Q&A: Wirt Atmar

Wirt Atmar: A Scientific Mode of 3000 Faith

NewsWire Q&A

If the HP 3000 world had to elect only one pure scientist of computing, you could make a strong case for Wirt Atmar. He's been at the heart of many of the most important advances in the HP 3000 capabilities during the past decade -- proposing big-concept items like IMAGE's SQL interface; working behind the scenes providing new capability like Postscript handling under MPE; providing grist, through articles in HP publications, for the mill of the users' revolt at the Boston 1990 Interex meeting which led to HP's Customer First strategy; appearing in front by organizing last year's World's Largest Poster Project to celebrate the 3000's ability to set records. Atmar approaches anything he does with the discipline of a scientist, because at his heart that's what he is. He is much more than president of AICS Research: a biology research associate at Chicago's Field Museum of Natural History as well as an electrical and computer engineering professor in his home state of New Mexico. His dissertation for a combined interdisciplinary Ph.D in Biology/Electrical and Computer Engineering was entitled "Speculation on the evolution of intelligence and its possible realization in machine form." Atmar was writing about machine intelligence in 1976, before PCs were even popular.

It's those endeavors outside of the 3000 arena that have enriched what Atmar brings to his HP 3000 experience. In his younger days he worked in the Nuclear Weapons Effects Laboratory at New Mexico's White Sands Missile Range, where he said his job was to calculate how much destruction a full nuclear exchange would deliver. After picturing the end of the world in so much detail, he's been happy to lead a company whose application, QueryCalc, is often used for a task as mundane as sending invoices -- because "the HP 3000 sending invoices is an enormously pacifying event."Atmar's calm faith in the capability of the 3000 to serve as the most important resource in a company's information environment is strong, but it's based on a scientist's proof instead of raw zeal. We asked him about the alternatives and what's good for IMAGE in the second part of our interview.

What is it about the HP 3000 at its fundamental design that makes it a good engine for commerce -- say, better than a choice of Unix or Windows NT?

There are a number of people who regularly post on HP 3000-L that I hold in high regard, even though I've never personally met most of them. One of them is Jim Byrne. About a year ago, he wrote the answer to your question:

"I run HP 3000's, an HP9000/8xx server, a number of WinNT 3.51 WS and a clutch of Win95 WS. Of the four groups of machines, my time in support breaks down roughly like this: 10 percent, 60 percent, 10 percent, 20 percent. Notice who gets the biggest chunk? Now, the HP 3000 is by far-and-away the most important machine of the group. It runs the entire company. The thing is, it just doesn't need me."

I don't believe anyone could write a more powerful and succinct answer than that. But Jim's experiences aren't unique. His story has been told a thousand times over. That common response nonetheless represents the opportunity for a truly significant productivity boost for a company and remains the best possible reason to promote the HP 3000 as the platform of choice. If you're the person who's responsible for the business and signs the checks, you couldn't pick a better reason to either choose or stay with the HP 3000 than Jim's statement.

During Glenn Osaka's tenure as General Manager of CSY, he organized HP 3000 customers into various lists: the "A" customers, the "B" customers, and so on. The "A" customers were the very large accounts, the ones that would represent the most lucrative contacts for the least effort, the "B" lesser so, and so on.

I've always disagreed with that philosophy. The customers that Glenn was seeking out are intrinsically the most fickle, least committed customers that one could find. I've always thought that ultimately, if they were the only customers sought out, the death of the HP 3000 was inevitable.

We, too, have large-organization customers, but I've been constantly disappointed in them as imaginative users of the HP 3000. The problem doesn't lie in the people who work in these large organizations. They're no brighter or dumber than anyone else. Rather, the problem is intrinsic to their situations. The people in data processing are so isolated from the business of their businesses that they don't generally have any clear idea of what the organization does nor do they have the opportunity to exploit the HP 3000 for its maximally profitable use. Neither do they feel sufficiently empowered to make good use of the enormous amount of data that fills their databases. Rather, they have resigned themselves to see themselves as merely caretakers of the data.

That, however, is not the attitude held by our smallest customers. These people know their businesses very well, and they know their HP 3000 databases inside and out --- and they can generally make their HP 3000s sing. These people would have been characterized as "ZZZ" on Glenn's list, but over the years, I've come to regard them as the most important users of the HP 3000, if for no other reason than they are the harbingers of the future.

Seventy percent of our customers run their HP 3000s without a data processing staff. They're business people. And they want to use their machines for business purposes.

Ultimately, all technologies become simple. And it will happen to computers as well. From one very real perspective, having a data processing staff on hand that tends only to the needs of the machines is an overt admission of a fundamental failure in the basic design of the computing machinery and their use. It's much like requiring you to have one or two mechanics ride around with you in your car. If it were critically necessary for you to go to K-Mart, you would spend the money and hire the necessary technicians. And if everyone else did it too, it would actually seem as if it were a normal part of owning and operating an automobile.

Given the current intensity of demands placed on the average data processing staff, it may seem like the grossest form of heresy to suggest that much of that activity is superfluous, but eventually all of the turbulent hustle and bustle that characterizes today's computing environment must disappear it any of this to ever truly become a productive activity for the organization that originally purchased the equipment. And it will. It always has in the past for every other similar form of technology.

I worked for RCA Service Company from 1959 to 1963 -- and I consider that experience wholly instructive. RCA was first introducing color TV to the nation during that period. RCA's sister company, NBC, introduced in 1959-1961 a series of programs (Bonanza, Vic Damone's Lively Ones, The Perry Como Hour, The Jack Parr Show, Walt Disney's Wonderful World of Color, and NBC Saturday Night at the Movies) in color solely to sell color television sets.

Color television receivers were quite expensive then, about one-third to one-half the price of a new car (just as the first HP 3000 I bought was three times the price of my house). And when we sold a color set, we would sit with the customers in their homes for two or three nights in a row to be sure that they knew how to adjust the sets to get the best possible color. And we were busy as hell. The sets -- as were all other TV sets of the era -- were tube-based and were quite unreliable devices. They required an enormous degree of effort to keep them well tuned, converged and operating at peak performance.

But all of that has disappeared, as has the TV repairman himself. A thriving industry that was so busy that it would seem as if it would go on forever is now virtually extinct. Color TVs are now composed of only a few integrated circuits, self-converging, self-tuning, and are now so cheap that it is has become more economical to dispose of a failed set than repair it.

The same thing will inevitably occur in data processing -- or at least in the form of data processing currently being conducted in most businesses -- if any of this activity is ever going to pay for itself. The organization that makes this happen first will be the one that will prosper. Unix persists now because it is the operating system that has been used in universities for the last 15 years, but only because it was free.

The students who graduate from college have the same attributes as anyone else: once they learn anything well, they have the very human desire not to change or view the world in a significantly different light. And these are the same people that eventually become data processing staffs -- and consider it only normal that things are as complex and complicated as they are. But they're not the wave of the future; rather, they're probably the tail end of it.

Large data processing staffs will almost certainly go the way of the TV repairman. To a degree, you can already see the movement in that direction with the disappearance of the HP SE. Data processing managers, SEs, or TV repairmen all represent an enormous cost that gets invisibly bundled into the price of any machine. But the simpler, the more automatic, and the more reliable a machine can be made, the more inexpensively it can be sold. The agency that can create such a device first will be the one that will ultimately prosper.

Unix is no model for such a path -- and no one should be fooled into believing that it is. Although there may seem at the current moment an immovable amount of inertia associated with the intricacies and delicate natures of Unix shell programming, it has all of the future of a well-experienced color television repairman, who by his very nature, had to be an excellent electrical engineer during the earliest days of color TV. A color TV is at least as complicated a device as any computer. It is, in fact, a very elaborate analog computer. The level of demand placed on the earliest color TV specialists was an overt measure of the level of design immaturity (and outright failure of design) that characterized receivers of that era. But it is a stage that was rapidly passed through -- as occurs in all maturing technologies.

That day is not as far off as you might imagine. Indeed, in many offices, such as the great majority of our customers, it's already here. The key to long-term success will always lie in the evolution of very efficient, highly reliable and very simple designs, produced at the lowest possible cost, just as it was for color TVs, audio hi-fi equipment and automobiles. No other commercial computing platform is as well positioned to become that ideal device as is the HP 3000.

What's your opinion on computer productivity -- have we achieved some noticeable measure of it yet, or are we still working on it?

Productivity means using computing machinery in ways that reduce manpower and increase profitability at the least possible cost. In that regard alone, I've often wished that I give tours of many of our customers' operations.

A couple years ago, I "ambushed" several of our customers by calling them and tape-recording their answers, without giving them any advance warning or telling them that I was recording their comments, asking them precisely that question: How do you feel about your level of productivity?

I made these tapes specifically to send on to Harry Sterling, Jim Sartain and others at CSY, simply because HP never really gets the opportunity to honestly talk to customers such as these. However, I would be pleased to send copies on to anyone that asks. Simply send me an e-mail at atmar@aics-research.com. Please be warned however that the tapes are completely unedited and wholly unrehearsed and are of only moderate quality.

What you will hear on these tapes is a level of excitement and joy concerning the level of productivity that these people have been able to attain that is uncharacteristic of computers in general. But it happens more regularly and more easily on HP 3000s than on any other platform I know.

Is it time for the HP 3000 community to begin to accept other systems as essential parts of their information systems?

Heterogenous system mixes are more or less the norm nowadays in larger organizations, not so much because anyone plans it that way but because that heterogeneity simply occurs. Companies buy one another and inherit their systems. Or a data processing manager several generations back had a penchant for one system over another. But rather than embrace all of this heterogeneity, most organizations would be far better off doing whatever they can to choose one system and standardize on it. Rather than perpetuate heterogeneity, their lives would be enormously simpler and their costs of operation would be greatly reduced with a single-system solution.

A little while back, WRQ ran advertisements with the tag line, "If you only have to connect to one host system, count your blessings." Although WRQ didn't mean that line to be taken in the same way that I do, the sentiment is precisely correct. Much of the chaos that characterizes the modern computing environment is due to nothing more than the accumulation of the thousand little gotcha's associated with trying to tie all of this heterogeneity together. Heterogeneity intrinsically makes the entire system much less reliable and less stable, and tends to make it far less usable than it would have been if it were an architecture of a single host computer connected with terminals (or PCs acting as terminals).

In computing, the words simplicity, reliability and productivity are spelled precisely the same, letter-for-letter. The first great trick to making all of this work is to keep the system architecture simple. The last thing I would recommend to people is to "accept" heterogeneous solutions. To do so, I believe, is a fundamental mistake. Rather, they should work very hard to reduce heterogeneity.

Where does the PC-LAN environment still fall short of delivering what the HP 3000 provides?

PCs were originally adopted into the organization without caveat or fiat. They more or less just showed up. They promised an era of easier computing that would put computing power where it belonged: on the desktop of the knowledge worker. But mostly they were cheap. And the value that "cheap" brings to an organization should never be undervalued. All things being anywhere close to equal, cheap nearly always wins.

PCs are here to stay. They have specific benefits that truly do make them easier to use than a traditional terminal-based system. However, shared information is the lifeblood of any organization and evolutionary pressures developed almost instantaneously 20 years ago to link all of the PCs together in some form of communicating web. But it didn't take long for the problems associated with peer-to peer communication to become evident.

After 20 years of trying to introduce the PC into the organization and make it a cost-effective and productive tool, where we're headed now, as a group, is right back where we started 30 years ago: central host-based processing, interconnected with terminals. Only now, the terminals are likely to be rather uniformly Windows-based, Pentium-class processors with bright colorful displays, priced at about $800 or less per unit. In the 20-year interim, however, we've tried cooperative computing, client-server, and distributed computing. All of these approaches have fundamental defects that are intrinsic to their designs and which cannot be easily papered over. The worst of all of these is the PC-LAN. It is a fragile architecture that is enormously expensive to maintain. More importantly than that, it has essentially never been made to work in any practicable sense, especially when compared to what can be accomplished on a small host, such as a HP 3000. Anyone who would recommend PC-LAN-based design for their organization is going to cost their company a lot of money before their recommendation is finally abandoned.

You consider IMAGE intrinsic to the success of the HP 3000. What does it continue to offer that makes it a better choice than other databases?

IMAGE's overt qualities are, of course, its legendary reliability, robustness, simplicity and efficiencies. Running an Oracle-like database of equivalent complexity will generally require a much larger machine with a much larger attendant staff, both of which will generally prove themselves to be extremely expensive items.

An attribute of IMAGE's that is of equal importance however -- but one that is rarely spoken of, although it saves its users an enormous amount of money over its term of use -- is IMAGE's openness. "Openness" may sound like a strange adjective to apply to an HP-proprietary database, but it is appropriate. There are at least a thousand people outside of HP who profoundly understand the inner workings of the IMAGE database. This deep understanding benefits everyone. The people who possess this knowledge are often, of course, vendors who sell auxiliary products for the HP 3000, but as anyone associated with the HP 3000 for any time will tell you, these people have been traditionally extremely generous in passing that information on to anyone who cares to ask. The knowledge that this community imparts, both to itself and back to HP, represents a resource that saves every IMAGE user hundreds of thousands of dollars, although most of those savings are invisible to the ultimate end-user.

Advocates for the 3000 still watch for signs that HP's commitment to the 3000 is flagging. Do you think they invite such disregard for the 3000 at HP's highest levels with their doubts, or keep HP honest about the value of the system?

Unfortunately, a little bit of both. I don't mind saying now, a year later, that much of the impetus for putting together the World's Largest Poster Project at Anaheim last year was nothing more than a meager attempt at trying to redirect some of the bitter complaining that was being done on the internet at the time. Constructive suggestions are one thing, but vitriol is something altogether different. The people who run CSY are as human as anyone -- and listening to a constant barrage of complaints, some with merit, some without, has to be demoralizing.

There is an enormous amount of enthusiasm and respect for the HP 3000 out there in the user community. Doing something inordinately silly, but silly on a grand scale, to express that enthusiasm actually did some good, I believe. A lot of people, both within the user community and HP itself, were smiling and slapping each other on the back when it was all over. That's a much better way to represent the intrinsic qualities of the HP 3000 and the concerns of its users than having people scowling at each other. I'd very much like to see that tradition continue.

You were instrumental in getting Critical Item Update into HP's to-do list for IMAGE. Yet there hasn't been a lot of discussion among customers about the benefits of this relatively new feature. What are the advantages that CIU delivers that an application developer might have overlooked up to now?

The most important attribute of CIU is that it allows you to add indexes to existing databases without having to modify the original applications software. In 1977, a large database was one that contained 10 thousand records. In 1997, a large database now contains 10 million records. Databases are only going to continue to grow in size -- and the only attribute that is going to make extremely large databases possible is that they will be well-indexed. As databases grow larger, serial searches become less and less of an acceptable alternative. Proper indexing is critical to efficient use of a large database. Quite often, though, you don't always know when a database is being developed which items should be made search items and which should not. CIU allows you to be conservative initially and later invisibly add the proper indexes at any time in the database's life.

The second attribute of CIU is that it was a necessary hand-in-glove enhancement for the coming b-trees. B-trees in IMAGE are designed around the hashing nature of an IMAGE master dataset -- and b-trees in IMAGE are both efficient and fast. B-trees are one those enhancements that are going to make a truly significant difference to a great many users. I recently participated in some benchmarking tests of the new NM Query that incorporates b-trees. The performance gain using b-tree generic searches for queries such as, "find me all of the people whose name starts with AN," proved to be 30 to 100 times over an equivalent serial search. It's important to understand that, for most HP 3000s, you can't buy that kind of performance increase, no matter how much money you're willing to spend. But now you're going to get it for free. As a result, I'm sure that there's going to be a lot of added pressure to create new indexes in databases where few existed before.

Why do you think big-payoff ideas like an SQL interface, CIU and b-trees for IMAGE don't emerge from HP's database labs anymore?

I'm very pleased with the rate of progress that the database labs is making. Being completely non-responsive is clearly a perfectly valid way to fail. But being overly responsive, subject to every whim and trend, is as an equally effective way to get into a great deal of trouble. But more than that, it's going to take a little while to get all of these "big ideas" well tamped down. All-in-all, I'm very pleased.

What are the most important developments CSY can focus on for the further improvement of IMAGE?

As I said earlier, cheap has become the operative word nowadays. If a solution is too expensive, it simply won't even be considered. The singularly most important thing for CSY to focus on now is getting the price of a 2- or 4-user HP 3000 down to essentially PC levels.

Ultimately, what we're talking about is the capacity to build personal mainframes (PMs) for small businesses. There is an extraordinary demand for precisely such a reliable data platform out there. We have one customer just a mile or so from us, GenCon Construction, a construction management company, that has grown significantly over the 13 years that they've been our customer, but who, during the same time, has only grown from four to eight people. The office manager has credited their capacity to grow so much volume with such a small increase in people to their use of the HP 3000.

The people at GenCon are smart. And they're observant. They've watched their competition (other Arizona, New Mexico and Texas construction management firms) go through a dozen PC-based systems during the years they've been on the HP 3000, but they just keep cooking along with totally reliable data capture and data manipulation capabilities. To my knowledge, there are only about ten HP 3000s in all of Las Cruces, a city of about 60,000 souls. And we own at least half of them. But there are at least 300 hundred businesses within five miles of me that could profit greatly from having a reliable, robust, easy-to-use, easy-to-maintain PM-like device, connected to a half dozen or so cheap PCs, used more or less as nothing but display terminals.

The second great benefit of having inexpensive, small machines readily available is the saluatory effect that it will have on the user base. We've lost a generation of developers on the HP 3000s simply because the machines have gotten large. Large machines with a thousand users online are psychologically overwhelming. No one wants to be the cause of crashing such a system, thus development stops and the depth of understanding necessary to building applications easily and well very rapidly decays in the user community.

Moreover, large machines give you the impression that developing applications for the HP 3000 is difficult. Nothing could be further from the truth, of course.

There really are only three parts to developing an application: designing the database to best mimic the flow of information within the business, designing the input screens, and writing the necessary reports. Having a lot of small developers actively developing a lot of small projects is the key to getting a lot of small HP 3000s into a lot of small businesses. And having a lot of small developers developing small projects is similarly key to getting some very good new applications onto the machine.

The HP 3000 celebrates its 25th year of service to businesses this year. What should the customers who own a 3000 derive from that kind of anniversary?

Maturity. If you were a business owner or manager, I can't think of a single word that you would want to seek out and celebrate more than a mature solution, one that can easily demonstrate that it can do what it says it does. Immature solutions, on the other hand, are going to cost you an awful lot of money -- and a growing segment of the business community is beginning to understand that. You can only be led down the garden path so many times before it begins to dawn on you what it's truly costing you.

Copyright 1997, The 3000 NewsWire. All rights reserved.