DTACK GROUNDED, The Journal of Simple 68000/16081 Systems
Issue # 32 - June 1984 - Copyright Digital Acoustics, Inc
Someone - we don't know who - once observed that 'when it is time to railroad, railroads will appear'. Meaning that when the time for a particular technology is ripe, that technology will appear.
On Monday, May 14th we attended an amazing seminar presented by Weitek here in Santa Ana at the Saddleback Inn, an inconvenient 0.9 miles from Digital Acoustics. We learned more about 'Wonderful Weitek's' 32-bit floating point chip set and its applications AND we learned that the support provided for that chip set is CONSIDERABLY more extensive than we had realized. We also learned to our great surprise that Weitek plans to introduce a 64-bit floating point chip set toward the end of this year! Next issue we will try to give you more detail on this stuff - and why the IEEE (draft) F.P. standard bleeps when you are really in a hurry.
But two days later we receive the latest issue of Electronic Design and SURPRISE!, there is a series of articles on what appears to be that chip set. Then we do a double take and SURPRISE AGAIN, it is NOT that chip set but a nearly identical set made by Analog Devices! It appears to be time to railroad!
These chip sets offer performance more than ten times faster than even a 10MHz 16081 (if you could get one) at a price which is also more than ten times as great. Next month we will tell you more.
is endangered as a local tinsmith is busily chopping up, punching, and bending 100 sets of sheet metal. The sheet metal is genuine stainless steel. Not a plating or finish mind you, but 100% genuine stainless steel. Turns out to cost only two bucks more than a case made of common CRS (cold rolled steel) after it is plated to prevent rust and then painted. We have already ordered 100 ea line cords and line filters and tomorrow we are scheduled to place an order for 100 ea switching power
supplies with Lambda. Next issue look for a picture of a well, no longer mythical case on the front cover (and, we tactfully add, a price list). Anybody got a good name for a non-mythical case?
Having examined several alternatives, we are proceeding with the design of a medium resolution graphics board based on the 7220. It turns out that some startling economies can be achieved by designing around two of the bugs in the existing 7220, namely pan and zoom. If we pitch these out, it turns out that we can make an RGB monitor (3 bit planes) using only one set, 16 pieces, of 64K X 1 DRAMs. And since the price of the 7220 has dropped (considerably), that means the board can be made for a very moderate price (as 7220 boards go). We cannot give you an exact price but look for something in the $700 - $750 range. (We build boards first and price them later when we know exactly how much they cost to make.)
This board will be compatible with the Tandy 2000 color monitor to the extent that the Tandy cable for that monitor will just plug right in!
HALGOL is progressing more slowly than we would like. Our full-time HALGOL programmer has dynamic table allocation completed and operational and he almost has the error reporting routine modified so you are told the line number in which the error occurs. But a certain goof-off - the one we have in mind writes this newsletter - has plain not done much HALGOL work for two months now. It's not going to do much good to ask him to turn his hat around evenings (that's when this newsletter gets written) so it looks like either we wait until our shipment level drops significantly or else we go out and find another HALGOL programmer (implementer?).
We do not expect our shipment level to drop significantly in the short or medium term. Why does this give us less pleasure than one might expect? Dammit, we LIKE implementing HALGOL! You will repeat after us: "the sole objective of being in business is to make money." Well, we're doing that.
Motorola has seen first silicon on the 68020. As you would expect, initial parts are neither operational nor are they COMPLETELY unoperational. We are right on schedule for our predicted 50% chance of having a working sample on 1 Jan 1985. (* The 60-foot tall gorilla's tent. Remember?)
We lied to you last issue, and we apologize - we don't do that very often. On the front page we stated that we were leaving in TWO things we had written much earlier about the criticism we were receiving over HALGOL. In fact, at the last minute we yanked one of those, which constituted all of page 2, and substituted something we had begun for this issue. We don't want to tell lies because we would inevitably be caught at them and then nobody would believe us when we discussed business levels, or costs, or...
On the other hand, there is a clear distinction between a lie and a mistake. For instance, UNIX and PASCAL may take over the world starting tomorrow morning. That gentlepersons, would be a mistake! Boy, would it!
Enough with the mea culpa, back to the good old long-standing fight between Intel and Motorola: a few months back an Intel supporter suggested that our graphics demo program D3.FASTER, which all DTACK owners have in their possession, could not possibly run as fast as we claimed (11.8 seconds on the static RAM board) since that performance could not be approached by the 8086/8087 combination. So we dumped a bunch of source code on him and he went away for a while.
He has resurfaced recently, sending us an 80-page (appx) rebuttal to the 80-page (appx) Motorola architectural comparison of the 68000 to the 80286. Along with it he appended a note asserting that the marketplace has seemed to overwhelmingly select the Intel microprocessor line over Motorola's.
The area in which the 68000 exhibits the clearest superiority to the 80286 happens to be our particular area of interest: large memories, implying large programs and/or data arrays, in single-user and mostly single-tasking environments. (If the situation were reversed, this publication would bear the title "The Journal of Simple 80286/80287 Systems"!)
Please do not permit yourself to be confused by the fact that Motorola and Intel can each prove the superiority of their respective product.
Back to our Intel supporter's assertion that "the marketplace has spoken!". Boy, has it ever! And we certainly agree that the marketplace seems to favor Intel-based microprocessors and their substantial 8080-based software support. However, this is no recent development. Intel micros were favored when we started this newsletter and the situation has not changed in the three years which have passed since then.
As we asserted on page 2 of last issue, Intel dominance will continue (in our judgement) for about three more years. Emerging full software support for the 68000 and Intel segmentation woes in the midst of one-megabyte mass-market personal computer memories are then going to reverse the issue. (There are other contributing factors, such as Intel's failure to match Motorola's - and Nat Semi's - mastery of CMOS. The 68020 will be built using 2-micron CMOS.) However, today is NOT three years today, hmmm?
So let us see what is happening today. That's easy; almost all companies which have chosen Intel micros are hurting and some are going out of business for lack of parts. This is especially true of those folks who are locked into designs based on Intel single-chip microcomputers, but it will shortly also be true of folks locked into the 80186. Intel simply cannot supply enough product to meet demand. One largish company which had been a high-tech high-flyer, Convergent Technologies, is being very, very badly hurt by the shortage of 80186 chips.
In the meantime, IBM now owns 20% of Intel and seems to be having no problems getting microprocessors for the 2 to 3 million personal computers they are planning to ship this year. Intel spokespersons will look you in the eye and with a straight face assert that IBM does not get special treatment and has to stand in line for parts like everybody else. Ho ho ho! (Go ask the Convergent Technologies folks their opinion of that!)
How does this affect us? Glad you asked. We make a 68000 board which has lots of memory - up to a megabyte. We make a VDHR-2 graphics board set as an accessory to that 68000 board. A certain corporation making CADD workstations uses our boards in its product. Since our boards do not use any Intel products, we have not had our - or our customer's - production shut down!
Continuing along those lines, it just happens that the CADD workstation which incorporates our Grande/VDHR boards is proving very popular. So popular that our customer had been having difficulty purchasing enough of those 20 inch monochrome monitors from Video Monitors of Eu Claire WI. In fact, for a while they flat were not getting enough monitors (called 'bulbs' in the vernacular) and delivery times were stretching
out ominously - customers will wait just so long for delivery of a CADD system before turning elsewhere. Guess what happened?
Another customer - a big, big company well known in the CAD/CAM world which happens to use the exact same 20-inch 'bulb' - suffered a short-fall in delivery of certain essential Intel parts and cancelled a month's supply of those big 'bulbs'. So the Video Monitors folks called our mutual customer and asked if they would like to take delivery of a large bunch of 'bulbs' over what had originally been scheduled. You betchum, Red Ryder!
Of course, that put a lot of pressure on us to produce a lot more VDHR systems than had originally been scheduled. We (ahem!) did tell you last month that we had been spending a lot of #%$&*@! time chasing &<#@*$ parts, did we not?
But you do not have to accept our word for this. There is a very simple way for YOU, personally, to track our financial health. We have a full-time employee, a computer science graduate, whose sole duty is to work on HALGOL. He represents a considerable overhead to the company and yet does not, in the short or medium term, contribute in any way to our cash flow, If we get in financial trouble he goes first. For the record, he (unlike Alan Kay recently) does not feel at all insecure. As long as we continue to make progress on (unprofitable, for now) HALGOL, you can be assured that we are in no financial difficulty whatever.
So it turns out that our selection of a microprocessor vendor (Motorola) which is not the marketplace leader turned out to be a good idea after all. And since Apple is making lots of Mackintoshes using 8MHz 68000s then there will be no shortage of 68000L12's, right? (Sigh) Wrong! When, two weeks ago, we went out to buy another 100 L12's, which had been available off the shelf until now, we were told that the factory (Motorola) was quoting 28 weeks!
What had happened to all the L12s? It turns out a couple of big outfits called IBM and XEROX had just bought up all the L12's between them. IBM is obviously
stocking up for the release of its 9002 late this year. NOBODY can figure out what XEROX, a company which builds ever-slower and ever-lower-performance small computers, is doing buying L12's. Our guess is as imbedded controllers in Century Data Systems' hard disk drives. As you might suspect, this certainly puts a kink in OUR plans to use L12's in the Grande and elsewhere, yes?
If you will ignore the wicked grin we are displaying and the self-satisfied manner in which we are rubbing our hands together, we would like to return your attention to one of those 'busted tricycle' ads AMD was running a year ago. Remember the one which asserted contemptuously, "Motorola has their second sources spread out from here to Tokyo"? We are now buying Hitachi S12's, which is what Hitachi calls their 12.5MHz 68000. In fact, we now have more than $30,000 worth of 68000s on order! And you thought we were a small business? (We will return to Motorola when they get their delivery times back to something reasonable. Us round-eyes have got to stick together, you know.)
If some companies are in big, big trouble over Intel's inability to produce enough parts, then Intel is the dirty guy, right? No, that's not the way it works. Let us examine, for instance, the problem with delivery of the 8-bit single-chip Intel microcomputers. Three years ago hardly any personal-computer keyboards incorporated dedicated micros; now nearly all of them do. Two years ago the demand for these $3.50 chips went up by a factor of 3.5, and Intel largely kept production up with demand. In the past year the demand has increased ANOTHER 5 times (!) and Intel has NOT been able to keep up with demand. Surprise?
In the case of the 80186, unforeseen bugs in the design held up the start of mass production for a year. That delay apparently obscured from Intel the coming extraordinary popularity of the part, an extraordinary popularity which was foreseen by at least one outside observer. It is rumored that the pay cut and freeze instituted by Intel in Nov '82 prompted the departure of the 80186 design team to greener [pun intended] pastures. Obviously, it is difficult to locate and eliminate bugs if there is no one around who knows how the chip works. What is REALLY scary is that apparently no one inside or outside Intel is certain that all the 80186 bugs are fixed even today.
The most visible 80186 customer who (like all other 80186 customers) is being hurt badly by the shortage of 80186s is Convergent Technologies. Convergent makes workstations for sale by OEMs. In 1983, nearly all of
its sales were to Burroughs for a workstation designed around the 8086. During '83, Convergent designed a new workstation based on the 80186. They called this the 'N-Gen' for 'New Generation', and signed contracts not only with existing-customer Burroughs but also a whole bunch of other OEMs to deliver lots of 'N-Gens' in 1984-5. Not only can they not get enough 80186s to even service Burroughs, but it turns out they underestimated their costs on the 'N-Gen' and are trying to retroactively negotiate a price increase (rumored to be 30%) with its OEMs. They have also tried, without success, to persuade Burroughs to accept revised 'N-Gens' based on the 8086. [Shouldn't an 8086-based 'N-Gen' be called an 'O-Gen'?]
In the meantime, the hush-hush $250 million deal they signed with A.T.&T. turns out to NOT involve the N-Gen but a totally different design based on A.T.&T.'s (former) BELLMAC 32-bit micro and also on A.T.&T.'s 256K DRAMs. Jack Scanlon, A.T.&T. spokesman, has just asserted that Convergent just plain is NOT going to get those chips to sell elsewhere, as Convergent evidently had planned. (Just what in the hell was that top-secret $250 million pact supposed to cover?) In other words, not all of Convergent Technology's problems can be laid at Intel's doorstep.
(We have just discovered that Convergent also has designed a graphics workstation called the GT around the 7220 and - surprise? - the Intel 80186. Congratulations, guys!)
Intel - and Intel's customers - are victims of Intel's success! Back to that Intel supporter who wrote us that "the marketplace has overwhelmingly chosen the Intel line over Motorola's": that happens to be true for now but only IBM seems to be celebrating that fact, even among Intel customers! (Regardless of protestations of innocence and purity on the part of Intel, IBM is obviously going to continue to get all the parts it wants when it wants them. Here is how this works: Intel asserts that it is only delivering scheduled shipments. So if IBM needs more parts, Intel can find an order dated Aug '83 for those parts, wait for the ink to dry, and ship them! No? Do you also believe in the tooth fairy?)
What wakes up IBM competitors in the dark of night in a deep sweat is NOT that IBM can get parts while they cannot; that is a given. No, the nightmares are caused by the thought that IBM may be able to influence Intel to NOT ship product to a too-successful IBM competitor...
Suppose that you are employed by Intel at a management level either one or two levels below David House, who
is the director of microprocessor development and production. That means you will be at the management level to directly interface with your equivalent at IBM. O.K., IBM calls you up and wants more micros and to hell with whose allocation they come from. You have a wife and two kids to support, a mortgage payment, two car payments, orthodontist bills... and an IBM representative sits on Intel's board of directors. Question: do you say "No" before or after updating your resume?
It is a lot easier to be ethical if one's family will not suffer as a consequence. NOW do you see why IBM is going to get all the parts it wants, and why some IBM competitors wake up screaming in the midst of night?
The reason Apple cannot ship enough Imagewriter printers to meet the demand is that its supplier, I.C. Itoh, cannot get enough Intel 8085 microprocessors. (We hope the Mack owners bragging about their 68000 do not mind that their printer uses a genuine 8085, which is half-brother to the 8080. And Apple, in turn, hopes that folks will not slow their purchases of Macks because of a shortage of printers to work with that really slick graphics program.)
It is a very good idea to join up with the marketplace leader in microprocessors if and only if you can actually purchase and get timely delivery of those processors - ancient proverb.
Peanut, aka PCjr, is stone cold dead. That's D - E - A - D. Dead. Unfortunately, every other personal computer pundit in Christendom has spotted that fact. Most of them are predicting that IBM will recover by replacing the keyboard and dropping the price. That is only partly correct.
Let us give you the full story, and tell you what is going to happen next:
There are times when we believe the general personal-computer populace is incredibly stupid. To this day nobody has noticed that IBM has not yet successfully introduced a personal computer. Yes, we know PC magazine publishes a five-pound tome every two weeks. No, we have not taken leave of our senses.
The highly successful PC, the 256K dual-drive model running 1-2-3 and expecting a Winchester any month now is NOT, repeat NOT the personal computer that IBM introduced. The personal computer IBM introduced had
16K RAM, BASIC in ROM and an audio cassette interface for mass storage - remember? IBM had no idea (literally) of the direction that personal computer would take nor did it have any idea how popular it would be (IBM could have sold the entire first year's production to its own employees!). Because IBM was unprepared for the success of the PC in a segment of the market for which the computer was not designed, shortages of the PC created the opening for the remoras which have affixed themselves to IBM.
Please note that the PC not only failed but failed miserably in its intended market. Also note that the public, influenced by that IBM logo, immediately moved the PC upmarket (as the British would say). Anyone with eyes to see and an IQ one higher than a rock could figure this out for himself or herself, of course.
IBM is a very smart marketing organization and definitely figured all of the above out for themselves. So when they re-entered the low end of the market, right back where they had intended to start before, they prepared their new offering accordingly. Translation: they damn well busted that product so badly that it would be IMPOSSIBLE to move it 'upmarket'! The PCjr is, in several respects, very lovingly and very carefully busted. A great deal of intelligent planning went into busting the PCjr!
And so the PCjr, being permanently limited to balancing checkbooks or calculating that a recipe called for 2.3333333 smidges of thyme for seven people, was finally presented to an expectant public. Guess what?
The public immediately jumped on the PCjr and (surprise?) attempted to move it up-market into 1-2-3 and WordStar and dBASE II land! And when they discovered that IBM had very carefully prevented that, they leaned back in horror! The PCjr, a computer which (please excuse the repetition) was very carefully designed so that it could NEVER run 1-2-3 etc. is now being dumped on by all of the personal computer pundits because it cannot run 1-2-3 etc.!
Since folks do not commonly criticize a Ford Escort for not being a Rolls-Royce, just what is going on here? IBM's problem and the reason for its success are one and the same: that three-letter logo and the perception that the populace has of that logo. Any computer which IBM introduces into the personal computer market will ALWAYS be judged by the public exclusively for its ability to run business-type programs.
IBM has twice attempted to introduce a low-end personal computer and both attempts have been miserable failures! IBM is playing the good Dr. Frankenstein to the PC's monster and the monster is out of control
again. As it happens, the perception of the public is that the PC is enormously successful. Almost nobody is noticing IBM's desperate attempts to fragment the PC marketplace with such tactics at the 370/XT, the 3270 PC, PC/IX, and forthcoming tactics such as LANs which tie to IBM mainframes and more proprietary operating systems (we predict that all of these tactics will be unsuccessful).
However, the failure of the PCjr is HIGHLY visible. Like on the front pages of the Wall Street Journal, a place where IBM does not like to fail. If we were IBM we would be feeling a bit frazzled and snakebit. You see, there is no logical reason for the 'failure' of the PCjr. It was designed to balance checkbooks and it can do that very well. It can even do a decent job of shooting rocks!
IBM is getting a lot of free advice for how to turn the PCjr into a "success". Translation: how to turn the jr into a vehicle for running 1-2-3 and WordStar and dBASE II! If there is anything in this world that IBM is NOT going to do, it is NOT going to turn the PCjr into a PC-killer! Neither can IBM permit the public to perceive the PCjr as an ignominious failure. This leaves only one avenue open to IBM, and fortunately for IBM the Apple Computer Co. paved the way here as well:
There is absolutely nothing that can be done to resurrect the PCjr which would not make it a dangerous competitive threat to the PC. Therefore, IBM is going to follow the Apple III strategy: make a couple of cosmetic changes, like a slightly better keyboard, drop the price a bit and then assert that sales of about 5% that of the PC constitutes a success. Then they can drop the thing three years later when people's attention is elsewhere, as Apple just dropped the III.
That is what is going to happen to the PCjr.
We wonder whether IBM is dumb enough to make a THIRD try at the low-end personal computer marketplace with a computer wearing the IBM logo?
Most of you will recognize that as a line from comedienne Joan Rivers' act. Well, we want to talk, too. And we do, like to other newsletter editors and other journalists outside of these pages. Just today we spoke to a newsletter editor who has long held a view of the UNIX marketplace which is at the opposite pole from ours (if we both held identical opinions on all subjects one of us would be unnecessary). He told
us he had deleted a paragraph from a UNIX item that was aimed at us - beginning something like "that self-appointed UNIX critic..." Well, we are sorry he deleted it, for several reasons.
First, we publish about 20 pages of personally opinionated material every month. Much of that material is on new developments and/or new products. It would be amazing - no, astounding - if there were not fairly frequent mistakes in what we publish. We mean substantive errors, not just a misspelled word or improper use of the word "hopefully" (the latter is endemic, by the way). In fact, if there aren't any mistakes then we are in the wrong career field - we should then be in theology at a fairly high administrative level.
We have never (we believe) hesitated to point out, with our readers' help, the mistakes we have made and we do not hesitate to publish opinions contrary to our own if they are well presented. (We probably would not reprint "Your opinions about UNIX stink!" Then again, we might print it - listing the full name and address of the writer. We're nasty at times!)
This brings us back to that criticism of our views on UNIX which did not get printed. The other newsletter editor urged us to read portions of a book on UNIX, asserting that we would not be disposed to criticize UNIX if we knew more about it. Gentlepersons, one of the two editors in question is missing a point badly. We HAVE studied UNIX one hell of a lot more than the typical mass-market personal computer buyer! UNIX, to achieve the sales figures that are being bandied about, must in fact sell to persons who are FAR less knowledgeable about UNIX than we are now! What in the world does additional training for US have to do with the mass-market sales success of UNIX, pray tell?
Why do UNIX persons always talk about next year's sales, never LAST year's? These are our opinions and we welcome criticism of them. But the UNIX supporters absolutely refuse to address these questions! They drop behind a defense of "well, if you really understood UNIX you would not be criticizing it this way!" What does 'understanding' have to do with a continuing conspicuous shortage of customers?
Yes, we do make mistakes. But can we please talk about those ridiculous UNIX sales projections which wind up falling short by a factor of 20 or 30? Can we, just once, talk about LAST year's UNIX sales? The Commodore/Apple/Tandy/IBM PC folks don't duck out the back door when you ask to talk about last year's sales, hmmm?
[This being a free country, aren't ALL critics self-appointed? Like us and Jean Yates and David Fiedler?]
We may have to make something like the above a permanent disclaimer, right above (below?) the one that lets you make all the photocopies you want. No, we have not kicked any dogs or stolen any candy from babies lately. We have been guilty of a lack of secondary hospitality. Oh? You don't know what secondary hospitality is?
Take, for instance, one of the several W. Germans with which we correspond (mostly in these pages). If one should drop by, we would be hospitable (honest). And we would certainly have a couple of hours to chat. On the other hand, if the same W. German sends a friend by to say hello to Eloi, that friend will not get sch**st! That is what we mean by a lack of secondary hospitality.
This message is directed toward Dr. Mal Hackson (of the pointed tail), Peter R., Bill E., and several W. Germans, all of whom will be treated (by us) courteously in person or over the phone or...
Before DTACK we spent nine years selling environmental noise monitors and did not have the problem of being too busy. Our customers were mostly county health employees who had spent the last several years making sure restaurants used the right rat poison and who had just been told by their boss that they were now an acoustician. On average, these were not the most interesting folks in the world.
We just got paid for the third-from-last noise monitor we will ever sell - we have two left and when they are gone, c'est finis! - but $10,000 checks are not as exciting as they used to be. This one, according to our calculation, will pay for a 9.5 day supply of DRAM this month.
Steve Wozniak wasn't always an unsuccessful rock concert promoter. Back in 1976, for instance, he (and friends) published two remarkable pieces of software in Interface Age. In the Sep '76 issue he and Allen Baum published a 6502 disassembler, for instance. This disassembler later became the 'heart' of the public domain Pet-related Micromon. It is humorous that a lot of Pet types vigorously demand that authors of public domain software be given proper credit, but NOBODY in the Pet world, up to now, has EVER given Woz (and Baum) credit for their very major contribution to Micromon!
An even more incredible piece of software was published by Woz (and Roy Rankin) in Nov '76 Interface Age. This
one was a set of floating point primitives (+, -, *, /, fix and float) for the 6502 which (if you believe this you probably shouldn't because it is incredible!) FIT INSIDE 255 BYTES OF RAM! Woz did that part. Another 240 bytes or so contained LOG and EXP routines which were written by Rankin.
But that basic floating point package inside 255 bytes is INCREDIBLE! We are VERY familiar with that package - figuring out how it worked was a major way we learned how to program floating point. (Sweet 16, Woz's software emulation of a 16-bit microprocessor, was published in Nov '77 Byte magazine.)
Its ignorance, that is. H.P. has just introduced a high-priced portable with, get this, 272K of STATIC CMOS RAM - that's 34 each 8K X 8 static RAMs! And they even built in a spreadsheet in ROM! But they showed their ignorance (right, John B.?) by selecting 1-2-3, which is incorrectly written in fast assembly, instead of MBA, which is correctly written in sloooow PASCAL.
Now, where is H.P.'s woman in the skimpy bathing suit?
In addition to Sage prexy Coleman and Jerry Pournelle (Mondays and Wednesdays), a lot more UNIX critics are coming out of the woodwork. For instance, Will Zachman has just blasted A T & T's UNIX strategy as "the biggest corporate blunder in history." And, he "decried the notion that UNIX would ever gain significant popularity on stand-alone personal computers." Will is the vice-president of corporate research at International Data Corporation, and these quotes are take from InfoWorld, V6 #22 p.9. In an article on page 17, reporter Tom Shea notes: "(UNIX) has been heralded as the next operating system for years. But so far, the UNIX bandwagon has been constructed mostly of talk."
In the magazine PC WEEK, columnist Peter Norton ran a two-part editorial which utterly demolished IBM's UNIX derivative PC/IX. See the 1 May and 8 May '84 issues. (We had thought that UNIX booster David Fiedler had done considerable damage to PC/IX when he revealed that it took four of the ten available last-place benchmark finishes out of 28 systems in contention for last place.)
Every week more announcements of Intel 80186-based products appear. In the May 8 issue, for instance, STM (the Pied Piper folks) announced a portable with a built-in printer and 25 X 80 LCD screen (!) and, oh,
yes: an 80186. Televideo announced a multiuser IBM compatible with both a Z80 and an 80186. And PC Technologies, Inc (lovely name!) announced a PC plug-in card which has BOTH an 80286 and an 80186 (honest)!
If these outfits think they are having a hard time getting parts NOW, wait until IBM (a 20% owner of Intel) announces ITS 80186-based machine! (And will all of you please mail a get-well-soon card to Convergent Technologies, which is really hurting?)
Hercules is running a very interesting ad for a PC-compatible color card these days. Their card is less than half the size of the IBM equivalent and includes a parallel port while using 30 fewer integrated circuits! It seems they are using some custom gate arrays...
While we emphasize the mistakes IBM makes, we really should acknowledge the correct decisions it makes. Having dumped Displaywriter without admitting it has done so, IBM is FINALLY going to un-bust the keyboard for the PC! Honest! Beta evaluation is going on at Coca-Cola headquarters in Atlanta. Imagine! An unbusted PC keyboard and from IBM no less! (We TOLD you, a long time ago, that IBM deliberately busted the PC keyboard to protect Displaywriter, remember? And now there is no Displaywriter to protect.) If you can't wait, go buy a Keytronic 5151 (May Byte p.79).
There are a few of our readers who 1) come from mini or mainframe-land and 2) apparently think we fell off a turnip truck with our VIC-20 yesterday. Almost all of them have taken the trouble to relate the following tale to us:
One first develops one's commercial application program (or language, or operating system) in an HLL. Then, using the rule that such programs spend 80% of the time executing 20% of the code (some say 90%/10%), one goes back and optimizes the slow part of the code.
We are invariably polite when this is related to us, even over the phone. We have never once burst out, "You simple twit! You are the 17th person to feed us that line of hogwash this month alone!" The reason we never burst out like that, in addition to the fact that we are invariably polite and considerate, is that it might be true. One cannot fault the principle of the argument, after all.
On the other hand we have four fingers, er, on the other hand we can question the end result. Was VALDOCS post-optimized for speed? LISA I? Was Microsoft's BASIC for the Mack, which runs 20-30% slower than
Applesoft, post-optimized for speed? Was Context's MBA post-optimized for speed? Was Microsoft's FORTRAN for the PC (widely famed for its sloth and busted in C) post-optimized for speed?
We did not really just fall off a turnip truck with a VIC-20, and being an engineer we are much more interested in results than in theory. If you are a relative new-chum to this rag, you will please remember that we have heard this story already. Honest. Seventeen times this month so far.
We tell you this now because Bob P. has just asserted that he has found a very fast editor (for the IBM PC) which is written in an HLL. Could be. Let US explain to YOU how this is possible. You see, any program spends 80% of the time...
The May issue of St. Mack, which is still (if barely) thick enough that you are not likely to cut yourself on it, has an article entitled "How To Become Rich And Famous Writing For Computer Magazines", Yes, St. Mack did capitalize "To", "And" and "For". The facing page contained four very humorous rejection slips and one (from BLYTE magazine) which was not humorous at all! Humph!
The same issue of St. Mack also has a very long article (with no ads, they have to print something!) on writing your very own book about Mackintosh. Boy, does this country need another book about Mackintosh.
There is one personal-computer industry book in preparation which is NOT about Mack. This one is a book entitled "How to Write a Newsletter", which is a joint effort by several persons in the industry. The chapter on "Choosing a Title" is authored by Hull, the one on "Writing Succinctly" is authored by Eloi and the one on "Meeting Deadlines" is co-authored by Fiedler and Jeffries.
Fat Mack is not a replacement for Mack but a new and different machine. It will have 512K, a hard disk, a serial port rate of nearly 5Mbit/sec vs Mack's current nearly 1Mbit/sec. It will be built in 1985 in a second automated plant in Fremont CA alongside the current plant, which is already nearing its two-shift production capacity of 2,000 Macks per day. Speculation is that the Fat Mack will come in at a base price of $2950 and that Mack will then be dropped to $1950.
Certain Apple critics are laughing themselves silly over Apple's sales tactics for Mack. Apple has been selling Macks for $900, for $1000, for $1250, for $2500... so which price does Apple really want for Mack? Apple does not care what price you pay for Mack - just BUY Mack, PLEASE! Now, about distribution channels: remember how Apple was fulminating against the low-service retail outlets discounting the IIe last year? And threatening to cancel them as sales outlets? Well, when Apple introduced Mack the dealer allocations were based on last year's sales of the IIe!
(When you can pick yourself up off the floor and control your laughter, we will continue.)
Naturally this means that the full-service Apple dealers, the ones who followed Apple's wishes last year and were NOT threatened with cancellation, are now getting precious few Macks. They are naturally deliriously happy over this turn of events. Having proved that it will prostitute itself to obtain market share, it will be interesting to follow the next confrontation between Apple and its dealers, some of whom are (gasp!) DISCOUNTERS! So what do we call Apple ITSELF when it sells Mack for $900, for $1000, for $1250?
In case you are counting, a production capacity of 2,000 Macks per day works out to 43,000 Macks per month or 516,000 per year.
Another way of looking at things: 70,000 units in the first 4 calendar months of 1984 plus capacity production the remaining 8 months equals 410,000 Macks to be produced in 1984. That assumes that Apple can find 410,000 rodent-lovers in 1984, something which is yet to be proved.
T.I., having long played the part of a professional virgin, has given up and climbed into bed with Nat Semi. The chip division will AT LAST be permitted to make a microprocessor which is software-incompatible with the programmable controller which T.I. has been making for 15 years.
Of course, this came AFTER the idiot systems jerks began producing not one but TWO computers which were also software incompatible - the 8088-based T.I. Professional PC and the 68000-based MIT-designed Nu Machine which (naturally) uses the MIT-designed Nu BUS. The microprocessor which T.I. will second source? The 16000 family, especially the 32032. Fairchild, which has an agreement to second source the 16032 but NOT the 32032, is not especially happy for some reason.
is the title of a fictional novel by a priest in which a Roman Catholic Archbishop engages in some hanky-panky with a scarlet lady. By a remarkable coincidence, that priest just happened to be in Archbishop Cody's Archdiocese (of Chicago). And it just happened to surface not long before Cody's death that Cody was using Catholic Church funds to support a certain lady in Miami Beach in luxury. (In an apartment? Heavens, no! In her own mansion (!) - paid for out of discretionary monies turned over to the Archbishop by the Catholic Church.) Why do we bring this up? Because of a misunderstanding over copyrights. Honest!
Back in the summer of 1976, the Southern Cal. Computer Society (SCCS) was getting into full swing. The hot subject of the day (before certain members' monies disappeared) was software piracy, a subject which you may have forgotten about. The then president of the SCCS, whose name we have forgotten, published an editorial in Interface, the official SCCS house organ, on the subject of piracy. His viewpoint: if a programmer has his software ripped off, the programmer should ignore it and simply write another program. Somebody - we forget who - wrote a rather hot-under-the-collar rebuttal which SCCS printed, after asking the writer (twice!) to tone the letter down, which the writer did not.
Back in 1976 there were two other landmark copyright cases in the federal courts. One was a newsletter writer suing the University of California for photocopying his $700/yr petrochemical newsletter at the UCLA library (which legally subscribed) and distributing those photocopies to the libraries of the UC San Diego, UC Riverside, etc. etc. The U.C. legal defense, presented in court, was that it was INCONVENIENT (awww...) to ask a researcher at one of the other U.C. branches to drive all the way to UCLA to see the newsletter! (Apparently it never occurred to U.C. to have each library subscribe. Do you know what the annual budget of the University of California is?) We lost track of that lawsuit, but we certainly hope that U.C. got its knuckles rapped by the courts.
The second lawsuit, again back in 1976, involved no other but our old buddy Archbishop Cody of the Chicago Archdiocese. It seems that the Chicago Archdiocese wanted to save some money, so they had their own hymnals printed, bypassing the publisher's overhead and profits. Like the UCLA library, they neglected to determine whether any of the hymns were copyrighted. In fact, many of them were and three (we remember) of those copyrights were owned by FEL Publications Ltd. FEL Publications brought suit in federal court against the Chicago Archdiocese for copyright infringement.
When Cody returned from one of his frequent visits to Miami Beach and found out about the lawsuit he went into a rage and ordered that 1) the hymnals be destroyed, and 2) that hymnals incorporating hymns copyrighted by FEL Publications would no longer be purchased by the Chicago Archdiocese, and publishers of hymnals were placed on notice of this fact. This latter action constitutes an illegal boycott under federal law, but Cody was unconcerned.
We now find ourselves back in 1984, eight years later. Until just now we had lost track of those two copyright cases. But the Chicago/FEL case is still in the courts! According to the L.A. Times, 20 Apr '84 p.2, a federal jury has just awarded $3.1 million to FEL Publications Ltd. The amount is the sum of three different verdicts returned by the six-person federal jury: $190,400 as compensation for licensing revenues lost, $1 million in punitive damages and $2 million as compensation for "past and future losses" caused by the illegal boycott (you don't think hymnal publishers are going to re-incorporate works by FEL starting tomorrow, do you?).
Naturally, the Chicago Archdiocese is going to drag this one out even further on appeals, etc. - which suggests that the newsletter/UCLA suit may still be before the courts. We ask anyone who has any knowledge of that UCLA suit to fill us in as to its status.
Another reason for mentioning this is that yet another (software) brouhou has sprung up on the Commodore side of the microprocessor media. This one involves a sometimes Protestant minister who has often moaned in print about others violating his copyrights. Now this same minister, in the space of a single page, goes from moaning about HIS problems to what appears to us to be an incitement to others to go out and violate somebody ELSE's copyright! We hope that this is an uncharacteristic slip brought on by a fit of indigestion and so will not discuss this issue here further until we see how that question is followed up in subsequent issues of that other Commodore-related publication.
Your FNE occasionally receives mail which incorrectly addresses his educational status. For the record, your FNE is a high school graduate (PLEASE take our word for this, we do not have the actual diploma at hand). After graduating from high school we played beach bum along the stretch from Hermosa to Zuma beaches during the day and were (barely) gainfully employed at night. When we asserted, a year ago, that we had some expertise with respect to the young cuties in bikinis along the Santa Monica/Malibu beaches we were being perfectly truthful. For about 15 months after
graduation, we had no desire for any kind of advancement. But even Odysseus tired of the lotus and moved on. In our case, we joined the Air Force and quickly became involved in electronics.
After our discharge we started going to school at night while working full time during the day. We kept this up for more years (8) than you might believe, even being admitted to Eta Kappa Nu and Tau Beta Pi, the only honorary societies open to an electrical engineering student at University College (that's night school) at U.S.C.
Regrettably, the company which we owned 45% of became much too busy and prosperous, eventuating in our dropping out of college without acquiring a diploma of any kind. The absence of a diploma can be terribly embarrassing under some circumstances, such as if we ever have to go ask somebody for a job (are you reading this, Chris A.?). However, we happen to be very well connected with the majority stockholder of Digital Acoustics and so do not expect any problems along those lines soon.
Still, it is very nice to be able to truthfully report our absence of a degree without fear of reprisal. Others are not so fortunate.
It would be very interesting indeed to learn how many career hospital orderlies (and such), when finding themselves crossing that 50 - 60 year age threshold have decided to forge themselves a medical diploma and have accepted a position which requires an M.D. in a rural county. Go to any cow county in the U.S. Sneak up behind the county medical examiner or coroner or whatever and suddenly ask, "can I confirm your medical degree"? We would be willing to bet that one of three would turn white as a ghost!
The fact is, lots of folks expand their sense of self-importance as they get older and sometimes they pick up diplomas they did not earn.
The IEEE governing board has nominated two persons to run for president-elect of the IEEE in 1985. Evidence which has just surfaced suggests that the (ahem!) Ph.D of one of those two candidates just might be bogus. The candidate in question has publicly asserted that it is beneath his dignity to produce proof that he really has a legitimate Ph.D! Coincidentally, the 1984 IEEE president-elect has just cashed in his chips (he was 65 years old). It will be most interesting to follow developments of this situation.
Lotus Development, which is the distributor of a spreadsheet program called 1-2-3, has just announced that it made a profit of $7,495,000 on sales of $28,269,000 in the first quarter '84. That profit of 26.5% AFTER taxes means Lotus made OVER 50% profit (!) BEFORE taxes. It really upsets us to see a mismanaged company fall apart like this, so we called Lotus prexy Mitch Kapoor, er, Kapor to find out what the problem really was.
Mitch, we asked, just drop the B.S. and tell us the real truth behind the difficulties Lotus finds itself in, O.K.? "Well, Felgercarb, we made a really stupid mistake when we started out. But it's not our fault, there were only six of us and we didn't have anybody around to give us expert advice. How could we know that by writing 1-2-3 in assembly that we would be locking ourselves out of the Olivetti M-20 and T.I. 99/4A markets?" Mitch answered defensively.
But you have learned your lesson now? we probed. "Definitely. In the future we will write all of our programs in a transportable high level language. We will probably choose PASCAL, considering the success Context has been having with its PASCAL-based integrated spreadsheet, the one they call MBA."
And that should result in a more conventional bottom line for your company? we asked. "Absolutely. As it is, the financial analysts don't quite know what to make of us."
Let this be a lesson for you other readers. If you don't write your mass-market application programs in a transportable high level language, the I.R.S. will come around every three months and take $7 million away from you! And that would be just awful, wouldn't it?
In the 19 Apr '84 issue on p.24 we learn that a West German outfit called Eltec Elektronik is coming out with a 68000 board (VME variety) with provision for a 16081 math processor. Gee! Why didn't WE think of that?
is the name sarcastically applied to the Pentagon by industry suppliers, presumably in private. Those idiot brass hats have actually just demonstrated an Apple computer configured as a nuclear weapons launching system to congress! Why? They want congress to pass laws requiring (pay close attention now) foreign countries (our allies, in fact) to restrict sales of Apple computers! Mama mia!
With millions of Apple IIs in existence, with the biggest problem Apple has is keeping far-east Apple clones from entering the country... words fail us.
What do those idiots in the Pentagon think they are doing? You had better believe that they will not permit us to ship 68000/16081 board sets to West Germany (we would be talking firing squad for REAL!). So what is happening? A West German outfit, Eltec Elektronik, will be shipping such a board set in THIS direction! And we somehow suspect that Eltec might even sell some of those super-secret (sarcasm intended) boards to (gasp!) West Germans!
By the Pentagon's own figures, there will be 80 million "nuclear weapons launchers" (personal computers) in the world in 1990 and they are DETERMINED that not one of them will be shipped out of this country or out of any of our allies' countries, by unilateral act of the American Congress! I yi yi! Did we ever tell you that this country has a serious balance-of-payments deficit (export deficiency)?
Is it possible that the Pentagon brass is really as stupid as it is acting?
Let us tell you about a different bunch of idiots. When Prime Computer prexy Kenneth Fisher left under pressure, he decided to start a new outfit called the Encore Computer Co. For you non-gallics, 'encore' is French for "Play it again, Sam!". To our surprise, Fisher managed to gather a very experienced management team from Prime, DEC and Data General. He even managed to persuade C. Gordon Bell to leave DEC, something which utterly amazed us at the time and led us to question whether DEC was in BIG trouble.
In fact, the management team Fisher gathered was so senior that it had been years since any of them had done any actual work. Instead, they (C. Gordon Bell, for instance) had been overseeing massive engineering teams who actually did all the work. What, you may ask, did Fisher intend to do with all of this senior management talent?
Encore announced that it would be taking over small computer companies which had developed good products but which did not know how to market those good products and provide the needed marketing expertise. In other words, they were looking for outfits 20 times larger and 20 times dumber than Digital Acoustics. The Encore team then braced itself for a massive number of applicants.
A lot of time has passed and Encore has said, "Well, that didn't work, so let's go to plan B." Plan B calls
for Encore ITSELF to develop a family of interconnected parallel processors. Unfortunately, Encore is top-heavy with senior management, not working-stiff engineers of the sort needed to design, develop, and manufacture a new line of computers. Accordingly, it has been attempting a private placement of 6 million plus shares to raise $40 - $50 million. This is the sort of placement that is often attempted after a company has actually developed a product and demonstrated its viability in the market, as Fortune Systems (for instance) did. That reduces the risk to the investor but does not ELIMINATE the risk, as witness Fortune's stock dropping from $22 to below $5 in the past year.
For some reason which we cannot understand (the headline is "Comic Relief", yes?) this attempted private stock placement is not being warmly received by the financial community. Encore has just halved the price of its attempted private offering, meaning the computer design engineers who have not yet been hired will have to work twice as hard, hmmm?
We are wondering what plan C will prove to be. Will Encore's (very) senior management supervise the installation of a fortune-teller's booth in every Computerland store nationwide? Stay tuned to this channel, and 'Play it again, Sham (hic)'!
You will recall that, two issues back, we took Britain's Micro Computer Printout magazine severely to task. We asserted, among other things, that none of the MCP folk would recognize a modern high performance microprocessor if one snuck up and bit them on the ankle. Faced with such devastating criticism, the magazine immediately went out of business!
Yes, MCP has now (sniff) passed on to that Great Press Room in the sky, where the never-ending press run is always finished on time. Let us all pause for one minute as a tribute to that late publication...
Gee! We really didn't think they would take us so seriously.
A horrifying thought has just struck us: some of you new-chums have joined the personal computer movement so recently that you may not remember when Adam Osborne wrote a column for Interface Age magazine. His column was brash, opinionated, and often infuriated others in the industry. For instance, he wrote one column which asserted that one should never purchase a computer which had not been on the market for six months (Jeff and John, turn in your Macks NOW!). Shortly
thereafter, he started a computer company of his own and discovered manufacturers had to SELL their computers during the first six months! Tsk.
In any event, Adam has just had a guest editorial published in PC magazine, the May 15 '84 issue. He presents some "facts" which aren't. Do you suppose Adam has suffered brain damage from staring at 5 inch, 50-odd column screens for too long?
First, he asserts that "Lotus has been shipping between 7,000 and 12,000 packages of 1-2-3 per month". Look, if Lotus bypasses distribution and delivers 1-2-3 to the retail chains directly, then they clear $250 on each $495 list price package (we have seen 1-2-3 discounted as low as $299). Since Lotus just sold $28,269,000 in the first quarter '84 and since Lotus was still a one-product company in that quarter, they sold at least 113,000 1-2-3s, hmmm? (If they go through distribution and so get LESS than $250/copy, their sales would be HIGHER. 113,000 over three months is more than 37,000 copies a month. Where in the world did Adam come up with 7 to 12 thou a month?
We leave it as an exercise for the reader to prove that his estimate of IBM PC sales for the same time period is high (but not by the same ridiculous ratio). In fact, it appears that 1-2-3 is selling one copy per five PCs, which is as much as ANY package can expect (the guy who has a PC at work and another at home, as we have two Eagle IIs, will only buy one, the 2 or 3-PC office might get by with one, and not everybody does spreadsheeting).
The entire point of Adam's argument, that Lotus is not selling nearly as many software packages as it should is exactly wrong - Lotus has in fact saturated the market! But what the heck, Adam, welcome back! You're entertaining even if you are often wrong.
Our suggestion that the Dtack Software Exchange acronym be changed was brought up at a recent DSEx board of directors meeting but was voted down 9 - 4. The operation which De-SExes the human male, qualifying him as a eunuch, is called an orchidectomy. (We plan to keep this up. We only need three more votes to gain a majority of the board of directors.)
Jeff says the $6 printout of Chet's manual is for folks who do not have access to an Apple ONLY. If you have an Apple you can print out your own manual, please?
We have learned that "Consulting Editors" are NOT Editors and hence do not fall under the ethics code
printed in Apr '84 BYTE. This matter is scheduled to be clarified - but BYTE has this pesky four-month lead time. (We have discussed this matter at length with a BYTE editor but we do not want to pre-publish something which will appear later on in BYTE.)
Fortune Systems has now shown a loss for four consecutive quarters. Here are the figures (parentheses indicate a loss):
CALENDAR SALES NET 2ND '83 $12.0M ($3.0M) 3RD '83 $9.0M ($9.2M) 4TH '83 $12.6M ($6.6M) 1ST '84 $15.1M ($3.4M) ------ -------- $48.7M ($22.2M)
Please forgive us for stating the obvious: that is the financial profile of a company with problems. WHAT problems, you ask?
Since Fortune has gone through two managements it is tough to blame their problems on management. They have plenty of cash, at least for two more years at a loss of $22M/yr. The product has been in production for a long time and received LOTS of free publicity; in fact it was the hit of Comdex a few years back. You don't suppose that maybe the limited market for UNIX might be at the heart of their problems?
We have been asked what we have against Fortune. Nothing! It's just that they are a highly visible company in the lower end 68000-UNIX market, which is the one which is supposedly going to really take off any day now (?). And, as a publicly traded company, they have to release financial results every three months and those results get audited once a year, so they have to be honest.
Fortune's problem, as we see it, is that it is correctly positioned to profitably service a market of 600,000 UNIX system sales/yr. Unfortunately that market exists only as a figment of imagination. The imagination of a limited few analysts and newsletter writers, we add. Jean Yates may get invited out to dinner a lot but companies which base their business plans on her sales projections are hurting.
"(A T & T is) aiming at what... ONE [guess who] industry analyst estimates will generate revenues of $7 billion by 1986." [A T & T Chairman] Olson stated. (May '84 IEEE Computer, p.102)
We have a message here for all of you. When we started the DTACK GROUNDED line of 68000 boards three years ago, Digital Acoustics had four employees, two of whom worked full time. The reason we were so small, even though the company was then more than 8 years old, is that there never was a big demand for environmental noise monitors. However, if one listened to the politicians etc. TALK you would have thought it was a VERY large market. (The similarities to UNIX are obvious, no?)
Because we were such a small company at that time we (your FNE) obviously knew everything that was going on. We knew who our customers were. This is not true today, as we realized recently when, by accident, we watched a customer's order being packed. A switching power supply was being included! The reason we were surprised was that we were not aware that we were selling power supplies! Since we did not know we were selling power supplies, we naturally did not know what PRICE the supply was being sold for!
That is a clear sign that our little company has grown some. Three years ago we (D.A. employees) mounted the sockets and bypass capacitors and hand-soldered all of our boards. Today this is true only of the tiny Apple interface board, the one which uses a grand total of six integrated circuits. All of our other boards are farmed out to a company which specializes in stuffing and wave-soldering boards (wave-soldering is mass automated soldering - we will not go into further detail). The guy who runs this other operation, by coincidence, is a guy we shot pool with a lot twelve years ago.
So all we do is stuff the integrated circuits into the sockets, test the boards, burn them in for 48 hours and retest them before shipment. This is done by the numbers, and we can ship a lot of product with few employees. Even so, we now have seven employees, five of whom are full time. Since we are only 20% of the full-time employees, we do not repeat not know everything that goes on around here. In fact, orders are routinely received and shipped without us knowing anything about them (unless the order is $10,000 or more - we still get told about those).
With this growth comes a certain lack of flexibility. As a few of you have discovered, it is pointless to write to us asking to purchase a special part - for instance, a bare (TTL-Iess) Stuffer board. Such an order would involve documentation which simply does not exist - nor do we have a department to produce such documentation. Those of you who have compared our prices with those prevailing in the general market will have discovered that we have the lowest prices around
for what we sell (12.5MHz 68000s, 150 nsec DRAM). One of the reasons that we can have such low prices and remain profitable is that we do NOT have a large overhead - such as a department to handle custom requests. This is the message we have to 'all'.
The message we have to 'one' is that the conditions which are beginning to prevail in OUR still-small company are magnified many times as companies grow larger. Let us take as an example a software publishing house employing 50 persons. If we, one of seven employees of Digital Acoustics, do not know everything that goes on around here imagine being one of 50 employees - even if you happen to be the boss!
Like a book publishing firm, software publishing firms have distinct departments. There are the packaging experts, the illustrators, the ones who write the promotional blurbs and the introductions, and the ones who interface with the 'talent' and finally the 'talent' themselves, who are not usually employees of the publishing company and who do not ordinarily live in close proximity to the publishing company.
How - this is addressed to 'one' - can you possibly believe that a software author also produced the packaging and directed or actually produced the promotional blurbs and introductory material on the finished package offered for sale?
The following is taken from a review by John Martellaro in Peelings II magazine, V5 #2 p.41: "The interface card that I ended up buying was the standard Epson parallel card - it works with everything. In my experience, I have had nothing but grief with fancy parallel cards like The Grappler and the Apple Dumpling. Maybe it's my karma, but things just never have gone smoothly for me using those kinds of cards, and I prefer the simple plain-Jane parallel cards."
Take THAT (Pow! Sock!) Bob P! See #29, p24 col 1. John forgot to mention how well his simple Epson parallel interface worked with his DTACK/HALGOL system - or does that come under "everything"?
V5 issue #4, to come out around July, will carry a review of lots of the Apple/68000 systems, including the DTACK, PDQ, QWERTY, SAYBROOK and probably some others. Peelings II is now running a special offer for persons taking out two-year subscriptions: get either all of 1982 or all of 1983's back issues free with your two-year subscription. They take Mastercard & VISA and even have a 24hr toll-free number (Hey! These guys are serious!) which is 800-345-8112.
The Apple Computer Co. has formally stopped all development work on the Apple III and, according to EN, is busy phoning around outside Apple to find employment for the engineers who were working on the III. The official word at this time is that production of the III will be continued as long as orders continue to come in from the small vertical markets where the III is still seeing a few sales. TRANSLATION: When the few IIIs still on the shelf are gone, bye bye III.
Anybody who is surprised so signify by going to the nearest corner and hanging your head in shame. SAY: You don't suppose that independent business unit failed to show a profit, do you? (see #27 p.25)
The latest advertisements by Nat Semi assert that ALL 16 and 32 bit micros now have floating point math support! How about that! Since we have always been interested in fast floating point operations, we will have to look into what Nat Semi has to offer for the 68000. It's nice to see that a big company like Nat Semi is willing to look after our interests.
We have to like the opening paragraph of Steve's article in May Byte magazine. We also like the fact that apparently all pretense has been dropped that Steve's articles are not in fact promotions for commercial products.
The article is about what is undoubtedly the world's best under-the-hood attached processor. This one is based on a 10MHz Z8001, and you will remember that we have asserted that the only chips you can purchase across the counter that are faster than a 10MHz Z800X are the 12.5MHz 68000 and the 10MHz 68010. One distinguishing feature of the Z800X is that even less software is available for it than for the 68000.
Despite this, note (on pp42-3) the substantial amount of software that comes with the commercial product which Steve promotes. Having had some recent experience in paying for the development of software, it is evident that this commercial product has a hell of a lot of financial backing. (One assumes that the Byte editors do not believe that Steve dashed off that software over a weekend.)
Again, this looks to us like the world's best under-the-hood attached processor. (The hood in question is the IBM PC, not the Apple.) Even the prices on p.54 seem very reasonable indeed - they want $330/256K/150nsec versus OUR $300/256K/150nsec DRAM.
True, the commercial product which Steve is promoting does not support the 16081 math processor, but there are, after all, limits to what one can do under the hood... (One warning: The Z800X uses segmented architecture like the Intel series and hence is not well suited to the megabyte-plus single-user memories of the future.)
Bill does not appear often in these pages because of his Intel/S-100 orientation. Yesterday (May 3) we read about him in the interview in Infoworld and again in Jerry Pournelle's column in Byte. Bill's attitude, as expressed in Infoworld, could give this industry a bad name. He says "I do not sell things that I don't own" and he "pays his bills". Doesn't Bill know that over half of the folks in this industry make a living selling stuff they don't own and that nearly half of them BRAG about it?
The two most obvious things that people sell which they do not own are the Apple design and the IBM PC design. That's not all by a very long shot; a couple of years back there were about 20 companies in Orange County CA alone which were making and selling 16K RAM cards for the Apple II. Anybody who thinks that as many as five of those 20 were original designs go stand in the corner.
One of Bill's S-100 competitors recently bragged that he was bringing out a board that was 'more advanced than anything Godbout makes'. When a reporter asked Bill about that he replied "It's about time somebody made something WE can copy for a change!" (Bill was kidding.)
Almost everybody 'pays their bills' if they have the pecunia. A lot of folks start garage operations in this industry with no knowledge of taxes, licenses, cash flow - all the stuff which comes under the category 'business'. A lot of these folks, with the best of intentions, have found themselves broke and with customers who have fully-paid orders. A careful scan through the small ads at the back of Byte will turn up a few deals (prices) which are literally too good to be true in the long run or even the medium run.
**** PAID ADVERTISEMENT ****
Digital Acoustic's prices are NOT necessarily the lowest prices around (for comparable products). Our prices are merely the lowest prices around for a company which is profitable, which has been around for 12 years and which plans to be around for 12 more years.
Jerry Pournelle spends considerable space in the May Byte 'User's Column' on Bill Godbout and one other. On page 74, right under the headline "Rod Coleman" Bill is quoted thusly: "unless you know what you are doing, don't fool around with 68000 machines." Let us place this quotation in its correct historic perspective.
In late '80 - early '81 Bill Godbout had a thriving business selling Intel-based S-100 boards. Digital Acoustics had a so-so business, a very small one, selling high-priced environmental noise monitors. Both Bill and Digital Acoustics (and a LOT of other folks) got interested in the Motorola 68000 and in its clear performance advantage over other extant microprocessors. (At the time the Z8000s you could buy ran at 4MHz while the 68000s you could barely buy ran at 8MHz.) Both Bill and Digital Acoustics approached Motorola for assistance in designing an inexpensive 68000-based product for the personal computer marketplace.
Neither Bill nor Digital Acoustics knew that Motorola had, at that time, a deliberate corporate policy to keep the 68000 out of the personal computer marketplace. Motorola's convoluted reasoning was that by keeping 68000 software prices high it could attract lots of third party software vendors to sell high-priced software and that lots of available software would be good for sales of the $30,000 EXORturkey. Therefore, Motorola saw to it that all of the application literature made available to Bill and to Digital Acoustics (and to lots of other folks) exclusively featured enormously complex designs. One suspects that Bill, like us, had occasion to see the rapidly receding back of a Motorola salesman.
Given that Bill Godbout already had a thriving business making inexpensive CPUs for the personal computer market, we suspect that Bill threw up his hands and left the 68000 strictly alone thereafter. His quotation on page 74, "unless you know what you're doing, don't fool around with 68000 machines" is entirely consistent with the stuff Motorola was shovelling out in those days. (Bill's company does manufacture a 68000 board these days but they do not support it with software and we suspect that Bill did not have a personal input into its design.)
Digital Acoustics, which did not have a thriving business making CPUs for the personal computer marketplace based on another manufacturer's micros, said "(heavily censored)" and grounded pin 10 on the 68000, since that pin appeared to be the central culprit in the baroque complexities Motorola was then shovelling out. Pin 10 is, in case you have not looked at the logo on the front page lately, called DTACK.
When Motorola reversed its decision to withhold the 68000 from the personal computer marketplace in early summer '82 it did not go back around and inform folks like Bill Godbout (or Digital Acoustics) that it had really been kidding, and that it really was possible to build a cost-effective 68000 system for the personal computer market. So Bill undoubtedly still believes the 68000 should not be used except in enormously complex systems. And so Digital Acoustics has a thriving business selling (relatively) simple 68000 systems.
the president of Sage Computer, is also heavily featured in Jerry's May 'User Column'. We were especially impressed with the way Jerry, having signed a non-disclosure agreement with Rod over Sage's future product plans, nevertheless managed to work TWO hints into his column about how well the 68000 would work with the 16081 floating point math processor. You see, we have reason to believe that Sage is working on a 16081 enhancement to its product. It's like this:
Four months after we published a newsletter with a picture of our first working 68000/16081 prototype, three months after issue #25 with its photo of a second prototype and a schematic and an explanation of the 16081 pins, two months after we published a six-page 68000/16081 software supplement, and one month after we began to sell DTACK peripheral boards for the 16081 we received a phone call. The caller asserted that he was Rod Coleman, the president of Sage Computer. We had a polite, general conversation for about 20 minutes during which 'Rod' asserted that, while he was aware of what we were doing with the 16081, Sage was working on interfacing the 16081 independently. (It seems a LOT of folks were 'independently' interfacing the 16081 in Feb '84.)
We are somehow certain that Rod, under cover of that non-disclosure agreement, impressed Jerry with how his company was at the leading edge of the technology with its upcoming 16081 enhancement for the 68000. Why can't WE ever do anything impressive like that? (To the best of our knowledge, Sage has not even figured out how to fix the 68000's 'zero page problem'.)
Since Sage is coming out with a new box (it would appear), and since Dimension is running lots of ads and Mack is on the street now, this is an appropriate time to discuss the features which are most desirable in a new 68000 computer for an experienced hacker. So let's interview one such: You there: are you an experienced hacker? "Yar." Would you take a moment to chat with us? "I got a choice (growl)?"
Please be calm, sir. Just tell us what you would like to see in a 68000-based personal computer. "Well, I'm a touch typist, so I need a Selectric-layout keyboard with a good 'feel'. Don't ask me what a good 'feel' is (you understand that I am not talking dirty?) but that is important.
"The next thing is the display. It has to have an 80 X 24 or 25 text mode with upper and lower case. The characters should be formed, as a minimum, by a 9 X 13 dot matrix. Next, the computer MUST have a graphics mode, and the graphics mode MUST have more than the 200 lines vertical resolution that the 1977 8-bit computers (such as the Apple II) had. I want to see 400 lines minimum, with 640 lines of horizontal resolution minimum. What I would REALLY like is a 600V X 800H display, but that (sigh) is probably too expensive today for a personal computer. I want color as an option. And I really would like for the display to have a unity aspect ratio.
"The display cannot, repeat cannot, be interlaced. I realize that means a non-standard monitor with a 50MHz bandwidth and a 30KHz horizontal scan rate, but that can't be helped. I am allergic to white canes and seeing-eye dogs.
"You probably think I am placing too much emphasis on the keyboard and the display, but that is what most people spend the most time with (is that not obvious?). Next comes the mass storage. In 1984 I want at least 700K per diskette and I would really prefer the Shugart 765 5 1/4 inch drive, which stores a full megabyte, formatted! It's about time the rest of the industry caught up with Commodore, which has been selling the 8250 dual drive, 2 megabytes of highly reliable storage on line, for over three years now. If I have two floppies with a megabyte each, I don't have such a strong yearning for a Winchester. Oh, yes: I want the DOS to read the diskettes without interleaving. CP/M version 3.X does that using a Z80; I expect no less.
"Now that we have the keyboard, display and mass storage accounted for, we can discuss the hardware (and you thought I would discuss this FIRST, I'll bet!). Well, you have already asserted that we have here a 68000-based computer. I will want it to come with 256K absolute minimum RAM, and I want up to a megabyte to be optionally available. I want the display RAM to be separate from the CPU RAM and I want an optional 16081 math processor. Add a parallel printer port, a serial modem port and a real time clock that stays on when the power is turned off, and I'm happy.
"That's about it for the hardware. Now, for software: I want a very, very simple single-user single-tasking operating system. The operating system should be so simple that you don't notice it's there most of the
time. And when I boot up, I want to see a blinking cursor and nothing else. I (censored) on trash-cans! And I want a BASIC that runs at genuine 68000 speeds, optionally at genuine 68000/16081 speeds. That means that it cannot be busted in PASCAL, C, ADA, or even in COBOL. I want genuine assembly language-based interactive BASIC - a 1984 version of that 1976 Microsoft BASIC, but with MID$ on the left hand side of the equation. And that's it."
Thank you, sir, for taking the time to... is there something else? "Yeah. Can I finish shaving now?"
You will not wish to miss the letter from Jef Raskin on page 22 of May Byte. Read it now, if you can. Else go to a newsstand and read it tomorrow.
We wrote about the Alan Kaye interview in St. Mack for last month's newsletter. It was clear to us that something was bothering Alan badly and we guessed that it was the high-level layoffs going on at Atari. Well, damme if Alan did not quit Atari shortly before we put the last issue to bed. (It would have been much better if Alan had been considerate enough to wait until you all had received the last newsletter before he quit.)
We also told you that Alan was about the only PARC employee who escaped Apple's clutches. So where is Alan working now? You guessed it, Apple! Sigh...
is the name of the UNIX-oriented newsletter written by David Fiedler. ($54/yr U.S., call (201) 625-2925). His March issue (which arrived in May - we wonder if David is related to Ron Jeffries?) focuses almost entirely on the A T & T computer product line. Since David looks at midscale computers with very different eyes than us, it is interesting to note that he has about the same low opinion of the product line, and its price/performance, that we do.
On the last page David lists some UNIX benchmarks, and you know how we love benchmarks. Ten benchmarks are listed; 28 computer/OS combos are listed. The IBM PC appears twice, running PC/IX and again running VENIX. The Sage appears once, running IDRIS. Out of the ten benchmarks, the IBM PC with PC/IX finished dead last four times. Of the remaining six benchmarks, the IBM PC with VENIX finished dead last three times. In other words, the IBM finished last in seven of the ten benchmarks. Sage/IDRIS grabbed two of the three remaining "honors". The Radio Shack 16B, no speedburner, was listed twice and did NOT finish last.
"...I continue to enjoy the Journal of Simple Systems. To the enthusiast it is almost like getting a combination of Playboy and the National Enquirer in thirty pages or less." Wayne S., Elk River MN
"You guys at Digital Acoustics seem to know what you are doing. Why not make your own Trash-68?" Derald R., Vancouver WA
Because we DO know what we are doing, Derald. Let us take that Trash-68 development project at what used to be Odyssey and now is the Philips Home Information Systems Division. With 40 folks at an average of $30K/yr (that's low for Silicon Valley but a buck goes a lot farther in Knoxville), that is an annual payroll of $1.2 million per year. Take a three year development period and apply the usual 100% overhead and you can see that it costs about $7.2 million just in personnel-related expenses to develop a Trash-68. Now we have to pay to set up the production lines and start laying in inventory...
Nope. Trying to enter the mass hardware marketplace would be very foolish for a little outfit like us. We'll just sit here and turn out electronic Cosworth-Ford engines, thank you - FNE
Rafael S. of Warminster PA has sent us a lovely purple bumper sticker with the following headline: "REUNITE GONDAWANALAND!" This message is signed off by the "PANGAEAN PEOPLE'S NATIONAL LIBERATION FRONT." What we cannot figure out is why Rafael would think the sticker would be of interest to far-out left wing liberals. To the contrary, this is obviously the work of mossbacked reactionary conservatives. But 'NATIONAL', Raphael?
"...as much as I hate to admit it, my FORTH compiler, while better (faster) than almost any other compiler, is still only a third the speed of carefully handcoded assembly for the 68000, if you judge by the Byte benchmark (equalizing the clock speeds of the two results). I'm happy with the compiler, but I don't claim that it will equal handcoded assembly. I would like to see what that assembly program was, though; I can think of improvements to the compiled code, but there is nothing I can think of which will triple the speed. Ordinary FORTH is more like a 10-1 ratio with hand-coded assembly for large programs. I've seen 2-1 claimed too (even 1.5-1) but never found it to be true in my test cases. Looking at what FORTH has to do, 2-1 would be pretty incredible." Bruce W., San Pedro CA
Bruce, VALDOCS is a large program written in FORTH. Enough said? (Bruce also commented extensively on the RISC evaluation controversy, knowing that both of us follow Computer Architecture News. Not many of you readers seem interested in the RISC, so we discussed that portion of his correspondence by phone - FNE)
"After puzzling over the timing of the 68000 using the min/max values of the data sheets and noticing many impossible timing relationships I, too, turned to a good scope and made some real life measurements. Now with your interpretation of what was happening I begin to see the light!
"Since Oct. '83 I have been using an IBM 9000 and have been referred to as the most demanding user in the Western region by my IBM technical sales rep. If you are interested in my totally biased comments on this machine let me know and I'll scratch out a few pages (for the newsletter)." Tim H., Chandler AZ
It's nice to know that SOMEBODY out there read the hardware stuff, Tim. Since IBM will be releasing a shrunk-down desktop version of the 9000 called the 9002 late this year, maybe you SHOULD give us a user's-view of the 9000 - FNE
"I know, I know, you do not appreciate the negative flak you are getting on HALGOL. And I am sure that not all your readers are careful to remember how much work and money you (sob!) are putting [sniff!-FNE] into the project. We appreciate you! We really do! We PAY for your monthly ravings, don't forget.
"...make it a more-or-less normal BASIC. An expression should be an expression! And if I want to write REALV = INTEGERV, I had better be able to, and not have to throw in that LET (for which you have no possible speed excuse).
"Enough there. You solicited opinions. I think $800 is reasonable for RGB, and would VERY MUCH like to have a better resolution picture at reasonable cost. As for unexpandable 68000: don't bother.
"About Blue Sky n: After reading #31 1 immediately called Bill Basham, creator of Diversi-DOS, to ask him his opinion of your 1024-byte sector. Unfortunately, I couldn't explain to him WHY you wanted to embark on this madcap scheme, because I couldn't understand it myself." Ralph D., Mt. Laurel NJ
Ralph, you CAN (on the next release) write:
100 CONV INT% TO REALV
We believe you intended to say:
100 LET X% = INT% + REALV
Which can be performed using two steps:
100 CONV REALV TO INTV% 110 LET X% = INT% + INTV%
Now, before you and about 550 paid subscribers puff yourselves up like a horned toad and proceed to jump all over us, let us call your attention to the fact that an obscure language named 'Modula-2', which was designed by an obscure furriner named Niklaus Wirth, has THAT EXACT SAME 'FEATURE' FOR THE EXACT SAME REASON HALGOL DOES! Like us, Wirth did not want to over-tax the compiler. While you are calming down, you can confirm our assertion by, for instance, turning to page 16 of the Mar/Apr '84 issue of the Journal of PASCAL, ADA & Modula-2. Where two persons have independently arrived at identical results in the technical community, the one who publishes first is given the credit. Consequently, all of you who wish to criticize this will direct your correspondence to the furriner Niklaus Wirth, hmmm?
Please un-puff yourself completely before reading further.
$800 is just for the non-interlaced RGB monitor, Ralph. Including a board to drive the monitor runs the price up. The one we are tentatively working on has a 7220 but deletes all the features with bugs, including a couple of desirable ones like pan and zoom. So the price might be $100 more than the $1450 we guessed for the monitor cum 'dumb' board.
We personally agree with you about the unexpandable 68000, and we suspect that 99% of our readers agree. Which is not the same as saying that 99% of the marketplace would agree.
About 'Blue Sky n': Every reader, without exception, who has written us on that subject [as of now] has missed the point completely. The choice is not between DOS 3.3 or ProDOS or Diversi-DOS. The choice is between a 68000 DOS, with the 68000 making all decisions, or a 6502 DOS with the 68000 waiting patiently for the 6502 to do everything and to let the 68000 know whether there have been any errors.
The issue is exactly and precisely the same as the question of whether the 6502 or the 68000 knows where the cursor is when running HALGOL. The correct answer is that the 68000 not only knows where the cursor is at all times but does not EVER permit the 6502 to share that knowledge!
Here is how HALGOL loads a HALGOL program if a 6502 DOS is used:
If we are using a 256K (or more) Grande with Blue Sky n, the procedure is to:
This process proceeds so swiftly using the 68000 that a 130K program could be read in a little over 10 seconds (over because of the delay in initially reading the disk directory.). The 6502 never overflows because we never keep blocks of data in the 6502. There is never a need to store one program as a series of smaller programs with fake names, as would be the case using the 6502.
The 68000 never has to ask the 6502 whether a disk error has occurred because the 68000, at all times, is the sole and only arbiter of whether an error has occurred. Your FNE asserts that he is correct and the rest of you, all of the rest of you, are wrong. Blue Sky n as a HALGOL DOS is inherently superior to any, repeat any, 6502-based DOS. How's that for a 'monthly raving', Ralph?
A HALGOL DOS is INHERENTLY incompatible with, say, DOS 3.3. Those readers who have urged us to make HALGOL DOS 3.3-compatible do not know what they are talking about. Or: if they DO know and understand all of the ramifications and side-effects of even attempting a PARTIAL compatibility (ever try to make a 6502 BRUN 68000 code? A 68000 BRUN 6502 code?) then they have to be out of their mind to suggest such a miscegenation.
Do you realize how much gall it takes to tell a whole bunch of intelligent, educated and experienced readers that they are all wrong? We do so with a great deal of confidence. The persons supporting a 6502 DOS - any one of them - represent the final rear-guard action of the folks clinging desperately to their super-intelligent 6502-based terminals. Gentlepersons, the 68000 is VASTLY superior to the 6502 in WHATEVER area - including DOSs.
[The above paragraph was written on the morning of Tues, 1 May. The noon mail that day contained three letters SUPPORTING our proposed 68000 DOS - and we normally get very little mail on Tuesdays. Perhaps we are not fighting a lone battle after all.]
"The Mackintosh is built in an automated (robotized) factory; has its own magazines (2); and has caused publication of several books (at least at this time with many more forthcoming). Is it a success?
"What IS an order of magnitude?
"Who is John Galt?" Jess E., Dallas TX
Mackintosh has managed to sell 70,000 copies in the 1st 100 days after introduction, as we are sure you have read. The two magazines contain almost no advertising and hence are shrinking or going to less frequent publication or both. The third issue of St. Mack had
nothing but a bunch of who-cares interviews because there was nothing else to write about. The promised deluge of third-party software has not arrived nor does that arrival appear imminent. Worst of all, Mack is a 68000-based machine without a wonderfully swift BASIC; such a BASIC is needed for Mack to become a success (We apologize to John Martellaro for stealing this last from his editorial in Peelings II, Vol 5 #2). With Mack essentially locking out hardware add-ons, those two magazines are going to have to do some soul-searching about what they are going to do for future advertisers. It remains to be seen how many mouse persons who like trashcan icons are around. We say the jury is out on whether Mack will be a long-term success.
On the assumption that your second question is serious, an order of magnitude is generally considered to be approximately a 10-1 ratio of whatever is being measured. 100-1 would be two orders of magnitude. If you really want an exact definition, the logarithm to the base 10 of the ratio of the two quantities being compared is the number of orders of magnitude difference. Most folks, including us, are not so exacting but the pin-head who calls a 2.8 - 1 ratio an order of magnitude deserves to be laughed at.
John Galt is an effeminate wimp who runs a legal clinic in San Francisco for gay welfare cheats and who has contracted AIDS from one of his clients.
What is a 'Caddirector'? - FNE
Part of the mail we get is an occasional HALGOL Review, which is a budding newsletter published whenever the whim strikes by Jeff H., the DSEx (ugh!) director. The latest issue, #6, is four pages short and has a review of HALGOL release #2 written by Pete S. and a rousing defense of Mack written by Jeff himself. Incidentally, Jeff seems to think we have been critical of Mack in these pages. We thought we were being moderate; for critical read the editorial in the fourth issue of St. Mac (which is down to 44 pages, by the way).
Jeff has rubbed our investment in HALGOL up against the number of DTACK hacker boards and has reached some conclusions. One, that we are in this business for the long haul rather than as a short-term expediency, is absolutely correct. However, nobody (including Jeff) has noticed that we have a silent partner helping us finance HALGOL - the U.S. Internal Revenue Service! You see, a dollar spent on HALGOL is tax deductible as a business expense. In the past we have spent 65 cent dollars and this year we are spending 50 cent dollars on HALGOL.
It's too bad several of our readers cannot distinguish
between an attempt to start a controversy and actual for-real criticism. Don't they know that feuds between publications can be mutually beneficial? But (sigh) Jeff has chosen to ignore our sallies...
Reader Terry P. has sent us a copy of a page of the C newsletter published by CWare Corp, publishers of the DeSmet C compiler. Here is their "OOPS!":
"When we wrote the s/w floating point library we used Taylor series to approximate the transcendentals. Not fast, but accurate enough. Well, speed does count in the Dr. Dobbs fp benchmarks. Our s/w time for tan(atan(exp(log(sqrt(x * x))))) is a stunning 19 minutes..." Terry comments "Good grief!" and adds that the time is 1900 seconds, which is over 31, NOT 19, minutes.
Speed does count? Gee, we didn't know that!
"My compliments on your ability as a salesman. You sold me a Tandy 2000! No kidding...
"(Re last issue's front page) Hey, don't be so thin skinned (and stop ogling that picture of me with my pants down)." Norman H., Princeton NJ
Norman, the letter just above should be of interest to you, you being a DeSmet C fan. We demanded a commission check from Radio Shack but they wouldn't give us one. Should we sue? We are seriously offended that you would accuse us of ogling a photo of a naked male, especially one with such knobby knees - FNE
"I saw a pile of Mackintoshes at the recent Trenton computer festival. The monitors seem to be sloppily adjusted. There is about a 3/4" gap between the edge of the image and the edge of the tube (on a 9" diagonal screen, this cuts the image area by about 40%) and some of the images were tilted or very square. Also the beam focus was noticeably worse at the corners. Apple must be using cheap monitors." Hal C., Raleigh NC
Hal, DSEx director Jeff H. will not permit criticism of Mackintosh in these pages. We have, however, persuaded him (finally!) that you do not work for Digital Acoustics - FNE
"Dear Mostly thick skinned FNE:
"Page 2: '1-2-3 is the runaway smash-hit winner' but runs inside 64K boundaries but runs on this 8088/8086/80186 which 'has very good performance inside 64K boundaries but is a miserable failure when
attempting to work with larger memory spaces . Hmmmmmm. Guess Lotus forgot or did not know what miserable failure it was destined to be when it decided it could use all that memory (about 420K on my machine). Wish that I could be such a miserable failure. Just how do you define 'miserable' if 1-2-3 is such a runaway hit?
"Page 3: Praise a thief? Is 1-2-3 less a thief of Visicalc than COMPAQ of IBM PC? You elsewhere tout the market as king but here seem to imply that Bunnell should ignore the market (which wants the praise of compatibles) and ignore the thieves. Hmmmmmm.
"Page 9: Methinks the skin of Software person grows thin. In previous issues FNE was able to use the richness of the English/American language and refrain from such trite as the excrement from a common bull in describing comments from his readership. One can see that FNE needs a vacation from those would be language designers."
"I await next version of HALGOL but for my own use, I will need to access much data in DOS 3.3 data files, knowing you, I may go live in the world of C." Bob P., St. Louis MO
Bob, 1-2-3 written in assembly for a better CPU, such as a 68000 (natch), would be a much better performing program and the algorithms used to access the large memory space would be both much simpler (even trivial) and, of course, much faster. Mack does not have enough memory and LISA I had virtual memory done in software as well as an operating system done in PASCAL. LISA II we don't know anything about yet (does anybody?). Therefore, we know of no suitable mass-market 68000 machine to do an assembly-based 1-2-3 on. In any event the idiot professional programmers have decided that the 68000 is so wonderful that programs can be written correctly for it in a transportable high level language while the 8088 has such low performance that (some) programs, such as 1-2-3, wind up being written in assembly and somehow become successful...
We think that within 3 years a small number of those software types will catch on to where the money is, and that suitable mass-market machines (e.g., a 512K Mack) will be available then. Right now, the idiot professional programmers are KILLING the 68000 by condemning it to run absurdly slow software. Mack BASIC slower than Applesoft indeed!
Mitch Kapor had (and still has) a contract with VISICORP that allowed him to write and introduce an integrated software package. He did. Considering the contract, we cannot see any way in which theft can even be implied over VISICALC vs. 1-2-3. The only part of the IBM PC design which IBM has claimed is proprietary
(aside from the copyrighted pc board artwork) to date is the ROM code. It has taken three clone vendors to court, accusing them of copying this code. COMPAQ is highly visible, to say the least, and IBM has NOT taken COMPAQ to court. In addition, COMPAQ is a portable with several innovative constructional features. It has a hard-disk mounting method which is vastly superior to that used by the IBM XT (you cannot safely slide the XT across your desk!). We would be hard put to call the COMPAQ folks thieves, and we do not. Now, the folks who just plain light-fingered the ROM code - and we include Franklin - well...
Which marketplace are we talking about? The one where drug addicts sell TVs, stereos and Selectric typewriters which were not obtained through traditional distribution channels? Or the one where Apple fans buy $300 Apple II clones, knowing they were imported illegally? Or the retail computer stores which openly sold Franklins with ROMS copied from the Apple II+? Do we see varying shades of gray here?
Methinks you are right and that FNE's skin was growing thin. We agree that the expletive on page 9 was out of place. (The one on page 16 is completely appropriate, however, as you would understand if you were a hardware person.) We were definitely feeling grungy for about ten days and that page was written during that time period.
On the other hand, YOU, Bob, are 4% (as one of the guys in those 25 cubicles) of the reason we were feeling so grungy. It is amusing to note that almost half of the 25 - including you, Bob - have written about this and have all - repeat all - concluded by asserting that their ideas are nevertheless valid and that HALGOL will be the poorer if those ideas are not adopted. Many, like Bob, have added that they may just ignore HALGOL if their pet ideas are not adopted! Awwww...
The first time we were told how DOS 3.3 stored numerical values of data type real it took two days to convince us that our leg was not being pulled! Bob, you better get friendly with C because there is not EVER going to be a HALGOL which stores floating point numbers as an ASCII string! If we need "a vacation from those would be language designers" then why won't those "would be language designers", including you, give us that vacation? - FNE
(We think somebody might be being unreasonable here and we are not absolutely sure it is not us. On the other hand, we ARE sure, ABSOLUTELY sure, that there is no way we can accommodate those 25 guys, including Bob. Have any of you any suggestions which do not include trying to accommodate all those guys while ALSO avoiding telling them to all go to, er, to go find a warm place?)
[The following material can stand alone but is also an introduction to some correspondence with Ulrich S. of W. Germany.]
Except for trivial operations (such as adding two and two) ALL floating point operations have errors. In fact, a floating point number is, essentially by definition, an approximation to a true value. When comparing two different results which might be obtained using identical arguments (using different algorithms or different hardware) the problem is not to identify which result is correct and which is in error, but to decide which is the lesser or more desirable (not necessarily the same thing) error.
There aren't any in the U.S. In Europe, the IEC has adopted, in a hasty, ill-advised fashion, draft 3 of IEEE 754. In the U.S., IEEE working group 754 has submitted draft 10 to the appropriate oversight committee for approval and has disbanded. However, draft 10 has NOT been approved nor is there any indication that it will be approved in the near future.
IEEE 754 is a proposed standard for floating point arithmetic using radix 2. Almost all of the former members of the disbanded working group 754 have set up shop as working group 854 and are developing a radix-free standard for floating point arithmetic. A radix-2 floating point package could then be tested for compliance with either 754 or 854. If you are guessing that both 754 and 854 would produce the same result when radix-2 arithmetic is used, you are guessing wrong. 854 has already diverged from 754 in a particular respect, one in which the members of the 854 working group feel is an improvement over draft 10 of the 754 group.
Therefore, draft 10 of working group 754, which has been submitted for approval, is already (slightly) obsolescent in the opinions of members of working group 854 .
In addition, persons developing real software packages have discovered that certain areas of draft 10 of the 754 standard are ambiguous. When this matter came up at a recent rump meeting of 854, an informal discussion among five (former) members of the disbanded 754 group all disagreed on the resolution of the particular ambiguity(s) to the extent that no two persons supported any one suggested resolution!
As Senator Barry Goldwater noted in a recent letter to CIA head Bill Casey, this is a hell of a way to run a railroad!
The only two hardware implementations which we will consider here are the Intel 8087 and the Nat Semi 16081. While the 16081 uses the data representation specified by 754 draft 10, Nat Semi folks are very open about the chip not meeting that standard. Instead, they concentrated on something they could build in quantity within a reasonable time at a profit. Some persons inside and outside Intel will assert that the 8087 meets the IEEE (draft) standard, but that is not in accordance with the facts. Among other things, the design of the 8087 was frozen long before draft 10 was written.
Cray 1s do not comply with the IEEE draft. Cybers don't. No existing mainframe with hardware floating point support complies with the IEEE draft nor will they do so in the near future; all that hardware will remain in non-compliance until it becomes obsolete and is junked.
Performance-oriented hardware floating point implementations being designed today and which will be designed in the near future are unlikely to fully comply with draft 10 because there is a significant time overhead associated with full compliance. It is more likely that lip-service will be given to the proposed standard to the extent of adopting the data representation - essentially what Nat Semi has done with the 16081.
If draft 10 is adopted as an IEEE standard, any hardware or software implementation of that package would not conform to the draft-3 IEC standard (and of course, vice-versa). It appears that incompatibility between European and American floating point standards has been institutionalized even before an American standard has been adopted. Surprise?
The Apple Computer company has developed an Apple Numerics Manual in three parts. Part one is The Standard Apple Numeric Environment (SANE). Part II is The 6502 Assembly-language SANE engine, and Part III is The 68000 Assembly-language SANE Engine. The three documents taken together are more than an inch thick and weigh more than two pounds!
These documents do not include the assembly listings of the various floating point routines but they do provide extensive examples of how to CALL those routines from assembly language.
One way of looking at SANE is that Apple has developed a software implementation of draft 10 of IEEE 754, implementing the extended precision calculations which are OPTIONAL under that standard. It is almost correct to say that they have developed a software implementation of the Intel 8087. (We had not previously known about the 6502 version.)
These software packages are necessarily very slow as compared to packages such as the Applesoft floating point package, optimized in 8-bit chunks, or our double precision HALGOL package, optimized in 16-bit chunks. The IEEE format uses a 53-bit mantissa and an 11-bit exponent and is not optimized for any hardware. (Some hardware - the 8087 and the 16081 - is optimized for the standard.) On the other hand, the precision is good.
This is the only floating point software provided by Apple for Mackintosh, and it is NOT provided in ROM. If it were in ROM then third party software developers could use it free and it would run 33% faster. Because it must be loaded into RAM from disk it must be LICENSED and it runs more slowly since the RAM is shared with the CRT for video refresh. BASIC and PASCAL for the Mack, when provided by the Apple Computer Co, will both undoubtedly use this math package. Again, this package is very accurate and very slow.
We have no idea who uses the 6502 version or even whether ANYBODY uses it - it would set world records for sloth, if only Digital Research's Z80-based BASIC2 were not around.
As we have told you before, we do not favor software implementations of the IEEE (draft) standard because they will necessarily be very slow. If you do not mind very slow then 'The 68000 Assembly-Language SANE Engine' has many worthwhile features. Don't write us for a copy because we won't send you one (over two pounds and more than an inch thick, remember?).
(In the last issue, Ulrich S. discovered a missing instruction in the old 62-bit F.P. multiply routine and was awarded a fiver for his efforts. In this issue, he observes a difference which he identifies as an 'error'. On with the story:)
"...the major reason for the 14% speed increase turns out to be based on an erroneous assumption. Of the nine partial products, the least significant one is not evaluated; apparently you think that it will not influence the final result. This assumption is wrong, as the following example shows:
10 ASSIGN A = $1000 FFFE FFFF FFFF 20 ASSIGN B = $1000 8000 FFFF FFFF 30 LET C = A * B - .5 40 PRINT C
"The result is
7.62916169705E-6 (old multiply) 7.6291616935E-6 (new multiply)
"This discrepancy is entirely due to the fact that the least significant product is ignored in the new multiply. As a consequence, you lose one significant bit in the final product, which is now accurate to 47 bits only.
"Thanks for the fiver! If you send me one for every bug I find this surely is an unconventional way to get rich quickly. I am looking forward to your new divide routine!" Ulrich S., Aachen W. Germany
Ulrich, you should have quit while you were ahead. Your new correspondence above is quite correct right up to the last sentence in the penultimate paragraph, which should correctly be worded "As a consequence, the result is rounded down instead of up, giving a result which is just as accurate as the original but differing by one least significant bit."
Line (A) above represents the 96-bit product of the two 48-bit mantissas which is obtained when all nine partial products are summed. Since the most significant but of the remainder is set, the 48-bit result is rounded up as shown in (C). Line (B) above represents the 64 most significant bits of the 96-bit product which is obtained when the least significant of the nine partial products is eliminated. Since the most significant bit of the remainder is cleared, the 48-bit result is NOT rounded up, with the result given in line (D). However, both (C) and (D) are obviously BOTH wrong! The actual result is very close to line (D) plus half a least bit! It should also be obvious that (C) and (D) represent equally erroneous results!
The mantissas to be multiplied together are:
FFFE FFFF FFFF 8000 FFFF FFFF
The nine partial products formed are:
FFFE 0001 FFFE 0001 FFFD 0002 FFFE 0001 FFFE 0001 FFFD 0002 7FFF 8000 7FFF 8000 7FFF 0000 ----------------------------- (A) 8000 7FFE FFFE 8000 0000 0001 (B) 8000 7FFE FFFE 7FFF ---- ---- (C) 8000 7FFE FFFF (D) 8000 7FFE FFFE
The simple fact is, eliminating the least partial product cannot possibly cause more than 1/32,768 of ONE LEAST MANTISSA BIT change in the result, and that the difference can be noticed ONLY when the result is teetering at precisely half a least bit - at which point a reasonable person would not care whether the result is rounded up or down.
Ulrich, we would not want you to believe that we were a minority of one on this issue. Therefore, if you will turn to Monolithic Memories' AN 111, by Ehud Gordon and Chuck Hastings [see #22, p.2 par 3] you will see that serious hardware design types agree with us that those insignificant bits in the remainder are best left alone. (Now, if you want to pick an argument with Seymour Cray of Cray Computers or Norm Winningstad of Floating Point Systems, Inc then you clearly should not be messing with small fish like us!)
We believe that we have presented compelling reasons to persuade you to contribute a fiver to the 'Eloi throat-wetting fund'. If you find that you are getting rich less swiftly than you might desire, you are in good company. To prove that we are a good sport, we are publishing the new divide routine less than one week after we first tried it out, thus giving you a very good chance of finding a legitimate error.
Like the new multiply routine, only the most significant 16 bits of the remainder is used, allowing other possibilities of "teetering half bits" - FNE
We have prepared the following form to simplify the task of those folks who have been thoughtful enough to tell us how to program HALGOL:
HALGOL MUST BE:
A [ ] exactly like Applesoft
B [ ] very sophisticated
C [ ] structured
D [ ] all of the above
E [ ] other (specify)
HALGOL CONTROL CODES:
A [ ] must follow Apple II conventions
B [ ] must follow IBM PC conventions
C [ ] must follow CBM 64 conventions
D [ ] must follow MAX 80 conventions
E [ ] other (specify)
HALGOL MUST SUPPORT:
A [ ] the peripherals I own
B [ ] the peripherals I own
C [ ] the peripherals I own
D [ ] the peripherals I own
HALGOL DOS MUST BE:
A [ ] exactly like DOS 3.3
B [ ] exactly like DOS 3.3
C [ ] exactly like DOS 3.3
D [ ] other (specify)
IN CASE OF CONFLICT, THE DIRECTIVES WHICH SHOULD BE FOLLOWED ARE:
A [ ] mine
B [ ] mine
C [ ] mine
D [ ] mine
IF FNE DOES NOT DO WHAT I WANT, HE IS:
A [ ] misguided
B [ ] stupid
C [ ] (censored)
D [ ] all of the above
IF FNE DOES NOT DO WHAT I WANT, I WILL:
A [ ] pout
B [ ] denounce him
C [ ] continue to explain how wrong he is until he comes to his senses and accedes to my modest and very reasonable demands
D [ ] all of the above
FNE SHOULD IGNORE ALL OTHER DIRECTIVES AND BE GUIDED BY MY INPUTS ONLY:
A [ ] yes
B [ ] yes
IF THE PRECEDING QUESTIONNAIRE OFFENDS YOU, try making up your OWN form, remembering that you must accommodate the wishes of at least 25 persons whose viewpoints are decidedly diverse. If you detected a redundancy in our version of the form, that is NOT an accident. We were merely trying to emphasize that matter which our critics believe is most important.
What is amazing is that those critics almost NEVER mention speed. Now, that is downright peculiar because speed is almost all that HALGOL is about (interactivity is what is left). To the six or eight writers who have written something like "Fast is what I like - go to it!" we thank you.
PERMISSION IS HEREBY granted to anyone whomever to make unlimited copies of any part or whole of this newsletter provided a copy of THIS page, with its accompanying subscription information, is included.
THE FOLLOWING TRADEMARKS ARE ACKNOWLEDGED: Apple; II, II+, IIe, soft: ProDOS, LISA, Mackintosh?: Apple Computer Co. Anybody else need a 180th million ack, have your legal beagles send us the usual threat.
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
REDLANDS THIS MONTH lists a new, faster 62-bit floating point divide routine for HALGOL - which only uses a 16-bit guard word!
Yeah, we dunnit again. Lied. Our nose is growing so long there is room for three avians to perch. In fact, three are: an ostrich, a pterodactyl and a great auk! We simply did not get around to the new divide routine until two days before press time. So no REDLANDS this month. **** Sigh ****
CMPA is one of the four forms of the 68000 comparison routine. This one contains a gigantic "GOTCHA". We will pass this one on to you, courtesy of our full-time HALGOL programmer. Suppose we have the following instructions:
MOVE.L #$00008000,A0 CMPA.W A0,A0 BEQ EQUAL
The branch in the third instruction will surely be taken, right? After all, we are comparing A0 with itself! WATCHIT! The branch will NOT be taken because the two operands (A0 and A0!) in the second instruction are NOT equal! Honest!
You will have to read the programming manual carefully on this one. You see, the source operand (A0) is, for CMPA.W, taken as a word value ($8000) and sign-extended to a 32-bit value ($FFFF8000) before being compared with the 32-bit address register A0 ($00008000). In this case, it appears that A0 is not equal to A0!
But we are not really comparing A0 to A0. By specifying '.W' we are taking only the least 16 bits of the address register and sign extending it. Chances are the extended result will not match the whole 32-bit value in the address register.
The point is that ALL 'CMPA' comparisons are 32-bit comparisons, even when '.W' is specified in the size field. And if you did not know that already, would you please not pick on us so much for missing REDLANDS? (Maybe the chapter on meeting deadlines should be co-authored by Eloi, Jeffries and Fiedler?)
All of you know that we raised the price of the Stuffer board considerably because of the shortage of LSTTL. We had Death March Dunkerson go out and make a special buy of TTL to complete the 63 Stuffer boards we have in house. As expected, she had to go to the gray market to get the needed parts. (To get reliable gray market parts we went to CAL-AIBCO to get good Japanese parts with recent date codes - and we paid a 2.5 times premium for those parts.)
We figured what it cost us, including these higher-priced parts, to complete those 63 Stuffer boards and we came up with a new price $20 higher than the old one, or $130. That happens to be exactly half the 'HANDS OFF' price tag we had for a while, but that is a coincidence. So the price of the Stuffer will be $130 until these 63 boards are sold. If the price of TTL goes either up or down, it does not matter because we have already purchased the parts! (Consider yourself lucky that we do not segue into an explanation of FIFO accounting!)
Actually, in its most recent quarter Eagle Computer lost a few cents short of $10M on sales of a few cents over $10M. This is a company which was profitable right up until it signed a consent decree to stop copying the IBM PC's ROM. At that point essentially all of its former customers sat back to see if the revised, un-copied ROM would be compatible with as much software as the original. While they are sitting back, Eagle isn't selling many computers.
In the L.A. area, the Saturday L.A. Times sports section is where the computer ads are concentrated. A reasonably careful search through today's (May 19) edition revealed just one ad which included two tiny lines advertising an Eagle computer - a dealer apparently unloading one at his cost. (Before the consent decree, Eagle ads were commonplace.) A reasonable person could conclude that what Eagle Computer was in fact selling was that exact copy of the IBM PC's ROM. According to EN, Eagle's suppliers are becoming concerned now that Eagle's cash flow is about 50% below its business plan.
To almost no one's surprise, IBM has decided that it wants a major slice of the PC software pie. Its release of the $149 DISPLAYWRITER emulator package turned out to be the first of a whole series which include file managers etc. and which are priced in the $149 range. The heavily-advertised Leading Edge PC wordprocessor, originally priced at $295, was dropped to $200 and a week later to $100. We wonder how those $500-$700 business packages are going to stand up to $149 packages bearing the IBM logo?
Yes, folks, we have a new standard here. Let us flock to that standard and prosper mightily by writing software and by building clones, yes?
Did we mention that there is no longer any shortage whatever of the IBM PC or XT or even jr? That one L.A. area retail outlet has correctly identified the true value of jr? (They give jr away with a PC purchase!)
George Santavana pointed out that "those who cannot remember the past are condemned to repeat it". Perhaps those folks who have been playing tick to IBM's dog should have studied the recent (past 20 years) history of the IBM plug-compatible industry at the mainframe level. What is happening right now is decidedly old-hat to us old-timers.
The easiest prediction to make right now is that IBM will be encouraged by its success in selling applications software to take the additional step of introducing a proprietary and incompatible operating system. The proprietary, incompatible operating system will prove to be a devastating failure! This has all happened before, folks. What's that? Oh, you don't remember IBM's attempt to replace OS/360 with a new operating system? And you don't remember that the attempt failed so ignominiously that it actually dragged down a new generation of IBM mainframes with it? And that IBM had to hurriedly introduce the 370 series, which was software compatible with OS/360?
All of you will undoubtedly be pleased to learn that we have a new police force. The new force is dedicated to good deeds; so dedicated that they are determined to 'do good' to you whether you like it or not. Let us present here a few vignettes which will reveal the value of these praiseworthy folk:
"Look! Some kids trying to play baseball in that vacant lot! Why, they can't be more than six years old! Quick, round them up and take them to the training center... There! We are shackling your legs to your desks for your own good, kids. We know how immature folk are not always conscious of the need to do things correctly. Now, before you start playing baseball, it is essential that you know how to play the game PROPERLY. You over there! Stop looking out the window! Today you will learn the infield fly rule...
"Those two kids must be twelve years old and they obviously don't know what they are doing! Walking down a shady lane holding hands indeed! Hah! Caught you just in time, didn't we? You two climb right in the van now so we can take you to the training center... We're chaining you to these chairs for your own good, kids. You are going to learn about venereal disease, community property rights, the social dynamics of singles bars, the seven most desirable positions for sexual intercourse etc. We will begin today with a film on childbirth. Those things the woman's feet are in are called stirrups. The doctor is using the scalpel to provide a larger passage for... Hey, Sarge: the boy has passed out!"
"Good lord! Look at that programmer over there! Why, she is actually writing a program in BASIC! Look at those GOTOs! GOTCHA! Young lady, we have come in what is obviously the nick of time to save you from a life of constant ridicule and degradation. Get into the van here and we will take you to the training center... We are chaining you to this workstation for your own good, believe me! Now, we are going to teach you the correct way to program. The correct way is in PASCAL, using structured programming. That means that you program from the top down. You will please not notice that PASCAL requires a subroutine to be written FIRST in the program before a higher level routine can call it. Remember that having to both declare and initialize variables before they are used is a virtue. Today you are going to learn about flowcharts..."
"I can't believe it! That guy is actually writing a programming language! Look, buddy we know about your kind. Just move it along, right snappy now, into the van... We are chaining you to this development station so you don't try to get sneaky, see? What the hell do you think you are doing writing a programming language without a license or so much as a nod of approval from the Computer Science priesthood? What's that, Sergeant? HIM? This is the one? Buddy, you are in big, big trouble. We have twenty-five criminal complaints here that you are developing a programming language in a manner unapproved by the complainants..."
Remember: When the 'DO GOOD' police get done with you, you will know for sure that you have been done good to! What right-thinking person could possibly fail to approve of that?
Well, we defensively assert that if everybody else can write about the 9900 then WE can TOO. Even though we told you everything you needed to know about the 9900 way back in issue #4 (pp13-14). InfoWorld is reprinting, as a two-part series, the absolutely hilarious story of the T.I. 99/4A home computer. The author is Texas-based and brings to the story the personalities involved. Please read this story if you get a chance; you'll love it!
In the meantime, turn to the letters column in Jun '84 80 Micro magazine. On page 10 a writer patiently explains to us that the 9900 is superior to the 68000 because the 68000 uses inferior register-to-register architecture while the 9900 features superior memory-to-memory architecture. One purchases technical magazines to learn, yes?