the top-down approach risks no progress, a fate far worse than future technology patches. Early pilots can be designed to ease later integration by focusing on established vendors' available technology that is flexible and can communicate in standard ways. 
However the choice is probably most influenced by personal styles. The more analytical (ratiocinative) strategic planning approach attracts the Business School graduates, while the more synthetic (creative) pilot project approach better fits the mental set of the Engineering School graduates, who fill our corporate and government decision-making apparatus.
We did a pilot project. In fact we did two pilot projects. This report refers to them as Phase One and Phase Two.
We were lucky: for Phase One of our experiment we found a "software house" (i.e. a firm specializing in writing and selling computer programs) -- which had a UNIX system installed not far from Coach House Press that was (temporarily) too large for their immediate needs. We rented "time" on this machine for nine months so that we could try out our ideas.
Five DEC VT100 display terminals and one line-printer were attached to the remote computer via "dedicated" telephone lines. These are lines which do not go through a Central Office switchboard -- hence you can't use them to make telephone calls even if you put phones on them. Our communications with the computer was simplified with the help of a truly marvelous product of Gandalf Data Ltd. (of Ottawa) called a "statistical multiplexer".
Three of the terminals went into the typesetting department. Most work done in that department was transferred from the old  Datapoint computers to the new timesharing system. One terminal was provided for Publishing order-fulfillment and promotion, and one terminal was placed on Stan Bevington's desk for accounting, production control and system development. The line-printer was a GE Terminet 340 which has been used for some years with the Datapoints.
One of the typesetting terminals could easily be disconnected so that a Datapoint could be plugged in in its place. This way files prepared for phototypesetting could be shipped to the Datapoint. At present only the Datapoint can actually feed files to the Mergenthaler VIP phototypesetter. But this will be easy to change when the time comes.
At the computer system -- a Zilog S8000 owned by Rhodnius, Inc. -- we had a maximum of 17 megabytes of disk storage for all our files. (Think of a megabyte as storage space for a million characters of text.) In addition, we used the UNIX system utilities, stored in another part of the disk which we shared with Rhodnius. As with all timesharing computers, there was only one central processor. It attempted to keep our programs running fast enough for us while doing the same for Rhodnius' programs. At times, when Rhodnius' staff were making major modifications to their software products, our programs took noticeably longer to run. Presumably Rhodnius programmers blamed Coach House when the system slowed down for them.
Every day, all files on the system which had been changed in any way were "backed up", i.e. copied onto magnetic tape, and  once a week all the software and data files on the system were copied to tape whether or not they had been modified.
This backup procedure is extremely important as a protection against disaster. It is not unknown for disk drives to destroy themselves, and most of Coach House Press' current printing projects as well as much accounting and sales data were located on the disk at Rhodnius. That some of that data was also on paper at CHP was little comfort: it is only really useful if it is in "machine-readable" form so that it can be manipulated by computer.
Initially we had a great deal of trouble with the Bell lines connecting us to Rhodnius, but once the lines were tamed we had no serious technical problems, nor was the computer "down" (unusable) for more than a few hours during the nine months of the project. This freedom from trouble reflects both the much greater reliability of modern computing equipment and also the prompt and expert attention to problems given by Rhodnius personnel, who themselves were dependent on the Zilog's smooth operation.
Over the years, CHP has acquired a large body of machine-readable records of book sales. The first priority for improving publishing efficiency was to transfer all this material from Datapoint diskettes to the new computer system and to build a database out of it. About four years of sales records already  existed at the start of the project. During Phase One another nine months of sales were added.
We used a software tool (a computer program) called the "Mistress Database Management System", developed by Rhodnius, to organize our sales information and to provide a flexible way to extract information from it. The Mistress DBMS began as a Masters' thesis in Computer Science at the University of Toronto under Prof. Dennis Tsichritzis. Then two of the three graduate students involved in the thesis project formed Rhodnius to develop and market Mistress as a commercial product. They have improved and extended it enormously since their grad student days, and the improved version is used now in research into database management systems at the University of Toronto. Hence there is a continuing symbiotic relationship between the commercial venture and the research activities at U. of T.
We were fortunate to be able to tie into this state-of-the-art software development. Because we were using the computer system on which Mistress is being improved, we were able to use the very latest new features, sometimes as much as six months ahead of any other customers.
Mistress belongs to the class of database systems called "relational" because all information is organized into tables, and tables have the form which mathematicians call "relations". Older types of database system were "hierarchical" (such as IBM's  "IMS" system which is used on most very large corporate or government IBM computers), or "network" (such as "System 2000", an alternative to IMS which will also run on some non-IBM computers). But these data-models, which developed in an ad hoc fashion in the pioneer days of computing, lacked a tractable theoretical structure that computer scientists could get their teeth into.
The newer, relational data-model, on the other hand, developed by Dr. Codd at IBM, has a relatively simple mathematical structure: as a result, virtually all recent scientific work on database management systems has been based on this model. This means that any database management system that hopes to be state-of-the-art will be relational in form.
For the publisher who uses one of these systems, though, all that really matters is to learn to cast information into tabular form, and to create new tables from combinations of old ones. This turns out to be quite easy to learn. A number of visitors to CHP had short courses in the technique on our pilot system.
In the database design used in CHP's pilot project, four tables contain all the information for publishing order-fulfillment and promotion:
adr (addresses) a table of customer information
invs (invoices) a table of general info about each invoice
sales a table of invoice "details"
titles a table of info about each CHP-published title 
Each row of the adr table contains a unique mnemonic for some customer, plus that customer's mailing-label data and, in separate columns for convenience, information which classifies him or her by province, postal-code area, and type (such as institution, bookstore, library, individual, etc.).
Similarly each row of the titles table contains a unique mnemonic for each title published by CHP, the year of publication, the full title, the author's name, the current retail price, and the current inventory level.
For every invoice sent out by CHP there is a row in the invs table which contains the invoice number, the customer's mnemonic code (matching some one row of the adr table), the invoice date, the total dollar-value of the invoice, and various and sundry other information.
The most important sales information is in the sales table. Each row of this table matches a line on an invoice. It contains a title-mnemonic (matching some one row of the titles table), the quantity of books ordered (or returned), the discount allowed on that title for that customer, the retail price obtaining at the time the book was sold, the invoice number, and (redundantly) the customer mnemonic and the invoice date.
A special program was written to provide a fast and flexible way to enter orders. This program checks, as far as is practicable, the validity of the information which is being keyed  in, and refuses to allow the user to enter information which is not consistent or not complete. It at least warns the user when the information seems unreasonable. To minimize the work involved in order-entry, the program supplies the most likely value for each piece of information and inserts that value if the user does not type in a different value. For example, if the user gives the first title in an order a 40 percent discount, the program assumes that subsequent titles on that order will have a 40 percent discount unless it is overridden.
After orders for a day are entered, the user can have invoices printed automatically on the system's line-printer while he or she goes on with other work. Packingslips are also printed.
All the information on an invoice or credit note is added to the body of information in the database: that is, a new row of data is added to the invs table, one or more rows are added to the sales table, and the current inventory levels are adjusted in the titles table.
Besides keeping track of orders, the database is available for special ad hoc queries. For example, when a CHP author went on a cross-country tour, we were able to equip him with a listing of all the sales of his books to bookstores in the cities which he would be visiting.
To give an idea how easy this was to do, here is a slightly easier-to-understand query, which gives the mailing-label  information for all bookstores in British Columbia. Someone would type the following line on a terminal:
select lab from adr where prov = 'BC' and inf match '*bks*';
In this query, "lab" is the name of the column of the "adr" table which contains the mailing-label. The "prov" column contains the two-letter standard province or state code used by the U.S. and Canadian post offices, and "inf" is the name of a column of useful information. The code "bks" has been added to this column for customers which are bookstores.
On the Zilog system, this query took just over 15 seconds to return 73 bookstore mailing-labels, selected from the 1,226 unique customer entries making up the "adr" table.
Much more complex queries can be made which involve information coming from several tables at once. The list required by the traveling author was such a query: it needed to reference both the "adr" table (for the customer addresses for example), and also the "sales" table which contains data on which customers bought which books and when.
The publishing database which we created on the Zilog can form an important resource for tailoring our promotional efforts for new books to those customers most likely to be interested. With an expanded body of sales information (perhaps by incorporating information from other publishers), direct-mail promotion can become a potent tool. 
Yet new uses of this database are appearing every day to surprise us. For example, recently Stan Bevington wished to analyze the collectibility of past-due accounts. With Mistress he was able to extract that analysis from the database in a few minutes. Indeed, it was so easy to do this that once he had the results on his terminal screen and had looked at them, he didn't bother to print out the analysis: after all, he could always recreate the analysis if he wished to contemplate it again!
For years Coach House Press has had programs running in its Datapoint computers to compute aged accounts-receivables, inventory-level reports, monthly sales reports, and to generate monthly statements of account. Considering the limitations of the Datapoint equipment, these programs did a good job. So our main concern during the pilot project was whether it was going to be very difficult to recreate these functions within the new, UNIX environment.
It turned out that with the Mistress DBMS these reports were much simpler to set up than they had been on the Datapoint. In addition, they were much simpler to use on UNIX.
Nevertheless, our reaction to these things on UNIX was more relief that they worked rather than excitement at something new.
In a way, our experiments with typesetting on a UNIX  timesharing computer were less exciting than our database trials. This was because typesetting at Coach House Press has been a highly computerized activity for many years so that the qualitative results were already well understood.
For instance, the way manuscripts are keyed in at CHP has become standardized in a form which is independent of the particular phototypesetting equipment or, for that matter, the particular computer equipment that is being used. This "style" has been used on Apples, Datapoints, and now UNIX systems. It concentrates on the intent of the author, editor, or book designer rather than the particular sequence of actions required of a particular phototypesetter in order to achieve that intended effect.
(In this respect the CHP input style differs fundamentally from the newly-promulgated Federal Government standard for word-processor prepared manuscripts. This standard frees Government typesetting purchasers from prior commitment to the particular set of codes used by any one typesetter, but still requires the typist to consider the "how" of typesetting rather than just the "what".)
Nevertheless, there were dramatic quantitative changes in typesetting with the new system. We confirmed what we had suspected for some years: that typesetting with small standalone computers was very costly because of the "housekeeping" required. That is, typesetting staff were spending many hours a week just managing the storage of manuscripts on the over 300 diskettes  which we have acquired over the years, and, because diskette drives are slow on the Datapoint -- they are much slower still on personal computers -- and the Datapoint computer is simply not very fast, people spent more time than they realized just waiting for the computer to do something.
With the UNIX system the "housekeeping" virtually disappeared as a regular activity, and waiting-time decreased because computers which can run UNIX are much faster than Datapoints or personal computers. Moreover, since all current work is "on-line" at all times, any typographer can access the files associated with any book from any terminal. Indeed, two typographers can work on the same book, so long as they take care not to work on the same "galley" (file) at the same time.
One typographer and his wife have just had a new baby. He will need to stay home much of the time for the next few months, so he has arranged to have a terminal at home, connected to UNIX, so that his work will not suffer unduly.
In addition to the effect of many terminals sharing the common pool of text files, there is the benefit of multitasking: one typesetter at a single terminal can often have several things on the go at once. For example, when a chapter has been typed into the computer it is desirable that the typographer check that chapter for coding errors or obvious spelling errors, using special computer programs for the purpose. With a simple one-task computer like the Datapoints or almost all personal computers, it is necessary to sit around waiting while the  computer performs these routine checks because it is not possible to use the terminal for anything else while the checking is proceeding. But a UNIX system -- no matter what brand of "hardware" it is running on -- supports multiple tasks at the same time. The typographer typically finishes one chapter, sets the spelling and code checking program running on it "in the background", and starts immediately on the next chapter. At some time later it is easy to check to see if there were any code or spelling mistakes identified by the checking program.
Some recent personal computers have a feature which allows the user to have a file printed out on a computer printer while the computer is being used for another task. This is a special case of the "multitasking" described above, which for UNIX is hardly special at all. This feature is called "spooling"; its special character lies in the fact that although any number of people can simultaneously request that files be printed, the spooler will queue them up and print only one at a time. With a computer which has many terminals attached to it this is quite essential, especially when, as at CHP, the terminals are scattered all over a rambling old coach-house out of earshot of each other!
The Typesetting Department has carried out one interesting experiment in combining typographic methods with techniques normally associated with data-processing: they have used the Mistress Database Management System to organize and manipulate  the text for a catalogue before typesetting it. This experiment was not without its problems, but it did demonstrate the principle.
It is our intention to develop a Unix-based, typesetter-independent hyphenation and justification package for use by Canadian publishers. Because this package is not designed to run with just one model of typesetter, the publisher using it is not committed to a particular typesetting firm by using it.
The idea is to provide a means for publishers to key in their books themselves (which many want to do anyway) and then preview the way the pages will look before sending the typesetting file to the typesetter. We expect to be able to negotiate heavily discounted prices for typesetting under these circumstances, while at the same time the publisher gains better control over the design of its books.
Of course we do not intend to develop such a software package from scratch. Instead we plan to modify and enhance a package from Bell Labs called the "device-independent troff". The work will take from six months to a year to complete once the installation of the new computers is finished.
The result will be similar to what has happened in virtually all large newspapers over the past decade: editorial personnel now have absolute control over the content and appearance of the newspaper, though at the price of having to specify many things  precisely themselves instead of merely "hand-waving" and having the compositors make the detailed decisions.
There is no black-and-white here: some publishers will still prefer to have their printer worry about details, while others will want to do it all themselves. And there are all the shades of grey in between those two extremes. But we intend to make sure that publishers have the choice.
After nine months of Phase One of our Pilot Project, we felt confident enough of our understanding of the issues involved to shop actively for a computer suitable for the needs of all three publishers. We were anxious to find one that would do for all three -- even though the needs of each were not identical -- because we had the best chance of negotiating the best price if three computers were involved.
In addition to the Zilog brand (on which Phase One had been conducted), we looked at the following brands: Wicat, Altos, DEC (of course), and Pixel. We chose to carry out detailed tests on Pixel computers for a number of reasons, not least because we discovered that Pixel and their Canadian distributor, Andor Systems of Toronto, were very anxious to get established in the Graphic Arts Industry in preparation for a new product which they felt would do particularly well in that field. Therefore, we felt (and we were certainly right!) we would get good cooperation from Pixel, and an excellent price. 
Phase Two of our Pilot Project was to try to "port" (i.e. transfer over) to a Pixel all the programs and databases and typesetting files that we had built up on the Zilog.
The only common medium that we could use for this transfer was the telephone line. The process was not without considerable suffering, but it was done -- all seventeen million characters' worth -- in the course of a month. If we had to do it all over again we could -- with hindsight -- arrange to do it in a week. So the experience was very instructive.
The Zilog computer, by the way, was an excellent machine. We did not choose it for our proposed permanent system because it was much more expensive than the Pixel and a small system suitable for two of the three firms could not be expanded later into larger systems without considerable additional expense. But we did find it a good, reliable machine with well-developed software.
The Pixel, both the hardware and its Unix operating system, is not as "mature" as the Zilog -- problems can still be uncovered with it -- but the Pixel factory is very responsive to our suggestions and we feel that having to sort out some problems now -- Coach House staff are experienced enough to cope with them -- is a small price to pay for the great dollar-savings we make by buying a Pixel. We expect that the problems will all be ironed out before any of us take delivery of our computers (in August at the earliest, assuming that orders are placed very soon). 
We found very few procedures which had to be changed when going from one Unix computer to another. What problems arose were usually accommodated within minutes of identifying them.
The Pixel computer performed some tasks faster than the Zilog did, and some tasks slower. The particular machine we had for testing did not have up-to-date disk drives and this slowed it down (in comparison to the Zilog) when the task required moving much data to and from disk. However, our agreement with Andor Systems requires them to supply us with faster disk drives as soon as they are available, and even if we place our order very soon it is probable that these new drives will be available before we take delivery in late summer. Thus we expect that the Pixel we actually buy will be faster than the Zilog computer on which all our experiments were conducted this past winter. We are quite sure that that is fast enough for the next few years.
Coach House is currently cooperating with Pixel and Andor Systems in cleaning up problems with the Pixel version of Unix. We feel that the good rapport which we have established will be of inestimable value in the years ahead.
In the course of the Pilot Project, two Canadian writers were given the opportunity to try out the UNIX operating system as an alternative or as a complement to personal computers. Their own accounts of the experience are presented here without comment. 
I've been using the experimental UNIX system at Coach House to do editorial work for the press and to prepare pre-formatted copy for the magazine Open Letter which Coach House typesets and prints. By accessing the system by telephone and acoustic coupler from an Apple personal computer running a terminal program called 'Visiterm', I'm able to do work from my home or office that previously I could do only by visiting the press.
As an editor, I now enjoy access to Coach House files similar to that of regular Coach House employees. I can inspect the text of books I am editing for the press, make editorial changes, and -- using the electronic mail facilities of the system -- keep in close touch with production workers about the progress of books I am editing for the press. I often prepare promotional materials at home on an Apple and mail them into the UNIX system over the telephone. Later, I can access these from my home or office and revise or enlarge them at will. In general, the electronic mail provisions of UNIX allow better communications than the regular telephone, because I can leave messages for Coach House staff 24 hours a day. I no longer have to trust to oral messages being relayed or wait for the necessary person to be available at the phone.
With Open Letter, I usually prepare computer files of an entire issue on an Apple computer, format these with the relevant Coach House typesetting codes, proofread them on the screen, and then transfer them over the telephone to UNIX. Once they are in  UNIX, I can access them on an Apple, check them for errors, run them through the UNIX spelling program, and then notify Coach House to proceed directly to type. Not only does this process result in Open Letter having more efficient typesetting than before, but it gives me as editor much more control over copy and format. It speeds production of each issue both by speeding my work and by allowing Coach House to proceed quickly in confidence that the files have been accurately prepared. It eliminates two stages of printed proofs, allowing us to proceed directly from electronic files to a typeset mock-up of the finished book.
One other thing that writers associated with Coach House have liked about the new system is that we can mail new poems to one another almost as soon as they are written. David McFadden and I have been doing this, and on occasion been able to give each other useful feedback.
As a fiction writer, poet and freelance journalist I've been doing most of my writing on the computer at Coach House Press for the past eight months and it has been an amazing experience, the latest step in a technological ladder that began when I started writing with a pencil as a child, then moved to a straight pen and ink, then to a fountain pen, then to a ballpoint pen, then to a felt pen, then to an old Smith Corona typewriter, then to a new Olympia electric portable, and then, for two years while teaching at David Thompson University Centre in Nelson, B.C., to an Apple II with disc drive and now to a large mainframe with more memory  than I can possibly use.
I use it only for writing so far, although I realize it has much greater potential, particularly for the bookkeeping and business end of my work, but I agree with the science writer Lydia Dotto who says that any writer working on a word-processor will immediately realize a threefold increase in output, or a two-thirds saving in time.
Further, I feel as if the quality of my work has increased dramatically, as it is quite easy for me to take greater chances with my writing, to branch off on tangents that I ordinarily wouldn't bother pursuing simply because of the ease with which the mechanics are taken care of and the simplicity with which something can be erased or altered. Much less energy is burnt up with the job of keeping track of papers, files, cutting and pasting, and that energy is therefore always on tap for purer imaginative and creative processes. I find I want to spend more time at the computer than I ever wanted to at the typewriter, and therefore my output is more than tripled. It would be difficult to go back to an Apple, never mind to a typewriter.
And the copy I deliver to my editor is always immaculate.
Our proposal is to buy essentially the same equipment and software for the three presses, with the one exception that The Coach House Press needs to accommodate more terminals than the other presses. 
Keep in mind that The Coach House Press has already acquired some terminals and other hardware and software for the Pilot Project and thus these things do not need to be purchased now for the final configuration.
The computer will be a Pixel 100/AP with one million bytes of main memory.
Secondary storage is specified as two 20-million-byte Winchester disk drives, but we expect that the supplier (Andor Systems of Toronto) will substitute for these a single 85-million-byte disk drive for approximately the same price by the time that we actually take delivery. This newer-design drive is much faster than the ones available today.
Two forms of "off-line storage" are in the specification. For preserving copies of all our on-line files ("doing backup"), we require a Kennedy data cartridge drive, which stores up to 23 million bytes of programs and data on a single data cartridge. (A data cartridge looks like a giant audio cassette and costs roughly $40 each.) For archiving the texts of books and other records in a form which is quick to retrieve from, we also have specified a floppy-disk drive. This drive uses the 8-inch variety of floppy-disk and can place up to 630 thousand bytes on one diskette. Diskettes cost only about $5 each.
At The Porcupine's Quill and Penumbra Press the computer equipment must be protected from "spikes" and "brownouts" in the local Hydro power. Special equipment is provided in the  specification for them.
The terminal and computer-printer equipment for The Porcupine's Quill and Penumbra Press is not as extensive as that for The Coach House Press. We have provided for only two terminals and a relatively cheap printer for each. The printer will plug into the computer via a "parallel port", and the terminals via "serial ports". The minimum number of serial ports available is four, and that is what is specified. One of these serial ports will be connected to the telephone network through a device called a "modem". This leaves one serial port spare for future expansion. (The port used for the telephone line could be shared with a terminal too, so both presses could expand to four terminals without additional equipment on their computer.)
At The Coach House Press, the computer will have two parallel ports and eight serial ports. This is the maximum number of serial ports available with the Pixel computer at the moment. However, as soon as 16 ports are available, Coach House will upgrade its computer to 16 ports since 8 is really not enough for the activities planned.
All of the above is "hardware". The software packages to be bought are, first, the UNIX operating system (for all three presses), and, second, the Mistress database management system (for The Porcupine's Quill and Penumbra Press: Coach House had to purchase Mistress for the Pilot Project). The programs which Coach House has developed for use with UNIX and Mistress will be supplied to the other two presses free of charge. 
A further software package which is important to our plans for the use of the computer systems, but which cannot be purchased as part of this proposal because of the timing, is the "device-independent troff" typesetting program developed by Bell Laboratories. Coach House intends to negotiate the distribution of copies of this program -- adapted to the needs of Canadian publishers and typesetters -- for a price which cannot be higher than $6,000 but may be under $2,000.
Installation of all three computer facilities -- and training of personnel -- will be overseen by Coach House staff and consultants. Coach House will continue to supply advice and software to The Porcupine's Quill and Penumbra Press via the UNIX electronic-mail facility, which uses the telephone network when the rates are lowest, and works automatically. 
Quotations are attached to this report for most of the pieces to be bought. Here is a summary:
CHP PQL PEN ======== ======= ======= Computer System $28,710 $27,415 $27,415 UNIX Operating System 1,170 1,170 1,170 Mistress DBMS - 2,250 2,250 Computer Printer 4,442 2,170 2,170 Terminals - 3,772 3,772 Telephone Modem - 1,000 1,000 Hydro-power Conditioning - 1,200 1,200 Diskettes & Tape Cartridges - 1,000 1,000 ======== ======= ======= $34,322 $39,977 $39,977 TOTAL COST OF PROJECT..............................$114,276 FUNDING REQUESTED...................................$68,565
– April 30, 1983
– July 15, 1983
In my report of April 30, 1983 to the Department of Communications, I omitted any detailed discussion of the workings of the network formed by the computers proposed for my three clients. I did not feel that the complexity of such a discussion was warranted because it would not contribute significantly to the viability of the proposal in the eyes of the D.O.C. It appears now that I was wrong. This Addendum will, I hope, correct the deficiency.
My clients have a long-standing tradition of cooperation in publishing projects, supported by jointly-held goals and demonstrated by their ability to share a common group of authors. It is natural, therefore, for them to prepare for the high-technology future through joint planning. Moreover, no notions of "proprietary information" inhibit them from engaging in the sharing of marketing information and in supporting each other according to their individual specialties.
Thus, the project now proposed came into being. Three fiercely-independent small literary presses are eager to join in a business symbiosis never before imagined in Canadian publishing. The three presses will each buy and own a computer -- therefore no one press is dangerously dependent on another -- and run software which gives each press access to the information-**resources of the others.
The full extent of this symbiosis can hardly be determined until it is tried. But, at the very least, the sharing of sales information and publishing plans on a daily basis should have a dramatic effect on the efficiency of marketing, and therefore on revenues.
The technology that makes such a plan possible -- computer networking -- is particularly easy to use within the context of the Unix operating system. Examples of this use are given.
Clearly, literary presses do not compete in any economic sense, and they never will. This is because the major hurdle to overcome in literary publishing is simply to reach the relatively-small and highly-dispersed market. Once the product of a literary press is in the hands of a prospective buyer -- a difficult feat to get that far -- price is probably the last determinant of the purchasing decision. The decision will depend almost entirely on the purchaser's individual taste, and since tastes vary so widely, a great many small independent firms are needed to fill all the ecological niches in Canada's cultural life.
So cooperation between literary presses in the prosecution of their business is natural and healthy. The Coach House Press, The Porcupine's Quill and Penumbra Press (CHP, PQL and PEN respectively in the rest of this report) have a sound basis for  such cooperation. CHP and PQL have the same type of printing press, and PEN will probably acquire a similar model in the near future. CHP and PQL have for years made joint purchases of printing paper from their common supplier in order to qualify for a volume discount.
The three presses publish from a common pool of authors -- several authors have published works with more than one of the three presses. And all three are active members of the Literary Press Group of the ACP.
In the future, with the aid of the proposed computer system, the three plan to share in the publication of anthologies drawing from material originally published by the presses individually.
Because the proposed computer system will make it feasible at last, an immediate project is to coordinate marketing and promotional activity, through joint use of sales information from all three presses. It is hoped that, through the interactive use of the combined sales database, promotion can be targeted more accurately to its intended audience, to the mutual benefit of the whole group.
Administrative tasks are the bane of a small enterprise, because often they involve as much work as for well-staffed major corporations. A case in point is the variety of title analysis forms (e.g., Quill & Quire, Telebook, Statistics Canada) which publishers are required to fill out. In the future, using the Unix network, one firm will develop (on behalf of all) methods of  extracting and formatting the analyses from the combined publishing database. Similarly, promotional mailing lists will be shared by the three firms.
Most small publishers today continue to pursue the illusion of economies of scale. The accepted wisdom is that the only way to operate more cost-effectively is to operate at a higher sales-volume per title.
But economies of scale, real though they may be in large-scale manufacturing, result as much from the particular technology of mass production than from the volume of sales. Given that the volume of sales is small, there are alternative ways to organize production to achieve lower costs. Now that computers can be involved in production, many industries are eyeing "flexible manufacturing": the use of a quite different, automated plant to produce very small batches (or even single copies) of manufactured goods.
The goal of CHP, PQL and PEN is similar: instead of putting all their hopes on achieving high volume per title (which is clearly impossible for avant-garde material), they are working on technology to make low volume production less costly.
The key to lower costs, we believe, lies in moving the function of typography -- including every detail of book design -- from Production (which most publishers farm out to graphic-arts firms) to Editorial. Editorial personnel must learn to  integrate design decisions with traditional editorial decisions, using powerful typographic software running on their computer as a guide to what is possible and desirable.
By integrating some of the traditional functions of Production with Sales, Editorial, and Management functions, not only intra-house but inter-house, and learning to reduce compilation costs, my clients will be providing all Canadian literary presses with the means to meet their avowed publishing objective: to place works of excellence in the hands of a targeted audience for a well-defined margin of profit.
This new arrangement of functions means that a very sophisticated computer program must reside in the publisher's computer system -- a program too large for the small-business and personal computers that are most common today. Only the next generation of 16-bit computers -- the "small mainframe" in a large shoebox -- will provide the capacity to deal with such a sophisticated software package.
More aggressive marketing also requires new techniques of information processing, which need the powerful computers. This is especially so since better marketing can come about only by many firms cooperating with the common aim of alerting the public to their varied (and to some extent interlocking) new material. This mutually-beneficial and highly interactive cooperation needs sophisticated data communications software, which only larger  computers can accommodate.
Furthermore, only "small mainframes" can, at this time, support the kind of user interaction my clients need for efficiency. Personal computer and wordprocessor screens look prettier and less frightening at first, but they do not support the "multitasking" (doing a number of things at once) needed to make maximum use of the human brain, and they do not have the capacity to run the necessarily-large programs which make the best use of the time of the person sitting at the keyboard.
Many of the most sophisticated applications which we plan to run on this system will actually run "in the background" while people do other things at their terminals. The three computers will be working along side people instead of always requiring their direct attention. Personal workstations like Apples and Decmates cannot support this background activity.
The high price of these new 16-bit computers is a deterrent today. But within a couple of years that price should drop to a quite affordable level, and unless some publishers start now to develop the techniques needed to put them to effective use, most Canadian literary presses will be unable to benefit from the new technology even when they can afford it.
The proposal is for a single, distributed computing facility shared on an equal basis by all three presses. "Distributed Computing" is a computer industry buzzword meaning that the "intelligence" (i.e. the computing capability) in the system is not located all in one place but is spread over a whole building or a whole continent, with all the pieces of the system working harmoniously together for common goals.
Most early examples of distributed computing were "hierarchical": there was one "central" site which possessed most of the capability to process large programs and to store files, and many "remote" sites which performed small tasks largely under the control of the "central" site. 
But, concurrently with the development of this approach, another kind of distributed system was developed by computer scientists for the United States Department of Defense. An experimental version was set up joining together a number of U.S. universities and defense research establishments. This network, called ARPANET, was wildly successful and is still in use today. Our planned distributed system will look very much like ARPANET.
In such a system, no one "node" of the "network" controls the work of the system. The network is "anarchic": things are arranged so that workers at the different sites can easily cooperate with each other to do things which require cooperation, and to continue on their own ways when cooperation is not required. Since there is no "central" site, each site has roughly as much computing capability as any other (though this does not preclude some sites having special capabilities which the other sites can use when they need to -- e.g. special printers or phototypesetters).
Ironically, the military-supported research which made ARPANET possible was motivated by a concern which touches our own pacific need: if there was no one "central" site there was no single location which, if bombed, would put the whole network out of operation. This argument is not without merit in the current proposal: there is no one firm in our network which, if it collapsed, would destroy the ability of the remaining firms to make use of the (reduced) system. Thus this kind of distributed processing minimizes risk as much for us as for the military. 
Each publisher will have his own computer. The three sites will be linked by telephone lines which allow messages, texts of books, and software to move easily between them at a rate of at least 100 characters per second.
The resulting "distributed system" is "loosely-coupled", for a data-transfer speed of 100 characters a second is really quite slow! But because of the sophistication of the part of Unix which deals with inter-Unix communications, messages between people at different sites is as easy as between people at the same site. If Stan Bevington at CHP wishes to send a note to Nelson Adams (also at CHP), he types:
and then types his message. If he wants to send a note to John Flood at Penumbra Press in Moonbeam he will simply type:
and then type his message. In the first case (the local message), nelson will be notified that he has incoming mail within seconds (and can read it any time after that, at his convenience). In the second case (message to a remote site), the delay before John Flood gets a "You have mail." message on his screen depends on the general instructions Unix is given about when it may make long-distance calls, and could be anywhere from a few minutes to a few hours. Note that "messaging", as distinct  from regular phoning, is often just as good as and sometimes even better at communicating, provides a natural record of the exchange, and is invariably much cheaper. This is because the message, composed at typing speeds, is transmitted at 100 characters a second: a four-page memo costs a minute and a half at most to transmit, but would take much longer to read over the phone.
Likewise, transferring text files or software is done almost as easily: To send a collection of files constituting the text of the book "Joeboy" from, say, Porcupine's Quill to Penumbra, Tim would type:
uucp Joeboy/* pen!~/Joeboy
and the files would appear on Penumbra's computer within minutes or hours, depending on policy set for sending files. (Files might be sent only between 11 p.m. and 8 a.m. to get the best long-distance rate.)
Finally, sharing of new software improvements, which will go on continuously for some years, can be handled the same way as text of books. Programming support will normally be situated at CHP (which has access to computer-science and engineering students at the University of Toronto), so the automatic coupling of the three computers is essential so that PQL and PEN can readily communicate their needs (via the "mail" command) and so that programmers can install new or improved programs on their computers. 
It should be emphasized that Unix-to-Unix communications uses a sophisticated "packet-switching protocol" which guarantees that messages and files sent between the computers are not damaged or changed in any way by bad or lost telephone connections. This is in contrast to the use of the telephone to transfer files from, say, an Apple or a Decmate wordprocessor. Since the slightest fraction of a second of "noise" or other bad connection will damage the data being transferred, we have found no better solution so far than to transfer each file from an Apple or Decmate twice and then to compare the results. This is a time-consuming process requiring careful attention (though the actual comparison is done by a standard Unix utility program). Better solutions can certainly be designed, but they require expensive programming-time and would have to be re-done for each of the great variety of personal computers and wordprocessors.
Also, Unix-to-Unix communications is handled automatically. All the user has to do is tell Unix to send a file or message. Unix itself arranges the telephone connection -- at times which it has been told are acceptable -- and "tries again" if necessary when it cannot get through or the line is lost. The users at both ends are unaware of all this activity, and, of course, go on with other work (or go home) after typing the mail or file-transfer commands given above.
It is even possible for Tim in Erin, Ontario to run a program on the CHP computer and get the results sent to his own (or even John's) computer, without involving any person at CHP,  simply by typing (say):
uux chp!ls > pql!~/chpfiles
This command, typed in Erin on the PQL computer, would cause the program "ls" to run in Toronto on CHP's computer and to send back to Erin its results: a list of all of tim's files stored in Toronto. A similar approach will be used by each publisher to use databases stored at other sites.
These examples may seem complicated, even arcane, but only a day's exposure to Unix would make them quite understandable. Moreover, it is common for such commands -- if they are used frequently -- to be simplified into one- or two-word commands (called "shell scripts"). Individual workers at CHP already have, typically, one to two dozen such shell scripts which they have created for their personal use. Thus each person using Unix has, to some extent, his or her own private "computer language" after a time.
We have already done a great deal of research and development, and, once the three computers are installed, will carry out a great deal more. Thus we will develop a smoothly-functioning networked system, so that literary presses may cooperate effectively and may benefit jointly from all the new software tools which can now be provided. 
We are doing this at a time when the cost of the necessary hardware has dropped to the level where it is just affordable -- provided we are given the requested joint grant. By two years from now, hardware with the same capability will almost certainly have dropped in price to half what it is today -- and many other presses will be able to join in networks (hopefully linked to our network) for mutual cooperation, provided we have developed the software, procedures, and instructional materials already.
We are paving the way (with research conducted while the technology is still generally too expensive) so that industry can use that technology when the price drops (as it always does with things electronic). In this case we are paving the way for an industry which is too poor to fund the research itself, even if it had a general appreciation of what might be accomplished.
It is our hope that our pioneering experiences will motivate many other small literary presses to emulate us. Indeed, the unusually-high cost of our project could properly be thought of as R & D costs amortized over the many future project proposals which achieve similar results but at much more affordable cost.
We would suggest that even next year, some other firms be funded to join our network with similar (but already cheaper) equipment. This will increase our experience at spreading the technology (and undoubtedly give us a few more headaches to deal with once-and-for-all), so that when the "deluge" comes a year or so later we will be able to hand over our experience neatly-packaged. 
We assert again that all small literary publishing firms really need the capabilities of the "small mainframe", and that these capabilities (including network functions) will be available in even personal computers within a very few years. Our goal is to make our impoverished industry ready to make use of the powerful computers of the future as soon as they are affordable -- instead of waiting in ignorance until well-financed industries have already adapted to them.
– July 15, 1983