DTACK GROUNDED, The Journal of Simple 68000 Systems
Issue # 23 Sept. 1983 Copyright Digital Acoustics, Inc
Things seem to be picking up in software land. And, as we write this, we are just about to personally get out from under a big, big hardware burden and get back on the track in re: HALGOL. In the meantime, some other folks aren't waiting for us.
"NATIVE CODE COMPILER FOR FORTH COMING!
"Is FORTH still too slow, even on the 68000? A new compiler produces 68000 machine code for low level functions for the ultimate in speed. It handles arithmetic, stack operations, shifts, logical operators, and simple IF-ELSE-THEN and loops to produce a speedup of around a factor of 5 over ordinary FORTH. E.g., the BYTE sieve benchmark runs in 1.6 seconds, compared with 10.68 for interpreted FORTH on a 12.5 MHz 68000 and 1.55 seconds for DEC FORTRAN on a VAX 11/780.
"The code is compatible with the FORTH already available for the 68000 and should be available in October."
Bruce W. Walker
900 W. 14th St. #10
San Pedro CA 90781
MINOS 1.0 ---------
"Not content to rest on our laurels, we at PHASE ZERO, LTD. have created a new product, bigger, better, more sophisticated than anything the world has ever seen! (NB: our copyrighters are rented.) MINOS 1.0 is a simple operating system for the DTACK Grounded 68000AP and Apple II (+/e). Perfect for software development, MINOS provides a Monitor for direct machine language control of the 68000AP, and an I/O kernal containing a large (and growing) set of I/O routines for accessing Apple hardware and peripherals.
"Best of all, MINOS has no licensing requirements. Software based on MINOS can be sold and distributed with no fees, no hassles. The reason is simple: if we provide good support, and make MINOS easy enough to use, then WE may not have to rewrite our software to use someone else's operating system! Clever, huh?
"So what do you say? If you're thinking about developing software for the 68000AP, then MINOS is probably just what you need. If it's not, we'll help make it so, and provide source code to make modification easy. Do you know what it costs to license CP/M, let alone the source code? This has got to be a bargain, at only $40 (U.S. and Canada; $45 elsewhere; U.S. funds) for MINOS 1.0. You get a DOS 3.3 disk and a 40 page manual. (Does not include source code.)
"MINOS source code, on disk, will be available in August. Price is another $40 (U.S. and Canada; $45 elsewhere; U.S. funds)."
PHASE ZERO, LTD.
2509 N Campbell Ave Ste 13
Tucson AZ 84719
"I was writing a manual for my compiled BASIC when issue #21 of the newsletter arrived. I was having to go to the source code more and more often to remind myself of the syntax. When I read Dr. Hull's offer to act as a clearing house for software, I decided to modify the manual so that it could be read, hopefully, by others and throw it into the software pot to be used until a more complete BASIC becomes available.
"A copy of the 41-page manual is on the disk. The introduction states what software is used from a DTACK disk. I assume from what you have said in the newsletter that you don't care about that [Right! - FNE]."
Chet S. Warwick NY
From Jeff Hull:
"Yesterday I received Chester Sensenig's compiled structured (tiny) BASIC for the DTACK boards... Obviously a fair number of folks will want this package. It creates some real impetus to the software exchange idea... The idea of sending disks to get software has perhaps caused some to delay. Additionally, the catalog now reads three disks, so it could be cumbersome for people. So I'll offer to send
(continued on page 26)
On the very day we mailed issue #22, we received both E.E.T. and E.H., the two electronics industry weeklies. Both featured long articles suggesting that T.I. had goofed in the home computer market. Both featured short articles about troubles in the Intel 186 chip. EET's article (18 Jul, p.102) had a few interesting sidelights on this matter. For instance, the 186 reportedly has over 400 design wins already, and Intel is hoping for a 1000 wins by the end of the year. This will come as a surprise to you readers?
The 186 has had three mask revisions and is currently undergoing its fourth, described as the "final revision" by Intel's Tony Barre.
A company called Victory Computer Systems (not Victor) had a board all ready to go for the 186 last November. When they could not get the 186, they redesigned the board for the 68000 and are now profitably selling the resulting 68000-based computer. Roger Vass, Victory honcho, says "...it is ironical that our success with the 68000-based machine should be a result of Intel's misfortune."
In John Dvorak's column Mitch Kapor, chief programmer of 1-2-3, reveals that the program was originally programmed in C but "it was a dog", meaning that it had very poor performance. This is partly due to the fact that the 8088 is a poor match for the C language. The 68000 is a better match but the PDP-11 is best. Surprise?
Atari has just surprised everybody by the size of its second quarter losses: $283 million! That includes a loss of $310.5 million in its home computer division. That division reported a profit of $358 million total for all four quarters combined in 1982, when the division was generally perceived as successful. The L.A. Times quoted one financial analyst thus: "It's mind-boggling. I never expected anything of that order of magnitude."
Two days previously, T.I. had reported a pre-tax loss of $126 million for its second quarter, somewhat higher than the $100 million estimated earlier. This is the first time in the 30-year history of T.I. as a public company that it has reported a loss in a quarter. Having lost all this money in the home computer market, T.I. has naturally officially reaffirmed that it will stay in that market.
To increase their profitability, T.I. is threatening to
sue any third parties who have the sheer effrontery to write software for T.I.s home computers. Will someone explain that to us? We miss the logic of that plan...
Mattel is projecting a $65 million loss for their second quarter, vs. profits of $67 million for all four quarters combined last year. Mattel is predicting "substantial" losses for this full fiscal year.
Most of you have already heard about the new wonderful coming down the software street, Modula-2. This is going to become conventionally 'popular' more slowly than you might think because its logical clients are Pascal programmers and the Pascal programmers are going to need a face-saving device before they can drop Pascal and announce that Modula-2 is the latest wonderful.
Are we saying that Modula-2 is going to bomb? Heck, no. We don't know enough about it to make a judgement. We will say that we like what we have learned about the underlying theory of the language, but we want to learn more about its implementation. Motherhood is a wonderful concept, but thirteen welfare babies is a poor implementation of that concept.
Allen Munro has written a long article about Modula-2, which begins on page 109 of July (IBM) Softalk. Here are some excerpts from that article:
"The many benefits of Modula-2 are partially offset by some disadvantages. The simplicity of Modula-2's syntax imposes some extra burdens on the programmer. One example is the lack of generic I/O procedures. In Pascal, the program can use the single procedure write to output integers, real numbers, characters and (in UCSD Pascal) strings. In Modula-2, variables of each type must be output with a procedure that takes parameters of that type. This means more lines of source code, but the compiled code is no less compact or efficient than that produced by a Pastal compiler. The Modula-2 programmer must do some of the work that is handled by the compiler in Pascal.
"Many programmers may find that Modula-2's case-sensitivity is an annoyance. ...the Volition Systems implementation provides a compiler option to turn off case-sensitivity. [is that an extension? - FNE]
"Modula-2 is an advanced prograssing language. Its enforcement of structured program development makes it an appropriate vehicle for attacking major software projects."
We still want to know whether the dimension of an array is part of its data type in Modula-2.
This is ol' POPS here with some more fascinating insider stuff about the personal computer industry, written just the way you like it!
Bad, bad news for most of you kiddies with 16-bit processors. Nat Semi is about to completely obsolete your PACE or INS-8900 based 16-bit machines. We have it on good authority that the 10MHz 16032 will be released soon. The good authority is a press release dated May, 1982 which asserts that the 16032 will be available the very next month! Since it is now after that very next month, the 10MHz 16032 is therefore available now (you will not believe how shrewd ol' POPS is, folks!).
Well, your PACE or INS-8900 based machine has surely given you good service these last five or so years, so you will not mind that it is now obsolete.
BIG NEWS from Intel this month. Over a champagne brunch at L'ESCOFFIER with our very good friend Bob Noyce we learned that Intel is about to release its silicon masterpiece, the iAPX 432. This is the chip set that is so easy to program, it is going to free up ten million programmers to work in the fast-food industry. A good thing, ol' POPS tells you: The wait for a BIG MAC has become intolerable lately. By the way: the eggs in the 'Eggs Benedict' were overcooked and the Bearnaise sauce was not freshly prepared! Shame, L'ESCOFFIER!
That's not all, folks: Intel is coming out with a chip next summer that will be three times faster than the 68000! We are referring to an 8MHz 286, which you have probably not heard of. Let ol' POPS give you the rundown on this one: the 286 is SO FAST that it is not only three times faster than the 68000, it will also add two 32-bit numbers almost as quickly as the 68000! Now, that is a truly remarkable accomplishment!
Even more exciting, the 286 uses an advanced memory addressing technique called 'Segmentation'. Our highly placed Intel contacts assure us that this technique is far superior to the 'flat' memory addressing used by the 68000. (FLAT! Sounds like flat-earth stuff! Yucky!)
Confirmation of the above two paragraphs can be found in issue #5 of PC WORLD, an authoritative and entirely factual magazine.
Not only has Bill Gates got a new president (Townes from Tektronix) to take all the dogwork off his back, but he is about to reap an ENORMOUS WINDFALL of, get
this, THREE BILLION DOLLARS this year in XENIX licenses! Since Microsoft's sales have been running in the 25 to 50 million dollar region lately this will place them solidly on the profitable side. (Our thanx to MINI-MICRO SYSTEMS magazine for this tip!)
Let's hear it for Ronald Reagan! Finally, we have a President who can cut taxes, increase spending and balance the budget! As Ronnie was telling us over a late dinner at Ma Maison, he strongly supports education. The problem, he explained to us, is that everytime you give a schoolteacher some money he/she will spend it on frivolous things (like food and rent) that take the teacher's mind off his/her main job. This is why Ronnie has taken HUGE cuts out of the Federal education support budget each year he has been in office even though he strongly supports education.
We told Ronnie that a wise leader like him was welcome to join us for a late dinner any time. By the way, the Medallions of Beef served by Ma Maison were cooked properly medium-rare, but the Bearnaise sauce was not freshly prepared. And the waitress turned down our very friendly offer of companionship later in the evening. Obviously, Ma Maison has no class.
The next day we had a luncheon appointment with Pullman Carr, prexy of Context, the folks who market the highly successful MBA 2nd-generation spreadsheet package. As soon as we walked into Le Petite Moulon and saw the five martini glasses that Pullman had stacked in front of him, we suspected that we were in for a difficult interview.
"POPS!" Pullman ejaculated. "Over hear! Here?" he said, trying to hide a hiccup. Why the celebration so early, we asked. "POPS, you are not going to believe this." Pullman stated. Try us, we replied, we believe anything.
"Well, truth be said, we have been worried at Context. We kept hearing about this other software house which was writing a second-generation spreadsheet package that had about the same features ours has, and that they planned to offer that package for about $200 less than we charge! Obviously, something like that could hurted, er, hurt our sales badly!"
So, we replied, you are really drowning your sorrows in those dead martinis there? "Of course not! We have just discovered some additional information about this competition and, they have blown it badly!" Pullman declaimed. "This is a celebration! Drink up!"
How so? we asked, retaining our journalistic distance.
"Look, our customers are no dummies! They know that some day they may move on to a personal computer with a differing CPU. You know, like the Nat Semi INS-8900 or the Signetics 2650 or that Electronic Arrays chip with 4K addressable space. And when they move on it will be essential that the software we offer be transportable! BUT,"
At this point our good friend pulled himself up to his maximum height and adjusted his wig - "BUT," he repeated, "OUR CONPETITOR'S SOFTWARE IS IN NO WAY TRANSPORTABLE! And so there is no way that he (well, they) can affect our sales." MARVELOUS! we replied. What is the name of this really stupidly-run company? "There's lots of numbers. Maybe 7-8-9?" replied Pullman, just as he passed out.
We complained to the waiter that he had failed to serve Bearnaise sauce with the Tournedos, but he cheekily insisted that Tournedos did not come with Bearnaise sauce. Ignorant savage! What can you expect of a nation that eats snails? Under the unpleasant circumstances, it seemed only logical to leave prexy Carr with the check.
We trust you readers are impressed with the wide circle of our prexy friends, such as Bob, Bill, Ronnie and Pullman. Ol' POPS here only deigns to discourse (you dummies are advised that that means 'converse', which in turn means - aw, forget it!) with those persons at the very top of the ladder. This is vitally important if you want the facts, because the presidents of companies and nations always know exactly what is going on! Unless they are taking their afternoon nap.
That afternoon we met with a famous leader of the personal computer game at a Taco Bell restaurant which features gourmet burritos. We must keep the identity of this famous leader a secret, but we can tell you that Coleco has named a computer after him.
Adam, we said, (OOPS! Did we say Adam? We can't say Adam, so you will please disregard our use of the name Adam.) what gives these days? "Why, there are opportunities lying all over the landscape for those who can maintain clear vision. And, as a well-known opportunist, I an pursuing some of those opportunities" replied, uh, John Doe.
How exciting! we replied. What, we asked greedily, is the most opportunistic area available?
"You will not pass this information on to others?" asked our source. Heavens, no! we replied. "Well, there is this operating system called UNIX which is an absolute perfect match for the Motorola 68000
microprocessors. And, since I happen to have a few 87K capacity floppy disks left over from version 1.0 of the OSBORNE ONE..." at this point Adam, er, John Doe, agitatedly asked whether he had revealed a hint of his true identity. We assured his that our readers would never recognize him.
"Well, it is obvious that the first company to offer UNIX on a 68000-based machine using floppy disks will be enormously successful. And since the company I have formed to build portable computers is generating such a strong positive cash flow, I need somewhere to invest these dollars."
We recommend, we told our friend, er, John, that you invest in a company called (appropriately) FORTUNE. These people are building a machine such as the one you are thinking of, and their stock is now publicly traded. This is how good a job those guys are doing: JOHN DVORAK of Infoworld fame has praised those guys as having the tenth largest stock placement in American business history!
Naturally, we continued, these new stockholders are doing nothing but heaping effusive praise on the heads of FORTUNE's officers lately as their stock stretches for the stratosphere. What, after all, could be superior to a 68000/UNIX system based on left-over 87K floppies? "Your point is well taken" replied John.
By the way, John, what do you think of the new operating system? we asked. "Which one do you refer to?" he replied.
Why, the one which is NOT used by the one million TIMEX-1000s or VIC-20s. The one which is NOT used by the half-million CBM-64s. It IS used, we apologetically pointed out, in about 1 % of the 800,000 or so Apple IIs and about 1 % of the IBM PCs. "What is this operating system called?" asked our friend. Why, this is called a Universal Operating System, we replied.
"Well, since you asked what I think" mused A, er, John, "I think that an operating system used by so few folks can't be very good and can't be very universal" John, we firmly explained, that is not an acceptable opinion. You will, therefore, change that opinion.
"Why should I change my entirely rational and objective viewpoint?" John asked. Because, we patiently explained, we are the POPULAR newsletter editor. We tell people what they WANT to hear. And they definitely do NOT want to hear what you just said, we pointed out. With that, we walked out of the restaurant, realizing that we had forgotten to ask for Bearnaise sauce to go with our gourmet burrito.
One of the people we chat with more than occasionally is the editor and publisher of a personal-computer industry magazine. Although we share many beliefs in common, we disagree often enough to make life, and conversations, interesting. (It is evident that if two persons always agree then one of them is dispensable, no?) When we analyze these disagreements it frequently appears that the reasoning behind his beliefs appears more logical than our own. However, he recently made what appears to us to be a truly fatuous blunder.
We were discussing what we will call a procedure. He described what he called "THE" method of performing the procedure. We corrected him, suggesting that it was "A" method. No, the publisher insisted on retaining the "THE".
Why, we asked? The publisher replied, "Because that is the way BYTE magazine will soon publish it!" Is BYTE publishing ex-cathedra these days, we asked? (BYTE and this publisher are not affiliated.)
In thinking this over, it occurred to us that there was a widespread belief in many fields that there was THE way to do something. Since we have personally been making a living for over twenty years now making devices which must work and must work well if that device is to enjoy a competitive market advantage, we have long since developed the philosophy that there is much to be said for things that work.
That is, we prefer what works to theories; substance rather than form; results rather than procedures. This philosophy, originally developed in a hardware context, has carried over into other areas, especially our attitude toward software.
In designing hardware, one can almost always build a device using several differing design approaches (there is more than one way to skin a cat). The same is true of software. It is possible to write a structural analysis package in COBOL and a payroll package in FORTRAN, but who would want to? A simple problem could be solved in Ada or Modula-2 and a complex software package involving dozens of programmers and a two-year effort might be written in BASIC, but who would want to?
Regrettably, there are some folks around who will tell you that there is One True Path to righteousness and correct programming. These people are in positions of authority over substantial numbers of programmers (i.e. students) and are doing their best to see that the One True Path is followed. We would feel that we were being railroaded!
When designing a product to perform a particular function, it is our practice to look into our 'kit of tools', compare those tools with the job at hand, and select what appears to be the most effective tool for that job. We favor the same approach for selecting appropriate software tools. We prefer to do structural analysis in FORTRAN, payroll in COBOL, to use BASIC to solve simple problems, etc.
We commonly design products using the 'bottom up' approach. This means that one first selects the output transistor devices when designing an audio amplifier, and then adapts the remainder of the design to those devices. When designing a high-performance microprocessor board, we first selected the microprocessor and then designed the board around the retaining criteria (simple design so static RAM, fit inside the Commodore Pet so 6.5 by 15 inches). You will recognize this as bottom-up design.
It seems that the cabal trying to enforce the One True Path favor the top-down approach. We agree that top-down can in some circumstances be an effective design method and we keep that method in our tool-kit for use when appropriate. But we also believe that the bottom-up approach can (often) be effective, whether we are discussing hardware or software.
We are not the only one to feel this way. In Mark Dahmke's book "Microconputer Operating Systems" we find, at the end of appendix III:
"About a year ago, I thought that structured programming was the ultimate tool in the analysis, design and implementation of a computer program... I was more wrong than right.
"...I suddenly recognized a major point that I had formerly not comprehended: structured programming does not encompass the entire process of programming.
"I also learned that certain design decisions within a given program are overlooked by the main ideas of structured programming... the quality of a program can often be greatly improved by attention to the design decisions that are made in the early stages of analyzing the program design. I am including a list of particularly helpful books and articles in the references." (Mark Dahmke)
'Decisions that are made in the early stages'? Sounds like bottom-up design to us.
The people who are trying to railroad us (uh, make that 'enforce the One True Path') are, of course, the computer science types at some universities. Apparently this attitude is more prevalent on the west
coast than elsewhere. It is interesting to note that WE were accused of the same crime recently (by a West German subscriber, as it happens)!
We have been trying to make the point that anyone trying to compete in the truly massive hardware personal computer marketplace that is developing is either going to have a fully automated factory or is going to fail. We have also tried, using 1-2-3 as an object lesson, to point out the corresponding fact of life that anyone trying to compete in the truly massive software market is going to write their code in assembly language or they are going to be rudely displaced by software vendors who do.
(We almost forgot: in July (IBM) Softalk, 1-2-3 was in first place with an index of 246.38, Microsoft's Flight Simulator (written in 8088 assembly, by the way) was in second place at 98.32 and Context/MBA was in 25th place at 12.87. Flight Simulator was outselling the 2nd place game, Frogger, by 5-1.)
As the folks at Context have discovered, transportable code such as MBA's Pascal hinders rather than helps sales. That contradicts the commonly accepted and prevailing theory. When facts contradict theories, it is time to reexamine the theories, no?
Two years ago the prevailing view was that the 68000 was exclusively useful as the engine of a complex device. We disagreed.
Eighteen months ago the prevailing view was that the Intel 432 was going to be a marvelously high-performance chip set. We disagreed.
A year ago the prevailing view was that the 68000 was such a fast device that operating systems could be written in interpreted p-code. We disagreed.
Six months ago the prevailing view was that UNIX was going to take over the 16-bit operating system market (some restricted that assertion to the 68000 side of the 16-bit market). We disagreed.
Today the prevailing view is that software must be transportable. We disagree.
Now, we are not always right and we FOR SURE do not print this newsletter ex-cathedra, but our batting average has been respectable. Following the crowd is the easy and expedient path, but sometimes one discovers oneself to be the millionth lemming to fall off the cliff.
Why, we even do ridiculous things such as assert that virtual memory is not a hot idea, which apparently
contradicts the opinion of almost all you readers. So LISA comes out as a virtual machine, which means the machine state must be maintained in RAM, which means LISA runs slowly.
LISA is a single-user device with a megabyte (!) of RAM, a 68000 and a 5 megabyte hard disk. There just isn't any excuse for that device not to perform like a bat out of Hell. (Or a hell-kite?) Instead it is slow enough that we get letters asking, "Are they sure there's a 68000 in that thing?" All because conventional wisdom says virtual memory is great.
We guess what we really need, therefore, is a 68000-based virtual UNIX system with the operating system written in interpreted p-code, floppy disks, a BASIC written in a sort of interpreted p-code and applications software written in transportable p-code Pascal. You know what? The Fortune 16/32 is close! As this is written, the $110 million stock offering placed by Fortune this spring is valued by the Street at under $60 million and a class-action lawsuit has been filed by an investor...
And Infoworld says Microsoft has sales of $50 million per year these days which means they are really going to have to scurry the second half to make that $3 BILLION in XENIX sales this year. Just how much of that $50 million do you think is in XENIX licenses?
Sometimes the prevailing view turns out to be right. When T.I. announced they were entering the personal computer market, a lot of people ran for the bomb shelters. The T.I. machine bombed instead. (Turns out the T.I. can only be sold for a lot less than it costs to take.) Then it was the Japanese who were going to take over. They were? And then IBM announced and everybody said IBM would sell a ton and guess what? Everybody was right! The prevailing view is not always wrong, but it has a poor batting average.
And, as we have pointed out, we are sometimes wrong. After trying for over two years, we still have not convinced any significant number of people that the 68000 is NOT exclusively suitable for incredibly complex systems. People who hear of us and our DTACK boards for the first time automatically enquire whether we provide UNIX, LISP and FORTRAN with our boards. And are VERY put-off to learn that we do not.
The only really successful 68000 systems to date are, in fact, the incredibly complex, expensive stuff - which, as it happens, generally means UNIX based machines! There are between five and ten companies that turn out maybe 50 systems a month each on average. One company claims that it has led this group of
companies for the last 18 months and brags in its ads that it has a total installed base of 1500 systems!
But this stuff is expensive! Although they would like to kid you with a price of $20,000, that is in fact a low-ball price which does not include everything you will need to get a system up and running. You will in fact discover that a single-user system from any of these companies will cost you about $35,000. The $20,000 figure is what another work-station costs after the initial system has been purchased.
AND, consider for a moment what the reaction of Uncle Jack or Uncle Clive, among others, would be to an assertion that a product which sells 50 a month is successful.
Well, folks, the price of the 8MHz 68000 has dropped to $42, quantity 100, in Motorola's published price list. And we understand the going price for production quantities is about $20. Which, as even the stupidest purchasing agent understands, makes the 68000 suitable for use in a small, simple system. We mean a system comparable to the Apple II, the CBM 8032 and the Eagle II we are typing this newsletter on. But no such system exists.
Perhaps the closest is our old friend, the 2-MIPS Sage II. As Infoworld pointed out in its recent review of the Sage, it is not useful for much other than as a software development tool. Part of the reason for this is that the Sage is a terminal-based system and has all the limitations of a terminal-based system, such as no graphics. And all of the software uses interpreted p-code and is therefore slower than the same program running in native code on a Z-80 machine. Z-80 machines are awfully cheap these days, and since the Sage runs over five thousand dollars with two decent floppies and the required Televideo 925 terminal, it is not effective on a cost-performance basis.
We have received numerous letters from readers pointing out that FORTH, for example, can in principle be written in assembly code rather than in FORTH. In practice, this is not done. A recent issue of the FIG newsletter carried a floating point package written, we kid you not, in FORTH! Talk about your basic slow code! And, in subsequent issues, that package was acclaimed by FIG newsletter readers! SLOTH IS THE INEVITABLE RESULT OF TRANSPORTABILITY!
We refer you to the very first issue of a new IBM-related magazine, the PC TECH Journal. Beginning on page 95 is an article on benchmarking which just happens to be our old friend the 'HAT' demo with an extra wrinkle at the top. The figure on page 97 will
look very familiar to owners of a DTACK board. You will keep in mind that we got that benchmark down to 10.8 seconds using the 12.5MHz 68000 and static RAM.
The time required to run that demo using single-precision floating point arithmetic with the IBM IV.03 p-System from Softech is 69:00. No, not 69 seconds, but 69 minutes! That's even slower than the original Applesoft version of the demo, but, remember that the IBM version does have that extra wrinkle. Then they ran the same Pascal code on a version of Softech's UCSD Pascal which has been fine-tuned by another software house and timed it at 21:37. That's 21 minutes, more than 3 times faster. How come?
If you will turn to the second column on page 116 you will discover that Softech wrote the p-codes for data type real (such as load, store, and the four math functions) in interpreted p-code! YE GODS! Talk about deliberately shooting yourself in the foot! This is exactly equivalent to what the FIG-Forth folks did recently. Now you know why those lower-end 68000 systems which have their operating systems written in Softech Pascal don't run very fast, 68000 or no 68000.
Oh, yes: they also tested a version of Pascal that uses the 8087 and native code. It ran in 3:35, or 215 seconds. That's 20 times slower than our best version of that demo, and about 17 times slower than the HALGOL version, D3.FAST. (Issue #7, Feb & Mar '82, p10. Time adjusted for 12.5MHz.) And, please note that the 8087/native code version does not meet the conventional test of transportablility.
(We had a vigorous conversation recently over the subject of Pascal transportability. The other fellow asserted that Pascal which compiles to machine, or native, code is best. We agree with this in theory, but in practice this can't be done because Pascal is so un-standardized. Remember, if it doesn't work the fact that the theory looks good is irrelevent!
As one result of the conversation just noted, we called the folks at Oregon Software for the second time, the first time being nearly a year ago. Their 68000-based native-code-generating PASCAL is now available and will set you back $7,000 for a single-CPU license including a program development helper package. Since it was a slow Friday afternoon, we spent some time chatting with a lady in the sales department about what we are doing and what they are doing.
They are obviously targeting their software package at those five to ten companies building about 50 very good 68000 machines a month each. So Oregon Software gains
about 250 to 500 potential new customers per month for their 68000 package. We do not know how many employees Oregon Software has but if they sell, say, 20 of their 68000 PASCAL packages a month they will generate an income from that one source of $1.7 million per year. Since they also sell a VAX version, that just might be close to what their sales really are for the 68000 package.
We pointed out the irony of the fact that we are offering a very fast 68000 board with a full megabyte of RAM for the Apple II computer, of which there are nearly a million in the field, but that we were not selling large numbers of boards because we have little software to offer. That $7,000 package would probably run very nicely on our half-megabyte board, which costs $1,245. It was apparent that the Oregon Software salesperson was aware of that fact. And also aware that you do not sell a $7,000 package to run on a $1,245 board.
Although we are swapping literature with the Oregon Software salesperson, there is no possibility that we can ever offer their software. And, is that not proof that we were WRONG when we asserted, two years ago, that the 68000 was also suitable for use in small, simple systems? You must spend $35,000 and up for one of those complex 68000 machines before you will be psychologically prepared to shell out $7000 for a good native code PASCAL compiler.
No solution to the 68000 low-end machine software dilemma exists AT THIS TIME that does not involve running at about the same speed as Applesoft unless you like FORTH or assembly language. And we assert that running about the same speed as Applesoft IS NO SOLUTION AT ALL! Although we would like to be able to offer the Softech package for our boards as A tool, we think that package is useless as THE tool. If you disagree, you can buy a Sage or maybe a SAYBROOK under-the-hood Apple-compatible 68000 board. (We can't offer the Softech package because it is MUCH too expensive for use as A tool.)
And you will recall that the industry-standard BASIC for the 68000 is badly busted because of two levels of interpretation. That's why you don't hear much about 68000 machines running Microsoft BASIC.
[LATER: an article has just been published in E.E.T. which asserts that Oregan's 68000-based Pascal runs $600 for just the compiler or $1650 for the compiler and the programming tools, and will be available in Sept. How this squares with the conversation reported above we dunno. Maybe $7000 is the single-piece end-user price and $600 is the OEM price?]
Do you realize that we have pointy-headed idiots in this industry who will risk a $100 million - plus manufacturing operation on the assumption that the 68000 is so powerful that you can write the operating system using highly inefficient code? That people who should know better have applauded that decision? That there are vested interests making damn certain that you cannot get a 68000 BASIC that runs at 60000 speeds? That there are to this day folks at Motorola who are deliberately doing everything they can to keep the price of 68000 software high? Do YOU understand any of this? If so, would you please write to us and explain it?
(Actually, sabotaging the performance of your 68000 BASIC makes the best - or worst - sort of sense if you have $50 million annual income from 8-bit software to protect.)
We can only see three possible solutions to the problem of 68000 software: 1) Give up and join the 8088 camp. 2) Get accustomed to running our 68000s at Applesoft speeds. 3) Go it alone.
We have decided to go it alone. Now you know why we have a full-time programmer whose sole responsibility is to implement HALGOL, and another hardware type who occasionally dabbles a bit with that language. Real progress is being made, but it is not the kind of stuff you can hold up and point to, like a math package. What we are doing now is developing the parsing basics.
Let us tell you, it is lonesome here with our fanny out in the cold breeze! But when the world's absolute greatest single-user single-tasking microcomputer (the 68020) becomes available over the counter two years from now, we will ALREADY have a language and an operating system - a SIMPLE operating system - available for that humongous chip!
And unless we make a great big mistake someplace, our language will outrun anything comparable because we are the only people we know writing a language for the 68000 that is specifically intended to be fast and efficient rather than transportable! (Phase Zero's 68000 BASIC is fairly close to our philosophy, we are told.)
And right now, we are making two different boards using today's absolute greatest single-user single-tasking microprocessor that can be bought across the counter. And that statement would be true even if the 8 or 10 MHz 286 were available across the counter this year.
But we sure wish we had an efficient language and a simple operating system for the 68000 RIGHT NOW!
by Steve McIntosh
Long Beach, CA
Back in the dark ages at the dawn of the computer age, Way Back in the sixties ANY computer was expensive. Everybody wanted one, but few could afford the luxury of spending a large percentage of their budget on projects that could be done cheaper (if much slower) by a few junior executives. Things that HAD to be done fast, such as payroll and market tracking, got the money to place them on computer and everything else got done manually or it was put on tab equipment.
As with anything expensive, there are ways to share the cost. Timesharing became popular. The words "multitasking" and "multiuser" were heard by the common user and the resources (and expenses) of a computer could be shared by all. It still cost, but not as much.
UNIX is without a doubt the finest multi-user operating system around. It, and its companion laguage "C" give the user power over the computer that he/she/it is connected to. (It also gives the user enough rope to hang himself and the system - but save that for later.)
A QUICK ROMP THROUGH THE HIGH SPOTS
I am going to quickly hit on the main features of UNIX in no particular order, and take a few stabs at pet peeves.
The UNIX system reeks with evidence of its origin as a time-share system to be used by computer professionals. The entire system is designed to let a knowledgable user (and they do exist) get on, fire off a burst at the keyboard and get off in the least amount of connect time, leaving tasks that can run on happily for hours (or days (or years)). In order to do this the command language has become very terse and quite unreadable by any that are not WELL versed with the system. This has carried over into the primary language in UNIX, which is C.
The commands that are available in UNIX have been described as "wild conventions flying in loose formation". Each command (read utility program) seems to have been written by a different committee. There IS NO standard command format, just poorly followed conventions. Each command has to be learned and remembered on a one-by-one basis. A new user has to keep the reference manual (all 600 or so pages of it) nearby at all times least he do something dangerous by typing in the wrong letter at some point.
In order to help (?) the user, the Command Shell was invented. This is simply a pre-processor for commands. It allows the user to type in even less and do much more. It will perform tasks such as remembering the last 20 lines you typed in or going out on a search-and-destroy mission against wayward files. The shells of course follow the precedents set by commands and are terse and cryptic to the uninitiated. They also follow the precedent of lack of consistency. The two most popular shells are the Bourne shell (the B shell) and the shell written in C, named appropriately enough the C shell. (OK everybody can now make the joke about selling C shells by the C shore and get it out of your system.)
The shells, of course, do much the same thing and as you might expect in just slightly different ways. An example is that where the B shell would use the notation 2>> the C shell recognizes >>! as the command to place error messages (if any) at the tail end of an existing file. Don't be discouraged by the fact that one shell can do things the other cannot. They can call each other so that you can use both at once! Of course you have to talk to each in its native dialect. And even more, if you don't see what you want you (or a high priced consultant) can go in and build a shell of your own so that you can talk your own dialect.
The best feature, in my mind, is the way the files are implemented in UNIX. A successful multi-user system has to be able to keep the files of different users separate, and still enable all users access to system files such as commands. The heirarchical tree structure of UNIX when coupled with the access permissions does this very nicely. When you get down to the bits and bytes of it, with INODES and directories you will discover a simple elegance which is quite refreshing after wading through the confusion of the rest of the system.
Since UNIX is a multi-tasking system with many users, the operating kernal needs to have enough smarts to keep things (and users) safe from each other. It is a sad fact of electronic life, that the smarter you make a program the slower it is. In the case of the kernal make that the slower the whole system is. UNIX has struck upon an unhappy medium which assumes intelligent and well intentioned users that will not do something stupid that the kernal would otherwise be preventing. It is very easy to degrade the entire system by doing something stupid by accident (or by malice). Files, however seem to be fairly safe from anyone but their owners and the system administrator.
Most of the commands are designed for the manipulation of character data in text files. This also is a hang-over from the timesharing days when all you had in a remote location was a character oriented terminal of
one form or another that had no use for any other form of data. UNIX lets you re-direct character input and output so that a program can take data from a file instead of the keyboard and put data into a file or "pipe" it to another command as input. These are very powerful features.
Some people have gone as far as to say that the "pipeline" concept is what makes UNIX special. You can take the results from one command and pass it as input to the next command and so on down the line for who knows how many commands and direct the output of the last command into a file, or out to your terminal. This is an absolutely wonderful tool, but there are a few kinks in the pipe.
Pipes in UNIX are implemented by starting up a task for each command in the pipeline. Each of these tasks are competing with each other and the system for run time. They cost the system a small amount of overhead in time and memory, but it can all add up. In addition each pipe link between commands is set up as a first-in-first-out line buffer. These FIFO buffers also take up system resources. Pipes are restricted to taking output from one command's "standard output" into the next command's "standard input". They cannot be used for passing binary data or records or anything like that. Only commands that use standard input and output can be used with pipes. (You thought all commands used a standard? They don't.) Pipes are a great idea that has been crippled by implementation. It's probably just as well that they are restricted in use or there would be even more ways to make mistakes.
You would think that with all this emphasis an text files, there would be a huge amount of effort devoted to creating text editors that are both powerful and easy to use? If so, you are half right. There has been a great deal of effort expended. The editors are extremely powerful. Don't try to learn how to use them on any data you want to keep.
There is a Very Powerful editor named, appropriately, ED. If you know how, this editor can with a few cryptic incantations do most anything you would ever want to do a text file. (Of course, if you don't know what you are doing... ). Do not expect to become an ED expert with any less than a month of regular use. Most importantly, ED presents the concept of "regular expressions". This means that you can define a very complex pattern matching "word" which can then be used to find and modify text. Many commands use this regular expression concept (as do the shells for filename generation, command construction and so on. Unfortunately the exact implementation of this concept differs from command to command and shell to shell, but only slightly enough to confuse non-experts. (Keep that manual very handy, folks).
There are quite a few other editors out and around the UNIX world. They are all expensive, and all have their own unique quirks and syntax.
Commands need not be written in anything as gross as a high level language. Most application commands seem to be written in commands. The shells have the capability of interpreting keywords to set up control structures such as WHILE and FOREACH and GOTO and to create and maintain variables such as file names and flags and counts. Of course the operators are commands and the operands can be generated by a whole slew of different mechanisms. These can all be passed to sub-shells (of any flavor) that the commands generate.
This makes it possible to generate VERY POWERFUL commands with completely arbitrary names and arguments. If these new commands are also useful, they get passed around and become a part of the system, arbitrary arguments and all. The manual grows a bit larger, and new users grow older a bit faster. This also has the side effect of placing an ever growing interpretive overhead on the system. There is really nothing to keep a poorly designed shell procedure from thrashing itself to death (and perhaps taking the system with it).
I am now about to alienate a good many people by comparing the UNIX system and its primary language C to another system/language, FORTH. In fact, I as going to call UNIX "the rich man's FORTH". Now before you reach for the phone to call your hit-man, hear me out.
FORTH and UNIX have three main points of similarity. (I am including C as part of UNIX as shell programs can easily be mistaken for C by the casual user.)
First, UNIX and FORTH share a common philosophy - never re-invent the wheel. Both systems rely heavily on previous work, permitting the system to grow in capability along the lines that the user wants, and to the user's tastes. This is a two edged sword that has people swearing by (and at) both systems.
Second, both systems require a fairly large start-up effort. They are both rather cryptic in implementation. (In fact the next time a UNIX person points to FORTH and decries it as obscure I may be moved to violence.) Both systems have the capability to be made consistent and readable. Users of either system rarely use this capability. Both have "standard" kernals of commands (words in FORTH) that are the building blocks for bigger and better commands.
Third, both systems can be used in what amounts to an "interpretive" mode where individual commands can be
invoked or strung together without first having to write a program, compile it, etc. You have the capability to write programs if you want on either system.
The main (conceptual) difference between UNIX and FORTH is that UNIX was designed and implemented to be multi-tasking and support many users. FORTH was designed for single task single user applications. The trend of both of these is that people want single-user UNIX systems ("Keep your hands off of MY cpu cycles!") and FORTH people want multi-tasking and more pre-built words. UNIX is for Big Fast and Expensive systems. FORTH is for small fast cheap systems.
(NOW you can throw rocks at me.)
With all of this bad-mouthing of UNIX I have been doing, you may ask yourself "If he's right, then why is it The most popular system, etc."
Well, in the beginning, it was one of the only games in town. It has a really HUGE amount of programming already installed. You have a Very Good chance of finding the program you need already written and lurking out on disk waiting to be discovered, and a really HUGE manual to help you figure out how to use it. Any system that distributes about a hundred games as part of the system and documents them in the manual can't be all bad.
UNIX is without any doubt Very Powerful once you become an expert. And after all, the experts can set up canned applications for the Great Unwashed, if you let the unwashed use your system.
You must invest such a great amount of time and effort to become and remain an Expert user that an emotional addiction sets in. You find yourself saying things such as "It's the greatest operating system ever! Why else would I waste time learning all the little ins and outs of all the commands?". This especially applies to the people who rise to the station of the System Administrator. (A good guess at the S/A's password for the root login is "god".)
If you set down with a prospective user who has been enamoured by the prospect of having UNIX, you will soon find out that the lure of those thousands of application programs and commands will overcome the fear of becoming enmeshed in the system. It can become a love/hate relationship real fast.
UNIX has a high-level assembly language called C which permits the user access to any and all of the system facilities. It is assumed that there is enough
hardware protection that a crazed C programmer will not be able to take anybody else with him when he goes.
As far as I can tell (and many others share my opinion) raw UNIX should never be served to a user. This is a computer expert's system, it was written, developed and used almost exclusively by experts. It is user hostile. Care MUST be taken to shelter the user from the system and protect the system from the user. The non-expert user is not to be given the Power of the system without risk to the system or a staggering amount of development cost to design and implement a smart applications environment for him to use.
Systems people are given the full measure of the Power. They are assumed to know what they are doing. But what do they have to do?? The bottom line is that the system is there for the users - people who have something important to do - Payroll, Finance, inventory and such. This means that the people with the most power have the least to do with it except carefully dole it out to the people who need it the most by writing applications.
For many applications, it can be easily shown that a close fit to a solution can be purchased, off the shelf, hardware and all from your friendly local computer store at a cost that is less than what would be required to develop it on an existing UNIX system.
The UNIX system, as it stands, was "developed" by a committee of committees and has all the problems and snafus that are implied. If I hadn't HAD to learn it as part of my job, I doubt I would have bothered because it does not meet my needs. I doubt that it meets the needs of many people who have been told otherwise, and I suggest that someone who sells a real end-user a UNIX system be ready to duck when he returns it.
UNIX has many good features that need to be isolated out and refined into something that is consistent as well as powerful. Anyone who tells a user that you have to take the bad along with the good is only asking for trouble and a loss of sales.
and have some phone conversations, too. A recent letter commented about a particular magazine: "(it) hasn't yet slid into the 'How to Get the Most For Your Adventure Game Dollar' trap, though it may be headed there." We had spoken earlier with another subscriber regarding the peculiar mathematics of the personal
computer marketplace, in that half of the existing personal computer owners purchased their computers within the last eight months. That (we asserted) was true last year, today, and probably two years from now.
We pointed to the vast success COMPUTE! was having going unabashedly after this beginner's market and asked what the implications were for any periodical publisher. "It means you publish something called 'The Journal of Simple 68000 Systems'!" was the reply.
In all seriousness, will there be ANY magazines left for us folks who already know what a RETURN key is? Is there a personal computer magazine out there that has NOT started a "beginner's section"?
"In response to William W's query about LISP... there is a PASCAL program which implements LISP. The title is "A Portable LISP Interpreter", the author(s) are L. A. Cox et al. This is available from the National Technical Information Service for $7. [Call (703) 487-4650 and give them your credit card # - FNE]
"As for books, William M. White's "Implementing Software for Non-Numeric Applications" (Prentis-Hall) is certainly worth looking at because of its blood and guts approach to the booting up of list processing languages.
"Just a comment about microcomputer FORTRANs. Microsoft's F80 compiler for the 8080 was a real gem. The code runs in 27 Kbytes and directly generates machine code with no intermediate disk files being written. The language allows a rational approach to strings and is simple and fast. The version of the language for the 8086 is a far different beast. The newly released version will not run in less than 160 Kbytes of memory and generates two intermediate files during compile and yet another during link. Yep, you guessed it, this version of FORTRAN was written in PASCAL. Well, at least it allows an ELSE clause after the IF." Murray S. Dallas TX
But it's transportable, Murray! - FNE
"You ain't going to have many buyers of a FP ARRAY unless there is at least a FORTRAN that can use it (if you are really greedy, ADA will run on such a design). But who besides three freaks in the world will program such a design at the ASSEMBLY level - all those guys into VAXes?)
"Someone else seems to believe in 0 interpretive overhead languages but calls it a REAL TIME COMPILER SYSTEM - and is trying to patent the design (on 8080/Z80 though) - ref MICROCOMPUTING Jul 83, pp52-63." Bob P. St. Louis MO
Bob, both FORTRAN and letters to the editor are supposed to have equal numbers of left and right parentheses. As you read this, the number of 68000 boards we have sold without FORTRAN will be 390 +/- 10, about 23 of them for the PET. Maybe there are more than three freaks in the world?
You don't suppose the guy trying to patent our very own original HALGOL idea would like a copy of "INCREMENTAL COMPILERS (a tutorial)" by Wesley J. Rishel, DATAMATION Jan 70, pp129-136, do you? (Yes, 1970!)
Here is the second paragraph: "The defining characteristic of the incremental processor is that each line of source code is translated at the time it is received from the on-line terminal. This allows immediate reporting of syntax errors. Since each line is compiled as it is entered, the whole program need not be recompiled and rerun to change only a few lines. Incremental processors always translate directly to core [RAM to you newchums - FNE], eliminating the loading pass. As a result the debug cycle consists only of 'type in statement; begin trial run.'"
And the last sentence of the first paragraph: "Since incremental compilers have a reputation for generating slow code, (this) article concludes with some arguments for attacking that problem."
Dr. John K (Lewiston, ID) has done it to us again. (See issue #10, the end of col 1, p.16) It seems we inadvertently combined two different Sherlock Holmes stories. The dog (NOT hound) which, strangely, had NOT barked in the night is from the story "SILVER BLAZE." The "HOUND (not hounds) OF THE BASKERVILLES (not Baskerville)" features a particular hound with a disposition midway between that of Kindly Uncle Jack (Trameil) and your FNE.
He also has an even lower opinion of the Intel line of microprocessors than we do, something which we had not previously believed was possible. John correctly asserts that the success of the 8088 in the personal computer marketplace is the result of marketing flim-flam, but apparently does not realize that the failure of the 68000 in that same marketplace is ALSO the deliberate result of marketing flim-flam - this time on the part of the Motorola marketing types.
"In a recent issue, you mention a screen editor that keeps track of which display lines are beginning-of-logical lines and which are continuation lines. This sounds very such like the way the CBM 40 col screens are used to handle 40 or 80 col lines. ...About the speed of the CBM disk vs. Apple: I think that the slowness of the CBM disk is due to the software IEEE
handshake within the computer." Charles McC, Minneapolis MN
Charles, you are one of several persons to correct us regarding the way the 40 column screen editor works. We only have an 80-col 8032 and thought it was typical. If they had the technique worked out, why the 80 column limitation on both 40 column and 80 column units? The size of the line buffer?
We are pretty confident in asserting that the IEEE handshake is not the speed limitation on the CBM disk drives. We have carefully analyzed the software, using a disassembler which also prints out clock cycles, and we have monitored the instrument with a scope while in operation. The IEEE software is capable of handling at least 10,000 bytes per second, about seven times the actual disk read/write rate. Maybe the 6504 which directly drives the disk itself has some slow formatting and error checking to do?
"...The Dtack hardware is intriguing, to say the least, but I fail to grasp something. I don't believe anyone is going to pay a thousand bucks or so for a 68000 board that simply tells secrets (rapidly) to itself. How do the Dtack units communicate with the outside world?
"The HALGOL project is interesting, though I don't see why anyone would want to pattern a language after BASIC. In my humble opinion, modern ideas about structured programming are not merely ivory-tower academicisms." Jim H, Pullman WA
Jim, the 68000 communicates with the Apple as an I/O device at about 71,000 bytes per second without the Stuffer board or about ten times that rate for block transfers with the Stuffer board. What happens after the info is in the Apple depends on what equipment you have bolted onto your Apple besides the Dtack board.
The Dtack board is not well suited (to put it kindly) to applications where the Apple is probably already I/O bound, such as a data base management program running in DOS 3.3. But there are a large number of computation-intensive applications where our board works very well indeed even though limited to conventional Apple I/O.
You have mentioned the Vectrix machine. That uses an 8088 to drive a 7220 which drives a 640 X 480 color monitor. For about the same price we will be able to offer, soon, a 68000 driving a 7220 writing to a 1024 X 796 color monitor, with a megabyte of program and data DRAM for the 68000 tossed in gratis...
About HALGOL: we are patterning the language more on
the BASIC programming 'environment' than on the BASIC language itself. There is much to be said for structured programming and some to be said against it; see our reference to Mark Dahmke's book elsewhere in this issue. Also, surely it is evident that a language which supports subroutine calls by name is well suited to structured programming? You can, if you wish, overlook the fact that we are including the possibility of using a (gasp!) GOTO.
Oh, yes: our secretary, who is in charge of all checks around here, has signed you up through issue #38. And we probably will need a new Eagle keyboard by then.
"Regarding your 'deploring' me for not photocopying your newsletter for my 'cohort(s)': I happen to work for this small, Fortune 20(00) outfit whose management is very sensitive about being sued. Said management has pointedly and specifically instructed me not to violate the U.S. copyright laws... in some quarters those laws are almost as highly regarded as those assinine export restriction laws of which you are so chary." Terry P, El Cerrito CA
Terry, we were going to point out that the correct spelling is "asinine" but on second thought we like your spelling better. Ned tells us he made twelve, not ten, photocopies of the last issue and happened to spot that item while making those copies.
As to why we retain a copyright notice: see issue #5, page 4, the first four paragraphs.
Our thanks to the two readers (so far) who have sent us (doubtless illegal) photocopies of "REAL PROGRAMMERS DON'T USE PASCAL" from July '83 DATAMATION magazine. We will have to see if we can ferret out our eleven-year-old copy of DATAMATION's very straight and scholarly article an the use of 'COME FROM' to compensate for the absence of the GOTO statement in modern structured programming. We can still remember some programmers trying to figure out whether the article was on the level...
"...Microsoft may be planning an 80286 piggyback board for the IBM's 8088 chip location that may 'transform the PC into a multiuser - XENIX super - PC.' If they do that, and you have not come out with something (68000 or whatever) before they do, we will buy one Microsoft (board) to speed up our single-user PC when theirs comes of age.
"I remember you wrote in a past newsletter issue that most IBMers were first time purchasers, and that they do not realize that their device is slow. Most of the
local owners I talk with know that they want any computer to be faster than it presently is. Most of them have had exposure to superminis and mainframes so they have a passing association with fast devices.
"I think that a market window is open for compatible speed-up boards in the IBM area." Sam B. Pittsburg PA
Sam, we think the Microsoft idea is a good one, even though the 286 will be severely limited by the PC's 8-bit bus bandwidth and its slow clock (4.77MHz), resulting in a bus bandwidth of only 1.2 Mbyte/sec. You will recall that Creative Computing found that the best Apple performance was obtained using a combination of the Number Nine 4MHz 6502 speedup device AND our DTACK board.
Rather than looking at our board and Microsoft's as an either-or situation, try looking at BOTH!
We agree that there is a market window for a 68000 add-on board for use with the IBM PC. Just as soon as we have a language - either the Phase Zero BASIC or our own HALGOL - we are going to do the five minutes' design work required to modify that small interface board to work with the PC. (See issue #6, page 1.)
Note that we will shortly have, an the hardware side, both static and dynamic RAM versions of our board and a very-damn-high-res graphics board that will work with either. And that we will add a floating point 'array' processor that will work with either board soon ...
We will overlook that multi-user bit this time, Sam...
"I resisted giving you advice on the 68008 under the bonnet design because I really don't think I'd buy one. The multiprocessor board is another story. For one of those I'd do anything short of reenlistment." Pete S. Palos Verdes CA
We'll drink to not reenlisting, Pete!
"I understand that I cannot expect you to send unsolicited material such as the diskette back to me at your expense, so I am including $6 for return postage. Mackintosh... Could it be that the end of the year finally sees the intrusion of low cost high performance computers whose power goes beyond the IBM-PC variety? Do I get it right that there are no TECHNICAL reasons that such computers do not exist today except at horrible prices? Are there any sound reasons that a low-cost system should NOT have similar capabilities as those mini-computer clones?" Bernhard M. Kandel W. Germany
Bernhard, we are returning your $6 plus an additional $6 to cover your expenses for the disk you sent us. (True, Digital Acoustics would want $10 for the same service, but Digital Acoustics pays taxes and other overhead, and you - presumably copying at home in the evening - don't.) We will, as you suggest, forward the small Lisp interpreter to William (or Bill).
Let us apologize for doubting that Pascal could load a disk in 10 seconds - your demo on the other side of the disk certainly proved your point. But some folks who should know have told us that Apple Pascal uses a 2-1 sector interleave, which is why we thought it would take about 18 seconds to load a full disk under Pascal. Question: does your demo use STANDARD Pascal, or a special jiggered-up machine language utility?
The short quotation above from your 1 1/2 page letter suggests that you are reacting to our reply to Oliver B's letter in issue #22. You should know that Oliver and us are old, and we think friendly, antagonists via correspondence. For instance, Oliver has in his possession a very recent offer to permit him to manufacture the static RAM DTACK board in W. Germany under an arrangement which would require no money whatever to change hands (he turned this offer down).
The minicomputer clones generally have graphics capability and disk storage which cannot be obtained cheaply, and most of them have lots of memory, like a megabyte up. That isn't cheap either. You might want to review the very last paragraph of issue #10. (If you do that, jack the price up $199 and substitute floppy disk for audio tape; times have changed.)
Two other obvious problems: 1) There is no cheap 68000 software to be found anywhere. 2) A manufacturer with a moderate-performance $10,000 machine might want his cheaper machine using the same CPU to have even lower performance...
"My No. 1 son is earning his bread this summer doing FORTH program maintenance on a system made up of three coprocessing 68000's. I understand that the firm he's working for is aiming for EIGHT co-processing 68000's. I don't know whether it's true co-processing, nor do I know other details..." Charles McC Minneapolis MN
Somebody stealing our ideas again, Charles? (Just kidding.)
"You just sent me an empty 9 X 11 envelope... I suspect it was supposed to contain your newsletter, since the 54 cents postage seems a bit much for an empty envelope." Eric K Ann Arbor MI
Eric, the phantom newsletter thief has struck again. I refer to the one who swiped REDLANDS from Bob P's issue #21, leaving the rest of the issue behind.
"I believe I (as a subscriber) am one of those unwashed fruity types so frequently mentioned in your journal. What I am looking for is (laundry list of applications software) and perhaps a rock throwing contest when I feel like it, and you're probably gonna get nominated." John L Mt. Pleasant MI
Boy is that letter overdue! John, none of the other 399 subscribers have had the nerve to write that... and as a paid-up subscriber, you are but definitely counted as one of the washed - FNE.
"Thank you for mentioning my all-time favorite hero, Charles Proteus Steinmetz... On page 19 of issue #22 I believe I found an error in your hexadecimal table representation. It seems to me that the 6th byte should be '05' rather than '03'.
"In the old days, Digital Acoustics offered a big, fairly expensive 68000 board and talked about a cheaper, more limited 68008 board... But Digital Acoustics, purveyor of simple systems, is not content to limit my choices to 2 items. I'm now offered Stuffer boards, a graphics board whose resolution would bring real tears to my eyes, and a QD-1 which offers all kinds of goodies including a port for my Epson. ...It appears that the 68008 board wouldn't be expandable [correct - FNE] to allow me to play with all of the new goodies which you are offering. Your expanding product line is leading me to consider expandability over price considerations. Volkswagons were simpler to choose when they only offered one color. Thanks for complicating my life." Pierre M Mt. Horeb WI
You are right, Pierre, it should have been '05'. A couple of other readers caught that also. And the IBS ad in Aug BYTE leads us to believe that a 68008 board would have to be VERY cheap (by having very little memory) and be non-expandable, of course. Truthfully, the under-the-hood Apple-68000 marketplace seems to be well served without intervention on our part.
And, while we also respect Steinmetz' considerable accomplishments, we were referring to the fact that had the unit of frequency been named after Steinmetz by using his initials, it would still be cps! (Or maybe CPS?)
"I believe that you were incorrect when you stated that Kernighan co-authored the C language. He did, in fact,
co-author the fine book "The C Programming Language" along with the language's author, Dennis Ritchie.
"From what I understand, Motorola is responsible for witing BASIC-09 which, in my very limited experience with it, appeared to be the best BASIC I've seen [you ain't seen Wang's BASIC-2 - FNE]. The GOTOs, etc. were semi-compiled at edit time; it has structured commands; allows parameter passing on subroutine calls and excellent debugging facilities. It even has a compiler feature in addition to the interpreted mode. OS9 was developed by Motorola and Microware as an environment for this language. I already know the answer, but I've got to ask anyway: If Motorola would spend the money and effort on a fine but last-8-bit-gasp micro, why can't they do the same for the 68K, which will prove to be a such more long term and profit-making venture? Arrgh!" Nigel R Torrance CA
You are correct about Kernighan, who also co-authored "Softmare Tools" with P. J. Plauger. But Nigel: "Arrgh!" is correctly spelled "Aarrgh!", or, under extreme provocation, "AAAARRGH!". Please take note of this for your future correspondence - FNE.
"I rashly assumed that some member of the Suite F staff might hit upon the idea of offering a printed text, topic being the language, system support software, and hardware features of the maybe soon-to-be DTACK system, said text being available for purchase...
"Ever since I purchased my first VIC - a machine that was deliberately undocumented - and only discovered what I had after buying the Programmer's Reference Guide, I have been quite wary of computers or computer-like processors [like the ENS board you bought? - FNE] that are not documented in a publicly-available form (book, etc.). Your careful and scrupulous avoidance of printing details about DTACK and your piecemeal documentation of developing software, via REDLANDS, may be fine for the classroom but it may not be the best way to sell a product to an impatient, inexpert and somewhat fickle public.
"I hereby resign myself to the DTACK WATCHERS ANONYMOUS organization." (signed) Faithless Correspondent Nils D Wethersfield CT
Sigh. Nils, having in his possession about 470 pages of this newsletter, asserts that we are secretive about our products. He suggests that we assign a member of our staff to produce a book which will accurately document our products and software. Accurately document WHEN, Nils? Early last summer when we had no graphics board or one megabyte Grande? Three months from now when we will have a case, crude HALGOL and a
QD-1W math/IO board? January next year when we will assuredly have a less-crude HALGOL and an 'array processor'? Perhaps we should turn out a new book every three months, each of which you will purchase?
Now, about that member of our staff: Who? Digital Acoustics has one yo-yo who signs checks, watches cash flow, decides what new products to develop and who also writes a newsletter. It has a young project engineer who has about a semester's worth of classes remaining before he gets his E.E. degree and a young programmer who just graduated from UCI with a degree in computer science. We have a secretary who works three days a week. Death March Dunkersan runs production, with an assist from a part-time college student whose major is in business administration and who has been with us about three years, beginning when he was a sixteen year-old high school student. We have a nearly-retired technician who works (?) four hours a day and who has the calm temperament needed to talk to ignorant (NOT stupid; there's a difference!) persons on the phone.
We have expanded our engineering and production capacity recently by having outside vendors (consultants?) lay out our circuit boards (to the tune of about $12,000 in recent months) and by having outside vendors stuff and wave-solder our circuit boards. The part-time college student has, accordingly, been understudying our technician recently.
That's it, Nils. That's the complete Digital Acoustics organization, unless you want to include the accountant who does our books once a month. Which staff member do we assign to produce a complete book (by next week, maybe?) describing our products and our software?
To the rest of you readers: Nils has "resigned" more times than Sarah Bernhardt had farewell tours. He will stick around because this newsletter continues, after more than two years, to be the only game in town.
A small organization? Sure. Profitable? Well, we added the new project enginner and the new programmer as part-time employees last Nov and Jan, respectively and both became full-time employees (with full-time salaries) last June. You will note that the pace of hardware development has accelerated in that short period, but you will probably not have noted that the pace of software development has accelerated also because the programmer has one job and one job only: to bring up HALGOL. And he has been developing the needed 'infrastructure' the Apple/DTACK system needs to do the job. What's that? You say we are side-stepping the issue of profitability?
O.K. In our last complete fiscal year we turned an after-tax profit on sales of 9.4%, very close to the
10% we'd like. Up until now, the most the company has shipped in one month is about $67,000. Today is Sunday Aug 7, and August is traditionally the slowest month of the year for us in sales. It may therefore interest you that we will ship over $92,000 NEXT WEEK, the next five days.
Which is why we are very confident that we will have the needed resources to turn the 'array processor' into reality and maybe even HALGOL (software is always more iffy than hardware).
Oh, yes: in our last newsletter we reported that we had only one order for the Grande. That was true at the time we wrote it (ALL of the stuff we write is, to the best of our knowledge, true at the time we write it), but did you notice that it was not until the last newsletter that we told anybody the board WORKED? And the orders are now coming in as expected, considering it is now August.
And we, along with the rest of Digital's staff, have done a heck of a lot of work to get ready to ship that $92,000, a fact which is not unrelated to why we don't have a case and power supply ready already. John K please note ...
Now, if we may address Nils again: just how does one "resign to" an organization, Nils?
"The GOTO instruction in our PASCAL p-Code interpreter looks like this:
GOTO ADDA.L (A6),A6 NEXT
"Where NEXT is a macro containing the interpretive overhead
NEXT MOVEA.W (A6)+,A3 JMP (A3)
"All kinds of loops are compiled to p-Code using this GOTO instruction. To keep the overhead small, all so-called p-Codes are just the addresses of the related routines in the interpreter, sometimes followed by some parameters. The generated code is still relocatable, since the compiler generates relative branch addresses. Oh, yes: unlike UCSD, we use long addresses..
"By the way, R. Usor told me that the UCSD interpretive overhead consists of about 8-10 instructions (68000). Even better regards," Oliver B Hamburg W. Germany
When will you have your Apple-compatible PASCAL interpreter finished, Oliver? We'll pubicize it here - ONCE - FNE,
It has become apparent that the deplorable tastes of the American reading public require that a successful novel include spies, intrigue, missiles and stuff like that. Since none of the above were included in our previous effort to outline a novel, we are turning over a new word-processor (not leaf - get it?) and present, herewith, the outline of our new novel. Which is, by the way, a comedy.
THE RUSSIAN SPY: This one is nominally a military attache with the Russian Los Angeles consulate. Being located in Southern California, he does not have a single trenchcoat to his name. He does have the traditional stainless steel tooth, however. He needs a good Russian name, so we will call him Boris Karloff.
THE MISSILE: This one is an intermediate range intercontinental ballistic missile with a single nuclear warhead. It is designed to sit in a hole in the ground (sort of an up-side down silo) as protection against attack. Since the missiles themselves are logical military targets, they are located far from any other logical target. Since they have an intermediate range, they are located in the part of the U.S. closest to Russian targets, meaning Montana (check a globe, not a traditional 2-dimensional map). We will call this missile a Secondperson.
THE MILITARY BRASS: We will need a project manager of the guidance package for the Secondperson missile: (bird) Colonel Crudcutter. And the project manager for the entire Secondperson missile program: Major General Rattlesword. Maybe we should also allow for a bright young Lieutenant, so we can have a military type with technical smarts. He can be anonymous.
THE YOUNG GENIUS: A rising star at the gigantic aerospace firm Amalgametics, located in Anaheim, CA. We will call his Adolf Dorfman.
THE ELECTRONICS (semi) EXPERT: We can use an electrical engineer as comic relief and spear-carrier. We will call him Felgercarb N. Eloi.
THE TIME: To avoid problems with the Pentagon (witness the current flap over WarGames) we will place this story safely in the past, with the principal action taking place between 1962 and 1966.
It is a lazy, drizzling Sunday morning, Oct. 6, 1957. Our electronics semi-expert is listening to a steady "beep.. beep.. beep.." arriving over short wave via a
BC-348 radio receiver. Felgercarb is beginning the last third of his 30 month vacation as a guest of his Uncle Sam at Nouasseur AFB, Morrocco, about twelve miles southeast of Casablanca. Since no C-124 or C-131 cargo planes are scheduled that morning, there isn't much to do in the electronics maintenance quonset hut but listen to the re-broadcast of the day-and-a-half-old Russian satellite's signal.
That morning there were very, very few people who knew that the entire 44-foot Russian launch vehicle had also entered orbit. Krushchev did; he hinted at that fact later in ridiculing Vanguard as a "pamplemouse" or grapefruit. Our military certainly knew because they had some excellent "photographs" of the launch vehicle taken via synthetic-aperture radar!
There were also only a few people who knew that there were some Russian generals worrying because they did not know where the San Francisco Naval Shipyards were, and some American generals who were worrying because they did not know where Vladivastok's submarine pens were. If your missile warhead has a three-mile kill radius, knowing the location of your target with respect to your launch pad to an accuracy of 6 miles does not cut it. Think about it.
The time is now early 1962. Boris, our Russian spy, is sitting at his desk in the Russian embassy busily spying. On his desk are the tools of his trade: that day's L.A. Times, the Orange County Register, the Wall Street Journal and the Commerce Business Daily. Boris neatly files the clippings indicating that an award of $XX million has been made to Amalgametics for production of the guidance package for the Secondperson missile.
He shakes his head. In Russia this would be a secret! The crazy Americans are broadcasting the news everywhere, including press releases by the company, proclamations by politicians bragging about delivering public dollars to their constituents, and full-page help-wanted ads looking for engineers to fill out the program!
Enough such clippings give a very accurate idea of the status and size of America's missile efforts. Boris and his scissors keep very busy. Trenchcoat? Who needs trenchcoats?
There are no Russian or American generals worrying about the locations of their major targets, since lots of Russian optics and lots of American optics have been orbited in the preceding five years. The problem now is to build very, very accurate guidance packages ...
Adolf Dorfman is assigned by Acalgametics to develop prototypes of the test equipment needed to check out the central item in the missile guidance package, the gyroscope. Commonly called a 'wheel' in the guidance industry, this is nothing but a metal ball spinning on low-friction bearings. How low-friction? The bearings are gas! The metal ball has little vanes, like tiny propellers, to suspend the metal ball in the 'air', which is really inert nitrogen.
The idea is, if the metal ball spins at an absolutely constant rate and with minimum friction, it makes an absolute inertial reference good enough to hit a target anywhere on earth very accurately. The American generals have therefore placed stringent specifications on the performance of this critical item, specifications that could be met only with extreme care in production followed by meticulous testing. This means it is expensive to build these things. Why do you think $XX million was awarded to Amalgametics to build gyroscopes, hmmm?
The problem is, given the presence of impurities inside the hermetically-sealed gyroscope the frictional component would not be absolutely constant, and the gyroscope would "wander", rendering it inadequately stable for its intended use. There were many other steps in the manufacture of the device which could cause the gyroscope to be unstable if not performed perfectly.
Adolf Dorfman had a brilliant idea: he placed an optical sensor on the inside of the gyroscope case, and a mark on the side of the rotating wheel itself. By comparing the output of the optical sensor with a crystal-controlled clock, he could tell whether the gyro rotation was slowing down or speeding up. If this information was used (automatically, of course) to appropriately adjust the operating frequency of the power supply driving the gyro, the wheel could be made to turn at a constant rate. This is called, in electronics, a "phase-locked loop".
This test procedure eliminated the first-order component of the "wander" due to non-constant frictional forces, leaving only the smaller second-order components. And this made it much simpler to produce test data showing that the gyro package met the drift standards imposed by the military production contract.
Adolf's engineering group successfully built and tested such a power source, called in the industry a "wheel supply". Now, Amalgametics really wanted to make guidance packages, not test equipment. So they carefully documented the design and drew up specifications which were about 20% better than the performance of the prototype wheel supply. Then they
sent this documentation out to several electronic test equipment manufacturers with a request for quotation (RFQ),
In the fall of 1962 one of these packages arrived at a small instrument fire in Burbank (unimaginatively named Elin, for ELectronic INstruments). The package is turned over to a project engineer, one Felgercarb N, Eloi, for evaluation. Felger reported to his superiors the next day that A) the design could be readily duplicated in production and B) the specifications could be met, if not easily. Regrettably, Felger continued, it would not be possible to do both of those things at once since the specifications were too stringent for the design.
Having learned to CYA (cover your ass) he then proceeded to conclusively prove his point, taking as an example a parameter of the specification (rise-time) that was familiar to his bosses.
Elin submitted TWO quotations, one to duplicate the design and the other to deliver a proprietary design which would meet the specifications. Both were rejected, and that wheel supply was built for Amalgametics elsewhere.
Back at Amalgametics, the success of Adolf Dorfman case to the attention of his superiors, and he was promoted from engineering group leader to section head. A small clipping announcing this in the Orange County Register is added to Boris' collection, since the announcement mentions the section in question is the inertial guidance test group.
As initial production of the gyros began, it became evident that the production yield was very high, due to Adolf's prescience in incorporating a phase-locked loop into the design of the wheel supply. Adolf was given a substantial raise in pay, a new, large office and an attractive receptionist and an attractive personal secretary. He developed a taste for expensive cigars.
Adolf was also given a key to the executive men's room, which did not make the local paper, and was "elected" to Analgametic's Executive Club, which did make the paper. Boris' clipping collection grew slightly.
A memorandum was circulated pointing out that the acceptance rate was 87% rather than the estimated 2.5% that would be achieved less Adolf's innovation. Since the acceptance rate was actually un-economically high, quality control procedures in the manufacture of the part was cut back, resulting in enormous profits for Amalgametics over the life of that fixed-price contract.
Missiles were built, production gyro guidance packages were installed, and both were dropped into newly-dug holes in the ground in rural Montana. General Rattlesword was naturally pleased with the approaching successful culmination of his program and his hopeful addition of a third star. He decided that a random Secondperson missile would be periodically selected, pulled from its silo, and test-fired. After replacement of the nuclear warhead with a dummy, of course.
Finally, enough missiles were in place that one could be safely expended for test purposes. It was pulled from its hole, trucked to Vandenberg AFB, and aimed due west. That's north of Hawaii and south of the Great Circle trade routes to Japan. When they launched the Secandperson the engines performed flawlessly as the missile flew on a straight line for Kansas City. The launch officer had to destroy it, of course. Regrettably, Kansas City is east, not west, of Vandenberg AFB. Oh, well, you can't expect missiles to perform perfectly every time.
General Rattlesword decided that the next random test would be with two missiles, and would be performed as soon as practicable.
The following week the (highly favorable) quarterly financial report reached the office of the President of Amalgasetics. Noting that the high profits accruing in the inertial guidance group were due to Adolf Dorfman's work, Adolf was (after consultation with the board of directors of the company) promoted to a full Vice-Presidency of Engineering. Naturally, he was assigned a larger yet office, this time with two secretaries in addition to the receptionist.
This was accompanied by full press releases, including photographs of Adolf, photos of Adolf being congratulated by Amalgametic's president, etc. Boris took out a new folder and penned 'Adolf Dorfman' on the tab.
Two new test-firings of production Secondperson missiles occurred on consecutive days at Vandenberg AFO under the personal supervision of General Rattlesword and a coterie of lesser officers, including Colonel Crudcutter. Although both were aimed west, one flew north and the other north-west. The engines of both missiles were performing flawlessly when the launch officer destroyed them.
A private conversation took place shortly after the second (third in all) test-firing. It was between General Rattlesword, seated, and Colonel Crudcutter, standing at attention. We do not know the details of
this conversation but it is possible that Rattlesword called Crudcutter's attention to the fact that he was the resposible officer in charge of the inertial guidance portion of the Secondperson missile program.
On the following Monday Colonel Crudcutter arrived at the offices of Amalgametics in Anaheim and demanded to review the test data on the production gyroscopes, especially the specific gyros whose serial numbers had been determined to be the ones installed on the three test missiles. The test data was found to be in order, well within the specifications required by the military contract. A random audit of other test data found no discrepancies,
It was decided to test additional missiles. In all, twelve production Secondperson missiles were pulled from the ground and test-fired. All twelve engines performed perfectly. The direction the missiles flew was random, one missile heading almost in the direction it was pointed.
A grim team of military brass arrived at Amalgasetics to personally supervise the final acceptance test of a gyroscope. After the test, Colonel Crudcutter personally installed the gyro in the guidance package, having practiced this procedure dozens of times. This was done under the watchful eye of General Rattlesword himself. A small military convoy then sped off for Vandenberg AFB, with the center automobile containing a driver, Rattlesword, Crudcutter and the guidance package.
The convoy drove directly to a launch pad where a Secondperson missile was ready for test firing. The guidance package was installed, the missile set to fly due west, and the personnel retired to the blockhouse for the test-firing. The launch went perfectly except that the missile flew almost due south.
Colonel Crudcutter had taken the precaution of having the company which had produced the initial test gyroscopes under an R&D contract (NOT Amalgametics) build several more. The next day one of these gyros was installed by Crudcutter in a guidance package, the package was installed in another Secondperson missile and the missile was test-fired. It flew forty-five hundred miles downrange and impacted within 1500 yards of its target.
Amalgametics was swarmed over by persons from Military Intelligence, the FBI, maybe some CIA types and a whole bunch of Pentagon brass. Engineers hid under their drafting tables as best they could. It turned out to have been a very good idea not to have ever signed anything related to the guidance package.
Those unfortunates whose signatures did appear had at least one, um, "interview" with Military Intelligence. Adolf Dorfman kept his office doors closed, even brown-bagging his lunch rather than taking his usual appearance in the executive dining room. Colonel Crudcutter set up a tiny office just outside the actual gyro test lab to direct the investigation. He was in phone contact several times a day with General Rattleswi)rd.
The investigation centered on finding the saboteur, it being obvious that only deliberate sabotage by an enemy agent could cause the problems which had occured. Unfortunately they could find no physical evidence or trace of sabotage, and the background checks being frantically performed by the FBI were not turning up any prospective 'moles'.
Ten days into the investigation, a young Lieutenant knocked on Crudcutter's ofice. "Colonel," he said, "here are two sets of test data which I took, personally, on the same gyro. The first one here is well within our requirements. But this second set shows that the same gyro is a piece of garbage!"
"This first set," the Lieutenant continued, "were taken using the standard test setup Asalgametics has been using. This second set shows the same gyro and the same power supply except that I changed the wheel supply to operate at a constant frequency."
"What do you mean, changed to operate at a constant frequency?" snapped Crudcutter.
"Well, the standard Amalgametics test setup uses something called a phase-locked loop" the Lieutenant explained patiently to his (very) superior officer. "What that does is keep the gyro spinning at a constant frequency. But it also means that the output frequency of the wheel supply is not constant, but rather changing continuously to accommodate the gyro."
Now, the fact is that Colonel Crudcutter had no education in electronics and did not understand the explanation. He annoyedly snapped back, "So?"
"So the power supply in the missile flight package runs at a constant frequency" replied the Lieutenant.
This result of the investigation was presented at a meeting of the highest brass. Colonel Crudcutter was the most junior officer present. A number of proposals were presented as appropriate punishment for Adolf Dorfman's role in the debacle. One involved a firing squad but the others were less friendly.
After allowing the others present a chance to vent their spleen, the senior representative of Military Intelligence pointed out that absolutely nothing could be done to or with Dorfman, as he had managed to get himself and his connection with the Secondperson guidance package well publicized. To do anything to Dorfman, even to fire him from his job, would be a clear signal to Russian intelligence that all was not well with the Secondperson guidance package.
And since a great deal of the deterrence credibility of the United States depended on all those Secondpersons in Montana, it would be well not to allow Russia to suspect any problems existed.
General Rattlesward addressed the group: "Our problem is that we have a whole bunch of missiles in place that are all perfectly good if only they had functional gyroscopes. What we need is a crash program to get some good gyros built and installed in those missiles. The outfit that built the R&D prototypes has proved that it can build good gyros, so I propose that they be given an emergency contract for that purpose."
Colonel Crudcutter cleared his throat and said inquiringly, glancing at Rattlesword to his right, "Er, if I say ... ?" Rattlesword nodded.
"As General Rattlesword has just pointed out, what we need are not just good gyros but good gyros fast. Northgroup, Inc. has proven that they can build good gyros but they have no in-place production capability." The Colonel directed his gaze straight down the table to avoid his superior officer's gaze. "There is a company that has the in-place production capability and the trained production and test personnel. They have it now. Under strict supervision and with some changes in test equipment and procedures which I have outlined here" - Crudcutter opened his briefcase and drew out some papers prepared by a certain young Lieutenant - "I believe that this company, Amalgametics of course, is best positioned to get us the gyros we need in an, um, timely fashion."
Some time passed; it became 1966.
And so it happened that Amalgametics received a new contract to produce those same gyros all over again. Because this was a crash priority involving unlimited overtime and almost unlimited capital expenditures, the new contract was for $YYY million. Oh, yes: the word was passed to Amalgametic's President that Adolf Dorfman was a good Vice-President of Engineering and should be kept on.
Because Amalgametics needed a bunch of new production test equipment immediately they approached a company
called NH Research in Santa Ana. The N in NH stands for Norman; we forget what the H stands for. A contract was let after a very short negotiation for a substantial amount of test equipment for what was becoming known as the Secondperson II guidance package.
Since the substantial amount of new test equipment could not be built in the short period required using internal NH company financing or available personnel and the existing production facility, NH was granted a contract calling for advance payments and progress payments. Miraculously, these payments arrived as scheduled. NH rented a vacant facility a half-block down the street and nearly filled it with production workers. Oh, yes: NH's chief engineer (Vice President, Engineering actually but he preferred the title chief engineer) was one Felgercarb N. Eloi.
With the new guidance packages installed, General Rattlesword's random test procedure continued. The new tests almost always resulted in the Secondperson missile impacting satisfactorily close to the intended target.
The board of directors of Amalgametics pointedly asked the President about Dorfman's suitability to continue as a Vice-President of Engineering. The President pointed out the following facts to the directors; A) Dorfman was essentially under Military Intelligence 'protection' and could not be replaced without potential dire consequences in future contract negotiations with the military. B) Very substantial profits were gathered from the first inertial guidance contract as a result of Dorfean's initiative. C) The $YYY million second contract was even more profitable, and would never have been received without Dorfman's er, assistance!
After the profits were totalled and the taxes paid, the chief engineer of NH Research walked into the Santa Ana Cadillac dealer and asked the only salesman on duty for a test drive in an El Dorado (more appropriate for a young bachelor than a Caddy sedan). The salesman looked up from behind the desk in his cubical, saw that the chief engineer was not wearing a coat and tie, and did not even bother to rise as he said no.
The chief engineer drove across town in his 8 month old Ford and ordered a new Lincoln Continental, paying 10% down. It seems the Lincoln people did not have a dress code. When the new Lincoln arrived, the chief engineer paid the balance C.O.D.
This completes the preliminary outline for our new novel "The Missile that Couldn't Fly Straight." This is the novel that is going to make us rich and famous, eventually (we receive no pay for writing this newsletter). Suggestions as to how to make the plot more realistic will be accepted gratefully provided they are offered free of charge.
As hard as it is to believe, some of you do not subscribe to BYTE magazine, including a local friend of ours. Well, you will have to buy the August issue if you are in that group. (Do not bother to rush down to your newstand to get the July issue.) In addition to all the good stuff about the C language and some more stuff about UNIX, there are two articles by Jerry Pournelle that will interest you. (They must interest you if you read this rag.)
The first article is the one beginning on p.312, "The Debate Goes On." Jerry seems to have opinions somewhat similar to your FNE regarding programming languages, except that he is not willing to try writing his own langauge to prove his personal opinions. This first. article of Jerry's concludes: "The language debate will continue." Surely.
(On p.315 Jerry tells us that, '...within two years, one will be able to buy the equivalent of a VAX for $6000 or so. This future "microcomputer" will run at 12 to 15MHz and have a half-million to a million bytes of memory." Within two years, huh?)
The second 'article' is his usual "User's Column". This one is primarily focussed on one subject, the Epson HX-10 and VALDOCS. True, the HX-10 uses a Z-80 and should not normally be of interest in these pages, hmm? Wrong. The HX-10 is a particularly nasty example of what happens when one succumbs to the prevailing wisdom that it is O.K., even desirable, to write operating systems in a high level language rather than assembly. A few quotes to whet your interest:
"The first problem is obvious from the other side of the room. The VALDOCS system is slow. It seems to take forever to do disk operations. ...It took me much more than A MINUTE [emphasis added] to save a one-page memo, then retrieve it to continue working."
"VALDOCS doesn't have any way to print except to store it first. That means it takes a minimum of TWO MINUTES [emphasis added] to address an envelope. ...Deleting the first three pages (of a six-word text file) takes 30 seconds. ...I'd love to be proved wrong, but I don't think Valdocs will ever run properly until something like the 8086 or 68000 is used."
(Jerry, writing your operating system in a high-level language is as big a mistake, maybe bigger, with the 16-bit processors than it is with the Z-80 the HX-10 uses. - FNE)
On a different subject: "In Ithaca (NY) I was taken to task by a professor of computer science. I have, it seems, been unkind to Pascal and have confused the language itself with particular implementations. I pointed out that I can only evaluate implemenatations..."
Jerry, you do not have the proper attitude. The prof is explaining to you that Pascal, like motherhood, is a wonderful theory. You are not permitted to notice that what one has at hand is thirteen welfare children. Theories are much more important than facts, as the math theorists took pains to explain to Oliver Heaviside.
Also: you will not miss the ad on page 346 and the lovely sidebar on page 84. That's at the top of the page, with a blue background. Read the last two paragraphs especially carefully. The reason we want you to read those two paragraphs carefully is that the logic is impeccable and DEAD MRONG!
Still in Aug. BYTE, if you read the small ad by IBS on page 391, note carefully that TWO products are being offered. For $1370 you can have a 384K under-the-hood 7MHz 60000 system (system = 2 boards). Also note that "available" does not equal "included", hmm?
Go next door and borrow your neighbor's kid's copy of Aug. incider, Wayne Green's Apple magazine. On page 1 you will see a beautiful ad for - gasp! - another under-the-hood 68000 board. Boards, actually, this is a two-board system (using two slots?). This ad lists "ETC" software, presumably furnished with the board; languages and software systems which are "supported" and yet another column of "optional software". Will someone please explain to us the critical difference between "supported" and "optional"?
This outfit is even going to have an 'optional' "Artificial Intelligence Laboratory incorporating a Deductive Reasoning System" in Spring, 1984. At 10:47 AM on March 31, we assume?
The point is, we received a lot of phone calls from people who thought that all three categories of software were included with those boards, and why didn't we provide the same?
THERE'S MORE: We bet you missed the very small ad in the upper left-hand corner of page 544 in July BYTE for a 68000 board, one you can buy assembled or bare for $69. Under-the-hood, of course.
EVEN MORE YET: Analytic Engines has actually released a board with software, Softech's P-code of course, to a Beta site for evaluation. (One of these days we are going to find out how A E has stayed alive and paid their bills these past 9 months. Those ads they are running in Nibble are professionally prepared and do NOT look cheap.)
Yes, we have some details about that software. No, we are not going to tell you about it before it is released. It's not good manners that's why. If Beta software didn't have problems, it would already have been released, no?
We have only one competitor in the outside-the-Apple category of 68000 boards, but that competitor's unit is a lot faster than ours at 8 MIPS. We refer, of course, to Acorn. Our static RAM board runs at about 1.3 MIPS and the Grande at about 1.1 MIPS, so we clearly are not competitive with Acorn in performance. Sigh.
Wayne is our favorite publisher, if for no other reason than the fact that, as long as Wayne is around, your FNE rates as a moderate. While you still have your neighbor's kid's inCider, turn to page 6 and read the publisher's column.
In this column he expresses the heretical viewpoint that Apple folks oughta stick their nose outside their Apple compound occasionally and see what the rest of the microcomputer world is doing. (Why do a silly thing like that?) Also: "...the hell we have reserved for the lowlife who keys in a program and then lets a friend copy it can't be put into print." Either Wayne has been taking lessons from Nibble's Mike Neck or the other way around.
Oh, yes: Mike, in his own column in a recent Nibble, suggested three ways of making money in the microcomputer world. Publishing was not one of them. In the meantime, Wayne Green has just sold off a bunch (but not all) of his magazines to CWC Communications for about $60 million.
BYTE is no longer the biggest magazine in personal computerdam, as measured by the number of pages. Looked at a copy of PC lately?
(Those of you who do not like reading about first principles should skip this section.) When comparing the performance of microprocessors offered by differing manufacturers, it is very difficult to determine what is or is not a fair comparison. Especially since most manufacturers do not want a fair comparison, they instead want a comparison which shows that their product is superior.
And let us note that us folks at Digital Acoustics are manufacturers who, presumably, would like to have our offerings appear superior; and barring that, would like to have appeared to have made the correct choice of microprocessor for its product offerings.
We ask that you grant the assertion that your FNE does not consciously flim-flam the readers in our comparisons. We concede the possibility of unconscious bias and permit each of you readers, based on our 'performance' over the past two years, to judge to what degree we are unconsciously biased.
Let us see if we can somehow devise a method of formalizing comparisons at least in some respects, so as to remove as such personal bias as possible,
Most of you readers are aware that there are those microprocessors which are real and those which are not real. Let us define this; a 'real' microprocessor is one which you can purchase at the chip level across the counter of a factory-authorized distributor. Let us further refine this definition: if the product is not currently on the shelf but can be delivered within, say, four weeks, we will consider it to be a real part IF AND ONLY IF that part has in the past actually been sold across that same counter.
Note that some products which do exist are not 'real' under the above definition. The unnamed HP 32-bit microprocessor used in some of its products cannot be purchased as a chip across the counter. 16MHz 68000s exist and have for well over a year now but you can't buy one across the counter.
8MHz Intel 286s are said to exist but they are even scarcer than 16MHz 68000s; you definitely cannot buy one. You can't buy a 6MHz 286 at the chip level either, although you night be able to buy one at the board level (for lots more bucks). The 286 you CAN buy now is a 4MHz part.
The Nat Semi 16032 part you can buy now is a 6MHz part. 10MHZ samples are available in quantities similar to 16MHz 68000s and 8MHz 286s.
Here is what is 'real' under the definition above, ranked in our estimated order of performance:
1. 12.5MHz 68000 2. 10MHz 68010 3. 10MHz Z800X 4. 18MHz 9445 5a 6MHz 16032 5b 4MHz 286 5c 10MHz 8086
It may come as some surprise to you that the 68010 is available, at 10MHz. This is because the Intel/AMD folks have been pilloring the 68000 for having poor performance in multiuser/multitasking applications (true) while ignoring the available 68010, which does have such capability. But note this carefully: our ranking above is for use in single-user, semi-single-tasking environments. (We don't consider print spooling and a type-ahead buffer to be truly multi-tasking applications, since both can be readily implemented on our Grande 68000-based board.)
Conspicuously absent from the above list are Intel's 432 and TI's 99000 (note the 3 zeros), which we believe are failed products.
If the above list were to be reshuffled based on multi-user applications, it is clear to us that the 10MHz 68010 would be in first place, and much more reshuffling would take place below. The 12.5MHz 68000, 18MHz 9445 and the Z800X (X = 1 or 2) would move considerably downscale; the 10MHz Z800X (X=3 or 4) would move into second and the 4MHz 286 into third.
(We are not absolutely certain that the 10MHz Z8003 or 4 are real parts. Since they are just spearcarriers in this episode, we aren't going to worry about it.)
If you cannot identify the 9445 or the 99000, please review the last several pages of issue #4.
It should be evident that the comparisons being pushed by Intel/AMD are for parts which are not 'real' against parts which are not only REAL but actually OBSOLESCENT! (8MHz 68000s = obsolescent; we don't even have them on our price list anymore.) Not being satisfied with such obvious unfairness, they use that $30,000 EXORMACS-turkey with its obsolescent 8MHz clock and its defective 4 wait states as representative of the performance of the 68000.
Two can play the Intel/AMD game. We propose another comparison of a part which does not 'exist' against parts which do exist.
Intel prefers to use as a target a machine which is MUCH slower than the best machines available using Motorola parts. The EXORMACS runs at 8MHz with four wait states, net throughput equivalent to 4.8 MHz with no wait states. That's 2.6 times slower than the 12.5MHz boards we have been shipping. So we will compare against an Intel part 2.6 times slower that the fastest 286 you can buy across the counter. That would be a 1.54MHz 286.
Against that slow 286 part we will naturally compare a part which does not exist. We choose the 18MHz 68010, a part which most certainly will exist once Motorola shrinks the mask (an event which may occur prior to the across-the-counter availability of 8MHz 286s and which probably WILL occur prior to the availability of 10MHz 286s).
So here is our comparison, folks:
18MHz 68010 vs 1.54MHz 286
Which do you think is the superior product?
THAT'S NOT FAIR, you may assert. Of course it's not fair! But it is EXACTLY equivalent to what Intel and AND have been advertising, but with the shoe on the other foot. Just as we have done, Intel and AMD feature products and performance which does not exist as purchasable products on their side, and products which clearly do not represent the best which is available on the side of their competition.
Our comparison above was carefully chosen to be EQUALLY UNFAIR. Surely there is much to be said for equal unfairness?
It is left as an exercise for the student to prepare an equally unfair comparison to counter the comparisons in the very recent Nat Semi advertisements.
We have finally tried something other than the wordprocessor on our Eagle(s). We purchased a data base management program (Condor 20, $295) and decided to learn just a tad about CP/M before trying that DBM program. So we dug out our book on CBASIC and a book on CP/M and tried to write a VERY simple BASIC program.
The CBASIC book suggested using CP/M's ED utility to generate the BASIC source text. Uh uh. ED we have read about. It was designed to edit paper tape on ASR-33s and hasn't changed since. So we used our regular word processor that comes with the Eagle, Spellbinder. And saved the program on a newly formatted disk in B.
Then we removed the word processor disk and inserted the CP/M & CBASIC disk. When we compiled the source program by running CBAS2 we got an error message. It seems there was an error in text position #19! So, if we had a paper tape punch, WE would know exactly where to fix the paper tape... seems that ED isn't the only program that hasn't changed since Moses was a pup.
So we removed the CBASIC disk and re-loaded the word processor disk. We fixed the error, removed the word processor disk and re-loaded the CBASIC disk. We ran CBAS2 again and got no error messages. Then we actually ran the program and it worked!
Thus encouraged, we turned to page 1 of Condor. It seems that the first thing we are supposed to do is copy all of the Condor programs onto a CP/M disk using PIP on the CP/M disk. Well, we didn't have a pure CP/M disk but we did have (a working copy) of that CP/M-CBASIC disk. We put the CP/M-CBASIC disk in drive A and Condor in drive B. Invoking PIP, we keyed in A:=B:*.*(RETURN). Sure enough, lots and lots of files were copied from drive B to drive A. Too many, in fact. The disk in drive A filled up and PIP was unable to copy the next file to drive A. So you know what error message we got?
WRITE ERROR ON DRIVE B
Friends, that is the wrong error message! We showed this to our project engineer, who was somewhat familiar with CP/M from his last job. He said, "Oh, that's normal. CP/M has lots of bugs! At my last job, we would get error messages for drive C, which was odd because we only had drives on A and B."
This is an operating system that has been around so long that many of its utilities are optimized for punched paper tape; it is an industry standard; and it has LOTS of bugs?
Ron Jeffries, in his Jeffries Report, asserts that the impending introduction of IBM's Peanut is going to be the most important event ever for the small computer industry. John Dvorak asserts in Infoworld that Peanut will not have a floppy disk. EN and EET report that Tandon has booked a single purchase order for $310 million from an anonymous customer for floppy disks. None of these matters are related, naturally.
Ron seems to think that Commodore is the only outfit that stands a chance against IBM in the home (not personal) computer marketplace, once IBM is IN that marketplace, as they will be shortly. We think IBM is the only outfit that stands a chance against Commodore in the home computer marketplace. The difference between Commodore and IBM is that Commodore does not deliberately 'bust' its low-end products to protect its high-end products. If Peanut is too good, who will buy a PC? If the PC is too good a word processor, who will buy Displaywriter? And so forth DEFINITELY ad nauseum!
The biggest result of Peanut's introduction is that Commodore will finally have to make a real floppy disk available for the 64. The 1541 is really a cleverly disguised audio cassette tape deck ...
Aug IBM Softalk is out and 1-2-3 is in 1st place with an index of 274.28. We don't know how Contex/MBA is doing because it is not on the list, but we know it is under 10.17 because the program with that index is on the list. The opening sentence of the Softalk presents the bestsellers column is: "Now that it's all over except the shouting in the spreadsheet competition..." Oh, yes: Microsoit's Flight Simulator, also written in 8088 assembly, is still in second place, this month with an index of 108.74. The next best game is FriendlyWare's multiple-game package with an index of 25.17.
Imagine how successful 1-2-3 could be if only it were written in a transportable language such as Softech's Pascal 4.1. (If that last sentence does not sound strongly sarcastic, it is because we are not a professional-caliber writer.)
Incidentally, Softalk's indices cannot be directly compared month-to-sonth as absolute magnitudes; they are only valid as comparisons between programs for a particular month. While Softalk does not reveal the exact formula used it appears to us that the index is actually ten times the percentage of the IBM software marketplace captured on a unit sales basis. If this interpretation is correct, then 1-2-3 has captured, for this latest month, 27.41 of the total unit sales. Since it is much more expensive than average, it seems to us possible if not probable that 1-2-3 is capturing better than 50 cents out of every dollar spent on IBM software.
The other thing you need to know is that 1-2-3 is doing better than that because Soitalk only reports sales made in retail stores, and Lotus now markets the program directly to major accounts with IBM's active assistance.
We have used the slang term "flim-flam" several times in this issue. To us, a flim-flam is a distortion of the facts for commercial gain. If words were sapped two-dimensionally by related meanings, "flim-flam" would be just south of "defraud" and just west of "lie".
No, not male chauvinist pig; we are referring to Micro Computer Printout magazine from England. MCP's columnist Julian Allason has a half-page item in the August issue asserting that Apple is about to dump the Apple III. A short quotation from this story: "Talking to investors in San Francisco last month, Apple's Vice President for Finance, Joseph Graziano, is reported to have dropped firm hints about its impending demise."
Meantime, we hear multiple-sourced rumors in this country that the IIIe (one calls it the III+) is waiting in the wings as soon as substartial stocks of completed IIIs are sold off. (Why does that sound familiar?) We have no idea which version is correct.
It looks like Macintosh is going to be a Lisa re-run, only a year later. Meaning it will be 'introduced' at Apple's annual stockholder meeting in late January (84) and then actually delivered considerably later. The software seems to be running a little late ...
NOOOO! 68000 software late? Gadzooks!
Yes, male chauvinist pig. Our secretary of 4 1/2 year's standing has given two weeks notice to start her own business grooming horses. (Kathy, er, Ms. Meziere hails originally from Cornwall in England, an area which is somewhat, um, rustic.) After waiting two days to see if it was maybe something she ate that day we had to place classified ads for a replacement in the two principal local newspapers.
Kathy had, naturally, already prepared a sample ad which listed all of her duties. Since we have a one-person front office, this is a long list. We explained that her description was commonly known in much shorter form as a "Girl Friday". So we turned to our copy of the Orange County section of the L.A. Times classifieds and picked out a model "Gal Friday" ad and modified it with special information about our company. Kathy had no trouble placing the ad at the Register but she ran into trouble with the Times.
It seems the (female) ad salesperson on the other end stated flatly that the words "girl" or "Gal" or any variant were not permitted to be used in Times
classified ads. She suggested to Kathy that the word "person" be substituted instead. We agreed and suggested the ad be re-worded "Person Friday". The ad salesperson took violent exception to such an imbecile-class expression and attempted to refuse the ad after a long tirade (to Kathy, remember).
At our end, we thought Kathy had been put on "ignore" because of the long silence. When we asked, Kathy placed her hand over the phone mouthpiece to give us a short synopsis of what the "girl" (Kathy's expression, not ours) was doing. Finally, the "Person Friday" wording was accepted under strong protest. We checked our model ad again: yep, "Gal Friday" in that morning's edition of the Orange County section of the L.A. Times classified. Apparently not all Times ad-salespersons are as feminist and/or as vigilant...
There are times when we do not believe what we see happening around us. We are well aware that "Girl Friday" is unacceptable in Gloria Steinem/Bella Abzug circles but since when is "Person Friday" objectionable? Shall we refer to "personhole covers"? Clearly your FNE is an insensitive MCP, as one of our female-type subscribers once pointed out. Oink?
(continued from the front page)
disks for $3 apiece in the U.S. and Canada, $5 elsewhere. (Prices are U.S. funds.) The original offer still stands, though." Jeff H
Disk #1: P.W. Soule's DTACK Applesoft, demos and whatever else is on hand to fill
Disk #2: C.B. Sensenig's BASIC with demos
Disk #3: C.B. Sensenig's BASIC documentation (41 pages single spaced)
DTACK SOFTWARE EXCHANGE
4326 Congressional Drive
Corpus Christi TX 76413
Jeff asked us to be sure that our readers (who own DTACK boards) understand that the Software Exchange will accept code which is less ambitious than a Fortran '77 compiler. Bubble sorts, whatever, will be welcomed as software submissions. If Jeff can distribute disks for $3 per (U.S.) you guys ought to help out, hmmm?
We have just received a marked-up copy of page 26 of the last newsletter from Oliver B, the irate Teuton from Hamburg. You will recall that he was attacking
us, Phase Zero and Brian Kernighan. Phase Zero he didn't like because they do not like to respond to information requests. We used the term "idiot" right in the middle of column 2, and Oliver underlined the word and noted "*&!#*%" in the margin, a comment which is easily understandable in German or American.
But two paragraphs later, when we suggested that our readers might like to write to HIM for information, he writes:
in the margin. It seems that Oliver, who lambastes people who do not respond to information requests, does not like to respond to information requests himself! Oliver: our dictionary uses "hypocrite" to describe that attitude.
So, if you want information about the new pricing of Oliver's cross-assembler, just write to this address:
D - 2000 HAMBURG 65
Oliver will be PLEASED to respond.
LATE SCOOP: The floating point package of Softech's 68000 Pascal interpreter is actually written - poorly, we are told - in machine code. The transcendentals are written in P-code. Apparently the folks who did the 68000 code for Softech were not really 68000 programmers. This is rumor, of course.
COMMODORE has just announced record sales & profits for the year ended July 1. Sales were $681 million, which means they will join Apple in the Fortune 500 next year. Profit is over 13% after taxes.
THE FOLLOWING TRADEMARKS ARE ACKNOWLEDGED: Apple, II, II+, III, soft, ProFile, LISA: Apple Computer Co. Anybody else need a 162nd million ack, have your legal beagles send us the usual threatening note.
SUBSCRIPTIONS: Beginning with issue #19, subscriptions are $15 for 10 issues in the U.S. and Canada (U.S. funds), or $25 for 10 issues elsewhere. Make the check payable to DTACK GROUNDED. The address is:
1415 E. McFadden, Ste. F
SANTA ANA CA 92705
The Phantom dunnit to REDLANDS again, Bob.