News and Review's theme this month is Open Systems. I previously wrote a column on my views of how and why open systems may eventually have a large and mostly positive impact on commercial data processing. To recap, it is my firm belief that today and for the foreseeable future, commercially open systems will not truly develop openness beyond a high level of data interchange abilities. I believe this because of the significant differences between the PC world and the commercial world. It is worth peering into these differences a little closer to understand the market mechanics creating these differences.
The PC world has set some really tough standards for openness. You can generally buy an unbelievably wide range of bits and pieces, plug them together and mostly expect things (both hardware and software) to work. There are lots of little reasons this seems to work so well in general. Many of them were accidents of design or marketing. And many are becoming less and less true by the day. For example, originally, there was only one highly compatible chip family involved in the PC product and until a few years ago, only one bus and one significant operating system available.
This single mindedness created a huge monolithic marketplace. It also made it easy for everyone and their brother to throw together low cost boards, device drivers, and software applications. The combination of price competition and high volume inevitably led to lower margins and good values for end users.
But as the market matures, some fairly predictable things are happening. First, if you'll forgive the exposed commercial biases, while the PC has certainly not been a "toy" for a lot of years, it is still maturing in its reliability and stability. The operative mentality in the PC world -- buy the best pieces from whoever make them today, don't worry about vendor stability (the product will be obsolete in four months anyway), and tinkers are pretty tolerant of incompatibilities, non-existent support and hardware glitches -- just doesn't translate into today's business world. The tolerance for problems which cause system outages (An unexplained system hang, the message "an error has occurred in your windows application, Cntl-Alt-Del to restart", etc.) goes down rapidly as the machine moves from the home to the office, and as it becomes an on-line data repository for others such as a department server. At the same time the number of CPU chips, buses, OS's, middle ware, applications, etc. is exploding.
At some point this combination of increasing complexity and reduced forgiveness causes vendors to do one of several things. They can choose to develop for only a subset of the new community (reducing their market economics), or support the ever expanding target hardware with the same resources (i.e. poorly), or raise prices to cover the same support for more options. Even if they choose the latter, it is inevitably harder and harder to do the QA on schedule and to do it on all the options to be really sure it works in all cases.
The net of these changes is as the PC invades the business world as a business data processing machine instead of a personal machine, it is bound to lose some of its price/performance and some of it's openness along the way. The big boys -- the commercial hardware manufactures, and the heavy duty database developers have been educated in the school of hard knocks for the last several decades. In the business world where your product is critical to the operation of an end users company, your reputation rides on the quality of the total solution. Few individuals in a company will know the last problem was caused by outside vendors -- "the computer is down" is a broad brush that tars the primary vendor regardless of the cause. Also, you can and do get sued over things totally outside of your control. The obvious and rational step then is to ensure as much as possible is inside your control. Do your own QA. No outside hardware boards, device drivers, or OS's. No third party maintenance, no unqualified middle-ware or databases. And heaven forbid, no released source code. In general, no openness if there is anyway you can avoid it! These are the facts of life -- if you are going to deliver a total solution, you seem to need to be in total control.
This was the underlying premise in my previous article regarding business "Open Systems". Hardware and software openness, a la the PC world, get translated by the pressures and exigencies of the business world to "you can have access to my data" but "I still have to control my hardware and software" in order to insure it works correctly and to protect the vendor from misplaced liabilities. As standards mature these issues will obviously begin to blur, but as I stated at the outset, for now in the business world, "Open Systems" has a fairly narrow data interchange definition.
Hardware openness in the commercial world has historically gone up and down. The old days, slower hardware innovation lead to limited hardware openness for at least the largest vendors such as IBM and DEC with add-in memory and peripherals. However, as innovation and competition sped up, it was harder and harder for third parties to design and develop quality products that could recover their investments. Today, there is relatively little hardware openness in the first tier hardware manufactures.
Still, the standards movement is progress for professional data processors. Over the last decade we have seen rapid progress from data exchange only with great difficulty, to easy batch interchange, and now to reasonable interactive interchange. This movement from batch to interactive or on-line open data exchanges is beginning to enable most commercial data processing shops to buy and use equipment, data bases, and even to some extent applications in the environment they are most efficient in and reasonable for. That's a whole lot more "Open" than we have been accustomed to for the last thirty years.
The acceleration of this process will continue to generate a significant improvement in your ability to respond to changing business needs and external requirements. You will be able to purchase full turnkey systems from vendors and simply plug them into a network to exchange data with your current systems and applications. If you can't find a reasonable pre-existing application, you will be able to develop it on the platform (hardware, OS, and database) that makes the most sense, deploying it with less hope and more confidence it will plug and play on the network with your other applications. Yes, this is to some extent a description of the dreaded "data islands" everyone is concerned about. But at least they are your data islands, under a degree of control and oversight, and with proper planning, they can be made to behave rationally at a lower total cost to the organization.
The new world of data processing is providing innumerable new options, but no neatly packaged new solutions. It is still up to MIS management, the application designers, and the programming staff to integrate all these options, select the appropriate ones and forge them into a high quality, reasonable cost solutions. But then, that has been you job all along, hasn't it?
Copyright © 1994, Holland Consulting.