This, said Dr. Aho, is a simplistic example, but over the past 20 years engineers have made Moore-like improvements in software feature verification. A problem that used to take seven days now can be done in seven seconds. Lucent applied this technology to a recent product—the telephony features that were verified using this system—and took the specification for these services, extracted from them properties that were written in linear temporal logic, and then in a matter of minutes verified that the code satisfied the properties of this kind of specification. A goal is to extend this technology to software more broadly because the ability to produce reliable software on a global scale will be essential to a robust economy.
As businesses come to depend on the interoperable infrastructure, such questions will become crucial to a sustainable economy. In general we will need an open, global network infrastructure that will be able to host these services. The Internet must become much more user friendly in terms of its interface devices.
He mentioned a visit to the Georgia Institute of Technology for the inauguration of a broadband institute, where he talked with a professor who was designing shirts embedded with computers that could monitor human vital signs. One such design was inspired by a mother who had lost her first child to SIDS and wanted to prevent a second tragedy. The professor made a baby-size shirt with monitoring equipment and a wireless link. Other applications might include monitors for extreme athletes who are trying to push the envelope of what humans can do or systems to help take care of elderly parents so they can live at home longer.
He reserved judgment as to whether there will ever be a network experience as natural as a face-to-face conversation and concluded with a proposed test of such an experience: Would you want to propose to your significant other over the network or in person? “That,” he concluded, “is my test for the future.”
SOFTWARE: THE CHALLENGE TO GETTING THERE
Daniel T. Ling
Dr. Ling said he would talk about the supporting technologies that are helping the software industry, some exciting new applications driving the software industry, and some of the difficulties in writing software.
In terms of supporting technologies, he said that several participants had discussed the good news at the center of the network, but that toward the edges the situation is not quite as rosy. Within a building, for example, wireless LANs and Bluetooth have a high bandwidth, but wireless wide-area applications such as cellular satellite do not have high performance. For residences there is some penetration by ADSL and cable modem, but he said that, for example, ASDL is not available at his own house.
Dr. Ling mentioned the concerns about the network effects of Metcalfe’s
Law21 and an even stronger network effect related to the number of communities that can be formed in a network with n participants. He suggested that this effect might be between n2 and 2n. Referring to magnetic storage, he said that Michael Lesk, now at the National Science Foundation, has studied these effects,22 as have Hal Varian and Peter Lyman at Berkeley, 23 and attempted to calculate how much information is produced in a year. They arrived at a number of 2 exabytes24 per year. At a few dollars per gigabyte, the expense of storing all the information produced a year online in magnetic storage would be only a few billion dollars. He said that considering Dr. Aho’s imminent transmission rate of petabits per second, this amount of data could be transmitted quickly as well, which he found quite amazing.
Dr. Ling then raised the topic of microelectromechanical systems (MEMs), saying that MEMs receive the benefit of all the semiconductor technology and will transform the information technology industry. He described the MEMs-based optical switch and MEMs-based displays made by Texas Instruments and other companies. The main idea behind MEMs is to be able to make inexpensive sensors and actuators, including displays, to bring computing into a closer tie with the real world. He cited special MEM devices that could measure structural changes in buildings after an earthquake, describing which sections have been damaged and which have not, and MEM devices for the body, mentioned by Dr. Aho. He said that MEM technology has the potential to bring dramatic change to information technology.
Batteries Do Not Obey Moore’s Law
Dr. Ling turned to technologies that are not increasing exponentially, especially batteries. Many optimistic scenarios are based on the availability of many mobile or isolated devices, all of which require batteries. The amount of power density that can be packed into a battery has not risen fast and for lithium ion batteries will soon approach a theoretical maximum power density of 5652 kJ/l. Portable applications, such as radios, need a certain minimum amount of power. Display applications need a minimum amount of light to be visible to the human eye. Audible signals need a certain volume of sound. In other words some requirements for power do not scale.
The Need for More Research on Writing Software
A second worrisome category of technology is writing software. Productivity has improved, partly through increased computing power, which brings more tools and makes the programmer more productive. Higher-level, object-oriented languages also make a difference, as do improved programming environments. Especially helpful in making programmers more productive are large, reusable components that include tremendous functionality and can be used repeatedly. An example is an operating system. In the past, operating systems were able to support disks and printers, but today they provide a number of services that programmers do not have to deal with, such as the windowing system, the graphical user interface, the TCP/IP (Web) protocol, and the HTTP (browser) protocol. Operating systems increase productivity, as do databases and Web servers, and these increases are experienced indefinitely.
Clearly, however, this area also calls for research. Dr. Ling recalled the PITAC report25 and the more recent NRC report,26 both of which mentioned the importance of software, in particular, the difficulty of making the very large-scale distributed systems that are fundamental to the advances envisioned today for IT. These systems are enormously complex and will require multiple cooperating firms to build. They also have to be secure, reliable, and able to enforce privacy. Once these functions go onto the Web they must be able to resist attacks and probing by hackers, which is an enormous challenge.
Some Pre- and Post-Internet Software
Dr. Ling said that just for fun he had listed a half dozen drivers of software and divided them into pre-Internet and post-Internet categories. In the pre-Internet category are computation and simulation, which were primary motivations behind the invention of computers. Such uses have transformed science in many ways, some of which can be seen in bioinformatics, genomics, molecular biology, and astronomy. He cited the Sloan Foundation’s funding of the Sloan Digital Sky Survey, which has spurred significant discoveries by comparing astronomical data from different points in time. Environmental science, engineering, and design have been transformed by computing. CAD models were used to design the Boeing 777, demonstrate the aircraft to customers, and simulate repair scenarios.
Other pre-Internet drivers include back- and front-office automation. Herman Hollerith, one of the pioneers of IBM, first used computing largely to do the census and moved from there to payroll, billing, and other basic applications. These came to include databases, which are still a hot area, now often called data
warehouses. These data warehouses allow a firm to analyze all transactions, to separate products that are selling from those that are not, and to decide what to buy and what to recommend. Enterprise resource-planning software allows firms to do their payroll, personnel, and other routine processes within the company in a single system. It includes customer relationship management and supply chain management. Finally, personal productivity software got its start with the personal computers as the classical spreadsheet. Word processing applications went on to functions such as diagramming, personal financing, and mapping software. More recently the pre-Internet technologies have moved to the “PIN” area of personal informational management, e-mail, calendars, and contacts.
Data Anytime, Anywhere
As we move to the post-Internet drivers of software, we see a dramatic change in that people begin to expect their data anytime, anywhere through a variety of wireless devices. The World Wide Web started with physicists who wanted to share information, so that browsers were designed to weed and find information. This was exciting for two reasons. The first was that it provided worldwide access to information and allowed people to share data with no gatekeepers. This alone was a remarkable change from the old days of closed, proprietary, vertical systems, such as Lexus and Nexus. This has developed into a push for large digital libraries and e-books that can be carried anywhere.
Another big transition has been the integration of other pieces of technology into the Internet function, such as GPS, to produce a new range of services. For example, agriculture—a very basic, old industry—has combined GPS with information technology to change tremendously. Knowing what seeds, fertilizers, and pesticides they have put into small plots of land, farmers can now track the yields of each plot by satellite.
Then came the first version of e-commerce, where customers go to Web sites such as Amazon, eBay, and Priceline to engage in some transaction. The auction site eBay is interesting because auctions are an ancient market mechanism that has suddenly become practical on a large scale because of the Internet. Priceline raises the question of whether the fixed price is passé.
The Digital Convergence of Media
Another step ahead is the digital convergence of media. The initial version of this has been streaming sound and video that comes over the Internet. The personal video recorder has the possibility of transforming the broadcast TV industry. The concept of prime time loses importance when you can record, program, and view it at your convenience. The differential values of those slots of time become uncertain. The Internet has also allowed companies, such as Amazon.com, to
learn a great deal about the behavior of their customers, which was very difficult in traditional retailing.
The Napster case shows that distributed, anonymous, peer-to-peer systems have arrived, whether we like them or not. Dr. Ling cited Napster as an interesting example, because it used centralized directory services, which created an entity that could be sued. Other distributors are not centralized and there is no one to sue. We must take this into account, he said, when we think of commerce in the next generation. The systems are here—anonymous, encrypted, completely decentralized.
Software as a Service
Version 2 of e-commerce, he said, will feature machine-to-machine commerce and less human-to-Web-site commerce. He cited a prediction27 that online retailers will become less integrated as their individual functions—customer contact, warehousing, billing, and shipping—are sliced off. New firms will be constructed from these slices that will operate by machines talking to machines. Customers will go to a service they trust that will recommend the products and they will go to another service to pick up the products. They may even fill the shopping cart from companies throughout the Web. New marketplaces will form with new market mechanisms such as compilatory auctions. New ways of measuring and maintaining a reputation will evolve.
Dr. Ling concluded by pointing out that software is changing from a commodity to a service. He said that researchers are concerned about their ability to quantify the amount of money going into software if people were to rent software that is tailored to their needs rather than buying shrink-wrapped boxes of software designed to broad standards. “Software as a service,” he concluded, “is definitely the trend.”
The Potential for Collaboration
Finally, Dr. Ling addressed the area of communications and collaboration, a “big, new application for computing” that started with e-mail and is moving on to AOL’s instant messaging and presence. There is enormous interest in additional conferencing and collaboration tools to allow remote collaborators to join as virtual work groups. Many firms are designing annotation functions to allow for “living documents” created by widely dispersed people who are able to make comments, ask questions, or post annotations. These annotations can all be searched and classified to distinguish private annotations, for example, from public questions from the audience or exam questions or answers to questions. All such tools have the ability to transform our educational process.
S. Jurvetson. 1999. “From the Ground Floor,” Red Herring, <http://www.redherring.com/mag/issue72/news-groundfl.html>, November 1.