5 Jul 87
DTACK Grounded Inc.
1570 PACHECO B-7
SANTA FE NM 87501
Dear EX-DTACK reader:
There have been a few changes.
All of us ex-Digital Acoustics folk now work for a New Mexico corporation whose legal name is DTACK Grounded Inc. It's located in Santa Fe, which is the capitol of the state, has a population of 50,000, and is at an altitude of 6,900 ft. There is no smog - none - and damn little traffic. Death March and Sure Foot used to take 40 minutes to drive home after work, and now they live two blocks from work. In CA I lived a short 5.15 miles from work; in Santa Fe I live all the way across town - 3.3 miles away!
Before you pack, be advised that it is very difficult to get a technical job in New Mexico- even Los Alamos, where much Star Wars funding is expended, has a hiring freeze right now. In Santa Fe, it is much more difficult to get a job than it is in the rest of the state of New Mexico! I am not exaggerating. I know exactly one employed technical person in the entire state who is not either his own boss or employed by a public institution, and his name is James Shaker.
But if you happen to be rich and looking for a place to retire, Santa Fe is a great place to be. Otherwise - I will say this one last time - bring your job with you, as we did.
'We' is me, James Shaker, Death March Dunkerson, and Sure Foot Tobias, who comprised 100% of the employees of Digital Acoustics Inc., a California corporation, as of 1 Jan '87. Yep, we're all here, and hoping to sell DBASIC to folks much like you, except they own an Atari ST and program in BASIC.
What happened to Digital Acoustics? Nothing, except that it is now dormant. I hope to close it out soon because it has about $60,000 in its bank account, just sitting there. And it has no debts. What happened to its (hardware) product line? That's been discontinued, for the simple reason that sales were trending down and we chose to jump (quit) before we were pushed (went bankrupt). As it is, we are very pleasantly un-bankrupt.
An outfit named Deusenberg once made pretty good cars. It stopped making them because not enough people bought those pretty good cars. I'm sure that after the Deusenberg factory closed, there were folk who wanted to buy just one more car. Sorry, too late. To those folk who want(ed) to buy one last DTACK board, I say: sorry, too late.
Let me modestly point out that very few people are smart enough to close down an operation before it starts losing money. For the fiscal year ended 31 Jan '87, Digital Acoustics reported a before-tax income of $190. Since we reported a before-tax profit of $152,005 the previous year, that may give you further insight as to why we quit the hardware business! Did I ever tell you that I hate to lose money?
We quit making hardware with exquisitely correct timing.
Page 2
Why did the other three employees decide to move to Santa Fe to work for a brand-new company with a brand-new product? Because I'm such a wonderful boss, of course! Uh, you ain't buying that one? Let me delicately hint: you should not compare your '86 XMAS bonus with Death March's, because unless you work for an investment banker, her bonus was much larger than yours. That did have some influence on that $190 pre-tax profit, too - XMAS bonuses are an expense. A legitimate expense.
Unlike many of our former competitors (remember Analytical Engines and the other under-the- hood vendors?), we still exist and can perform service work (both in and out of warranty) on DTACK boards. I just got a letter from a guy who bought a DTACK board a couple of years back who implied that his state-of-the-art board was becoming less so as time passed. Hmm. Two years ago, IBM's 6MHz AT was the hottest PC, the 386 didn't exist, and 256K DRAMs were just becoming affordable. We weren't even shipping the Beaucoup Grande two years ago (as this is written in late June '87). Two years is a very long time in the microcomputer business. We sold DTACK boards for over five years.
In fact, we stayed in the hardware business as long as Fortune Systems did, and we didn't have to go to the public for $110 million, and we don't currently have $62 million in tax-loss carry- forwards, as Fortune Systems now has. (You did know that Fortune finally tossed in the towel, didn't you?)
Digital Acoustics: hail and farewell.
Which brings us to the product which DTACK Grounded Inc hopes to sell for the next five years. Unless the mail got fouled up, if you received the original of this then you should have already received two DBASIC manuals. Two because I immodestly assumed that, as a (former) DTACK reader, you would keep one manual to gather dust on the shelf. I hope that you will seek out someone with an ST to give the other manual (and a copy of the disk) to so you can get an opinion on how lousy DBASIC is. If this doesn't meet with your approval, or if by some mischance you didn't receive those two manuals, I will very generously refund the money you paid for them!
(We even sent two manuals to each of our former overseas subscribers, via slow boat [book rate], of course. Those overseas manuals were mailed in the first half of May, and so you overseas types should receive them about the time you get this letter.)
You probably don't have an ST, so you can't read the documentation on the enclosed diskette, but it's about equal to an old 28-page compressed-print newsletter. One of the things the documentation points out is that our order desk phone number is an order desk phone number, and that the persons personning that desk ain't permitted to have a message pad, even!
Me? I'm gonna keep a large distance between me and that order desk phone. And if James doesn't keep his distance as well, I'll find me another chief engineer who will. Name one writer of a popular BASIC, who is known to have written that BASIC, who hangs around a phone with a published number, hmm?
Let me tell you that receiving two (not one, two) truckloads of books, on nine pallets, each pallet weighing 2,200 lbs, is quite an experience. That's a total of 17,682 DBASIC manuals. After we give away 1,660 to you (collectively speaking), and give 10 manuals to each of 400 Atari user groups, and then 50 or so to each of maybe seven of those groups, we will have given away 7,000 DBASIC manuals.
Page 3
I can't prove this, but I think we're giving away more DBASIC manuals than the total sales of all other third-party ST BASICs in the U.S. combined. So who are we going to sell to? Well, there are about 200,000 ST owners in the U.S. at the moment, and we are not giving a manual to 193,000 of them! What we are going to do is give away, over the next year, over 200,000 authorized copyrighted DBASIC disk labels so ST owners can legally use DBASIC whether they buy it or not. (We already have 65,000 labels on hand as we begin this mailing!)
Actually, our short-term business plan covers the next 12 months, and at the end of that period the number of U.S. ST owners will be close to 400,000, and we will not have given a manual to 393,000 of those good folk. We're in the enviable position of being a software house that's not concerned about next week's or next month's sales. Because we were smart enough to quit selling hardware before we started losing money, that's why.
We think about 30% of ST owners are BASIC programmers and hence are our prime audience. If that estimate is correct, there are 60,000 potential customers for DBASIC now and there will be 120,000 a year from now. We need to sell DBASIC to 8,000 or more of those folk in the next year if we expect to make a profit over that (non-fiscal) twelve month period. How many are we actually going to sell? Beats me. (Oh, yes: we think another 30% of the ST owners are potential BASIC programmers. If DBASIC is successful with our prime audience, we might snag a few of that second bunch a year later.)
The reason we're giving away a bunch of free manuals is that in the ST marketplace, folks are spending their own personal money, not company money. And they're no dummies. They know that any advert for any BASIC is going to say that the BASIC is wonderful and that they should buy it immediately, if not even sooner! So ST folk only buy software if their buddy Joe has it, and they can personally see it run. Our plan is to let the ST owner get his hands directly on the DBASIC disk. And if his buddy Joe doesn't have a manual, then Joe's buddy Bob has one - even if we had to give it to him! The folks who buy DBASIC are not going to be buying a pig in a poke...
How do we expect the DBASIC disk to be circulated? Why, by the very same network whose members (rumor would have it) are pirating ST software left and right. Except that they don't have to pirate DBASIC since we're giving away those free copyrighted labels!
It's going to be fun seeing whether our marketing approach, which is somewhat different in the mass-market personal computer software world, succeeds.
Hmm. The other U.S. third-party ST BASIC vendors tend to copy-protect their disks or require that you keep a key disk in your drive 1 at all times. We're not that smart.
Hmm. The other third-party BASICs all purport to support GEM and, of course, TOS. DBASIC ignores both GEM and TOS.
Hmm. The modern trend is to bust your BASIC by either programming it in a high-level language or by crippling it with icons and pull-down menus and other bit-mapped BS. That way, modern BASICs are assured of being slower than the old 8-bit BASICs that ran on the 6502 or Z-80. DBASIC is old-fashioned and uses its CPU as efficiently as possible.
Hmm. Computer scientists will tell you that if you must sell (ugh!) BASIC, then you have to be sure that it is really PASCAL, with no GOTOS, one statement per line, no line numbers, and religious schisms over use of the semicolon. DBASIC is obsolete, using GOTOS, multi-statement lines, line numbers... all just like the BASICs that came with those tens of millions of Apple IIs, Trash xxs, Commodore 64s, and IBM PCs. That's just awful! Isn't it?
Page 4
Obviously, we don't know what we are doing, since we aren't doing what all the other smart folks are doing. Well, after we go broke lots of folk will be able to say, "I told you so!"
(But please wait until after we go broke to bury us. [Former] PC magazine columnist and IBM PC guru Peter Norton publicly announced that we were defunct and buried us three years ago. Then he had to resurrect us last fall. He wrote about us under the code name Dutch Treat [DTACK], as one of his "kitchen software vendors." See PC magazines July 10 '84 p.111 and Aug '86 p. 104. Since we were busy piling up profits when that first column interred us, I didn't recognize us at the time. The Aug '86 column was a tad more obvious.)
Anyhow, I'm out of the newsletter, and even junk mail, business. In the past six months, I have done lots of writing... about DBASIC, as you can see if you look through one of those manuals, or look at the documentation files on the diskette. In the future (both short and medium term), I will do lots and lots of writing... about DBASIC. We need a collection of programming tips, an assembler manual, a beginners' manual, a revised and expanded DBASIC manual,...
Santa Fe has very nice, but very unpredictable, weather. While some evenings are nice for astronomical viewing, you can't predict that fact four hours in advance. Still, my 17.5 inch Dobsonian telescope should arrive in August, if Coulter's production schedule is correct. (Coulter has finally stopped advertising that unobtainable 29 inch bait-and-switch model.) I think I'll give up the idea of a 36 inch Dobsonian. Heck, with unpredictable weather you can't plan evenings in advance. So why have a spectacular telescope if you can't share it with anybody?
I sold my California domicile and put the money in the bank. I'm now leasing a two-story adobe house on the fashionable (by Santa Fe standards) and historic East side of Santa Fe. You see, the East side is the oldest part of Santa Fe, which is the oldest city in the U.S., having been established in 1609. I have no idea whether DBASIC will prove a fiduciary success and so I also don't know whether my next speaker system will be larger or smaller than my last one! (I designed my last house around the stereo speaker system, and plan to do likewise for the next one.)
Since the house I'm living in has two stories, it also has a staircase. Because this is Santa Fe, the staircase is brick. Because the architect was an idiot, the staircase is damn steep. I have, in the past, enjoyed a nip or three of a Seagram & Sons product in the evening. Given this damn staircase, that isn't practical. Not because I couldn't go up the staircase, but because I might come down the staircase much too quickly! So my waistline is down, and for all I know Seagram is down on the stock market (Business Week has just reported that the Canadian family which owns Seagram is down to their last 100 billion dollars).
I thank all of you for reading my past ranting and raving. Don't buy any wooden UNIX boxes.
Sincerely,
Hal W. Hardenbergh
Page 5
A TOTALLY UNBIASED BENCHMARK COMPARISON, Part 1:
(LDW did it!)
NOTE: These benchmarks were prepared by LDW and published in an
(apparently uncopyrighted) ad on page 45 of the May, 1987
ST Applications magazine. We added DBASIC, of course.
BENCHMARKS:
A) 1M empty FOR/NEXT loops
B) Integer calculations (see listing B)
C) Float benchmark (see listing C)
D) Calc. standard BYTE magazine benchmark (May 85)
E) Sieve benchmark size 7000; determine first 1651 primes
F) Screen output 1000 strings of 70 characters (50 x 20 lines)
Speed comparison (all times in seconds)
INTERACTIVE
rank DBASIC rank GFA rank Fast rank NEW ST
-------------------------------------------------------
A 1 17.6 2 48.1 3 66.0 4 303.0
B 1 50.1 3 527.0 2 526.0 4 1100.0
C 2* 10.0 3 10.2 1 6.0 4 15.5
D 1* 3.56 2 6.0 3 7.2 4 16.3
E 1 0.87 2 14.0 3 16.0 4 38.27
F 1 9.66 2 23.3 4 256.0 3 226.63
INTERACTIVE COMPILED
rank DBASIC rank GFA rank LDW rank Philon rank SoftWorks
----------- --------------------------------------------------
A 4 17.6 3 17.1 1 6.7 2 12.2 5 379.0
B 2 50.1 4 168.2 1 4.3 3 111.0 5 2542.0
C 3* 10.0 2 8.7 1 3.5 4 30.2 5 150.0
D 3* 3.56 2 3.5 1 2.8 4 8.2 5 22.0
E 1 0.87 2 1.3 4 1.8 3 1.5 5 33.6
F 1 9.66 3 21.9 2 13.9 4 58.2 5 62.1
INTERACTIVE COMPILED C's
rank DBASIC rank Megamax rank Mark Williams
----------- ---------------------------------
A 3 17.6 1 5.7 2 6.7
B 3 50.1 1 6.1 2 6.3
C 1* 10.0 3 58.8 2 37.2
D 1* 3.56 3 11.9 2 10.4
E 3 0.87 1 0.46 2 0.48
F 1 9.66 3 63.0 2 42.7
* DBASIC uses a double precision floating point format while the
other BASICS use a single precision floating point format,
except for GFA, which uses 1 1/2 (6 byte) precision.
Listing B Listing C
10 FOR J%=1 TO 10 10 X=1
20 FOR I%=1 TO 30000 20 FOR I%=1 TO 1000
30 Z%=2/5+3*2+I% 30 A=SQR(X):A=LOG(X)
40 Z%=2*3*5 40 A=LOG10(X):A=EXP(X)
50 Z%=2+3+4+5 50 A=SIN(X):A=COS(X)
60 Z%=Z%+Z% 60 A=TAN(X):A=ATN(X)
70 NEXT I%:NEXT J% 70 NEXT I%
Page 6
A TOTALLY UNBIASED BENCHMARK COMPARISON, Part 2:
(I did it!)
NOTE: The following comparison of interactive DBASIC to compiled C is completely unbiased. I know; I wrote it, just as I wrote much of the DBASIC code.
Only a cynic would suspect that I might be biased.
The secret to winning at benchmarking is to make sure that the things your program is good at are, um, well-represented. For instance, the LDW BASIC compiler is very good at integer arithmetic, so three of the six benchmarks LDW used in their ad feature integers (A, B, and E). I can make DBASIC beat any ST C compiler 99 out of 100 tries; I can also make a good ST compiler beat DBASIC 99 out of 100 tries. If you wanta win you gotta choose your benchmarks good. Here, I am going to use a different approach, and examine categories of performance:
- PORTABILITY: No contest. C wins hands down because C was designed for portability. After you finish writing your ST checkbook balancing program in C, you can easily port it to the Sinclair ZX-81, the IBM PC, the Tandy Color Computer, and the Cray 1. DBASIC, on the other hand, was specifically written to run as fast as possible on the 68000 microprocessor and hence is deliberately, as a fundamental design decision, not portable.
- INTEGER ARITHMETIC: No contest. C wins again because it is a compiled language, capable of global optimization. Also C does not check for integer arithmetic overflows, which saves time (DBASIC does check). As a result, a good compiled C is nearly twice as fast DBASIC on the BYTE Sieve benchmark, which is a good all-around test of integer performance.
- FLOATING POINT ARITHMETIC: No contest. DBASIC wipes out both the good ST compiled Cs in floating point. DBASIC wins both the speed contest and the accuracy contest. That's partly because the C language requires, as part of the language specification, that single precision arithmetic be performed in double precision. It is also partly because DBASIC has a six-year-old, and hence very mature, floating point math package.
- STRINGS: This is a tough one. Computer experts agree that BASIC, in general, has much better built-in string-handling facilities than C. DBASIC, which emulates the industry-standard Bellevue strings, wins on convenience. On elapsed time? We've never seen a reasonably standard string benchmark. C might win, but the programmer would have to make a heroic effort because C simply does not have good built-in string capability. We're going to award this one to DBASIC.
- DISK THROUGHPUT: No contest. DBASIC wins. All the ST compiled Cs use TOS because C is about compatibility. DBASIC is about speed on the 68000 and ST, and so uses a DOS which is nearly twice as fast as TOS.
- INTERACTIVITY: No contest. DBASIC obviously wins because it is interactive; the ST compiled Cs obviously lose because they are not interactive. What could be simpler?
- MATCHING ASSEMBLERS: Tie. Both DBASIC and the ST Cs include 68000 assemblers. (Ours is faster because it is written in assembly, but it's almost undocumented right now.)
- SCREEN OUTPUT: We're including this for the same reason LDW did: we beat everything in sight! (LDW did not know about DBASIC.) But Megamax must have an awful lot of lead ballast in its screen output code; it is over six times slower than DBASIC. The Unicorn is over four times slower. Well, I told you that if you want to win, you have to choose the right benchmarks!
- THE BOTTOM LINE: Obviously, the language you choose depends on how much time you want to spend learning it and what you want to do with it. I can't tell you whether DBASIC or a compiled C is better for you because I don't know what your needs are. I will say that, as interactive BASICs go, DBASIC goes damn fast!