Data Interchange Standards
Robert L. Marx

Today we are going to talk about standards. The Internet is awash in standards. There are standards, claim-to-be standards and wanna-be standards throughout every aspect of the 'Net. It is the foremost goal of every proprietary product to become a standard. We will talk about standards, what it means to be a standard and who cares what a standard is. We will talk about what standards do to help us and what do they do to hurt us. We will have three examples of standards to focus our attention.

We will discuss something called the ISO open-system model, which provides a way to think about and talk about some standards. I will give an example of standardization that is within the experience level of everybody here, even if you are just starting on the Web and the Internet. It is the eating standard involving the knife and fork. Finally, we will talk about what you can expect and not expect from this panel.

Someone once said there are two kinds of people: people who believe there are two kinds of people and people who don’t. There are also two kinds of standards: formal standards and de facto standards. A formal standard is endorsed by a standard setting organization, which has clear processes for adopting standards, keeping them current, and deleting them when they are no longer current.

Some formal standards are national. The American National Standards Institute or ANSI is perhaps the best-known American standards organization. For example, ANSI published the standard, ASCII, which describes how to represent the alphabet within a computer, and also the standard for transmitting fingerprint-based information over a network.

There are also international formal standards. The American National Standards Institute, which runs the ANSI standard program, is a member of another organization called the International Standards Organization (ISO), which is an agency of the United Nations. They have the authority and the mechanisms to formulate international standards, which are becoming increasingly important. The ISO is setting standards that affect the future of Internet and World Wide Web.

De facto standards are those not subjected to approval processes but which feature many of the characteristics of a standard, such as stability over a long period of time, careful documentation, and utility to a large number of people.

I will now display my first prop. This is an ordinary printer cable. It illustrates something about standards. One end of this cable is called the DB 25 connector, described by a formal ANSI standard. Everything about this connector is very carefully defined: the diameter and the pitch of the screws, the spacing and diameter of the pins, what wire carries what voltage with what meaning.

The other end of this same cable is called a Centronix connector. It is a de facto standard. It was standardized, if you will, by a printer manufacturing company called Centronix Corporation. This standard still exists even though Centronix Corporation no longer exists. Why did the standard survive the death of the standard-setting company? Because it was widely used by many other people. Because it does a job effectively and cheaply. Because it is as good as it needs to be and people didn’t need to invent its replacement.

There is one final peculiarity and it shows, I think, the flexibility of standards. The standard on one end of this cable calls for 36 pins to link up to 36 wires. The standard on the other end calls for 25 pins to link up with 25 wires. One of the great mysteries of modern life is, where are those other 11 wires? The fact is: nobody cares.

Why have standards? Standards and standardization carry both benefits and drawbacks. The most powerful benefit of standards is that they allow open systems. Standards are, in fact, necessary for the existence of open systems. Although we may not be able to offer a precise definition for an open system, we instinctively recognize one when we encounter it. Its parts and the interfaces between parts are well defined; there are multiple vendors for hardware and software parts; users and integrators can assemble systems using parts from multiple vendors; when changes occur there is a concern for backward compatibility and a clear migration path to the future.

The world looks a little different from the provider's point of view, and that leads us to the biggest drawback of standards. Standards are thought to slow development by building a box outside of which vendors dare not move. If a standard is perceived as preventing people from doing bigger and better and more wonderful things, the standard is considered bad. A related drawback is the broadly held perception that standards slow future development by fostering a commodity mentality. If everybody is building the same system, why not buy the cheapest one? (The answer is that standards, by and large, do not deal with quality. You can buy a high quality or low quality open system, and you will know the difference over time.)

In the real world systems are neither perfectly open nor perfectly closed (proprietary). An interesting example is the marriage-of-convenience between the Intel processor and Microsoft Windows. Intel processors and Microsoft Windows are closed, proprietary products, but each has clearly defined interface specifications which hardware and software makers can use to assemble open systems incorporating these closed pieces.

On balance, I think most systems people would argue that the benefits of standards outweigh the drawbacks. When a company says, "We exceed the standard," it really means "We have violated the standard." There may be legitimate reasons to do that. If the violation meets a real need and is clearly defined, it may be incorporated into the next release of the standard itself.

What are the characteristics of a good standard?

The first one is clarity. Can you understand what the standard is saying? The standard we are going to be looking at in just a minute, the ISO open-system reference model, is, to my mind, the most obscurely written standard I have ever read. I have read it many times, and I am never quite sure whether I have the English version or the French version. And yet, because it is so important to have a common vocabulary to discuss open systems, this standard is very widely used.

The second characteristic is generality. Do you remember the two ends of the cable that I held up previously? The reason those standards have persisted for so long is that they have a great generality. The DB 25 plug, the first one I held up, was originally designed to carry serial messages one bit at a time. It is now widely used to carry parallel messages.

Stability is the third characteristic. The examples I noted previously have lasted for a long time. Why do they last for a long time? Because they fulfill a real need and they don’t become irrelevant. The best standard-setting organizations, such as ANSI, have ways to identify and purge standards that are no longer in use.

The next characteristic of a good standard is backward compatibility. Requirements change and environments get richer. Can the standard flex to handle new technology, new insights and new user requirements? Some standards are more flexible than others. The more flexible standards are the ones that tend to persist in the marketplace.

The Final characteristic I will discuss is flexibility. Can the standard do a "Release 2" of the standard which handles "Release 1" stuff well? And can it move on to "Release 3" and expand its operational utility without breaking? That depends a lot on how the standard is formulated and how it is structured internally.

We now shift our gaze to the open system model. There are seven layers to this model, conventionally numbered one through seven, from bottom to top. The bottom layer is called the physical layer. The physical layer is made of stuff that you can kick, paint and measure with a voltmeter; very earthy kinds of stuff. The Centronix and DB-25 connectors I mentioned earlier deal with this layer.

The next two layers are called the data link and network layers, and they are concerned with describing the message elements (e.g. alphabetic characters) and message formation (Local Area Network operations). The Transport Layer performs error checking of data received from the network layer and sends it to the Session Layer, which understands how to read a transaction record and treat it as an entity. The Presentation Level handles encryption, formatting and further error checking, before handing off the data to the application level, which is the word processor or spreadsheet or criminal history system or web site or whatever.

Why do we care about any of this? Because the seven layers represent the ways in which multiple vendors can contribute to open systems which can grow and evolve gradually in response to changing technology and changing user requirements.

The other members of this panel are going to talk about real, honest-to-goodness standards used in day-to-day processes in criminal justice systems. They will describe the forward edges of criminal justice: Internet, intranets, extranets, Web sites and all the other buzzwords we live with. I, on the other hand, am going to talk about the application of eating dinner, using the framework of the ISO seven-layer open system model.

The "Eating System" is a de facto standard rather than a formal one. It is a national one rather than an international one. Here is my second set of props: a set of Chinese chopsticks and a set of Japanese chopsticks. The Chinese ones have a round cross-section and the Japanese ones a square cross-section. This is the physical layer of our open-system eating model.

The physical layer influences the diet of these two great nations. Japanese people can eat slipperier food than Chinese people can, because the square cross section bites into and holds the slippery food better. So the selection of the physical layer influences the choices that can be made in the other, higher layers. Isn’t that amazing? Hold on to that thought for a moment.

Picture yourself sitting down at Tony Roma’s House of Beef. You get the 24-ounce Porterhouse steak and you look down next to your plate, and you see - Japanese chopsticks. Oh-oh. You either need a different physical layer, like a knife and fork, or you need a changes in the higher layers, like cutting the meat into small pieces before it is brought to the table.

Let’s assume that you have changed to a physical layer of knife and fork, and consider two other great nations - the United States and Great Britain. Companies produce the same knives and forks and ship half of them to the U.S. and half of them to the U.K. Both start with the fork at the left side of the plate and the knife on the right. Both pick them up and cut a piece of food with the knife while holding it in place with the fork. Americans then set the knife down, transfer the fork to the right hand and use it to move food to the mouth, whereas the Brits continue holding both implements throughout the meal.

We are getting into the transport layer now. The Brit holds the fork tines-down, spears the piece of meat and transports it to his mouth. It works. He takes a vegetable, cuts a piece off, spears it and eats it. It works if the vegetables are overcooked, but not if the vegetables are crisp, so the English diet has been affected by the transport layer. (This may also explain why there are relatively few British people in California, where it is difficult to get overcooked vegetables.)

I will tell you one last story about the difficulty in the transport layer and how it affects the entire national life of Great Britain. The British love peas and they serve peas with nearly all their meals. They tried to get the peas to sit on the backs of their tines-down forks but it can't be done. The British, who are very inventive people, decided they had to do something about this, so they invented mashed potatoes. They put little dollops of mashed potatoes on their forks and they push the peas on to them and stick them into the mashed potatoes. This is my final proof that standards matter.

Things we don't notice way down at the lower levels still affect us. Standards matter, and, although they are hidden inside your system, they dramatically affect what your system can and cannot do. The Brits could have changed the physical layer, perhaps using a spoon to get at those peas, or they could have given up peas, or they could have altered their transport layer to the tines-up American standard, but instead they invented mashed potatoes.

There is a difference between a standard and a proprietary specification. A standard has a sense of openness that doesn't fit well with propriety. In this panel, we are not going to focus on database content and we are not going to focus on the look and feel of the application layer of the system. We are going to look at how one system sends data to another system. In many modern systems, it is hard to say whether you are crossing a system boundary or a subsystem boundary, because one system can reside on nine different computers (as it does in one that I am working with now) or one computer can host nine different systems.

Sometimes communications is required between dissimilar databases and dissimilar applications. That is where SEARCH is focused today. The phrase we are using is event-oriented inter system data system transfer. SEARCH is focusing its attention on two areas at this point:

1. We are attempting to describe the messaging requirement between all elements of a criminal justice information system at the county level: police, jailer, court, prosecutor, supervision agencies. We have formed a project committee, and are beginning to describe what events in the criminal justice process give rise to the creation of and need for inter-agency shareable data.

2. We are also designing an interstate rap sheet format with a transmission format for eventual adoption by all the states, the FBI and the National Law Enforcement Telecommunications System. This will allow interstate criminal history records to be sent around the country in a transmission format. Each state can then translate the transmission format into a page layout or screen layout, to provide its own look and feel to the data. This project is nearing the data test scheme, which will probably occur in the second quarter of 1998.