This paper will discuss the current state of affairs with the much hyped network computers, that industry leaders such as Microsoft, IBM, Intel, Oracle, and Sun are aggressively pursuing. It will discuss the evolution in desktop networking that has lead to the NC concept and the various viewpoints within the industry on the exact form the NC will take by looking at recently announced products from the industry leaders.
Then, the paper will discuss some possible future applications of the NC with respect to networked multimedia, assuming an installed base of 250 million units in 2007, and 500 million in 2012.
The Evolution of Desktop Networking
There has been a clear evolution in the past twenty years with desktop networking devices, from dumb terminals, to personal computers, and then to the client-server paradigm. At each stage, the problems associated with that particular paradigm became evident, which lead to the concepts of the next stage.
In the early 1970’s, networking was centralized, where all applications, data, and computer processing was done on a central system, which was generally a mainframe computer. These systems were accessed via non-programmable terminals (NPT’s), which were also known as dumb terminals. NPTs were very simple and reliable, both for the end users and the administrators of the network. However, they were very limited in functionality, as the centralized applications were generally low on features and only provided text-based interfaces. NPT’s were also proprietary, so that once a vendor was chosen, the customer was locked into whatever solutions that vendor had to offer. This proved to be inflexible and expensive. Dumb, centralized servers also proved difficult to scale to large numbers of users that were geographically disperse.
About 10-15 years ago, the personal computer emerged as the solution of choice for desktop networking. They offered much more flexibility than dumb terminals by empowering users to control their own computing resources. Applications, data, and compute cycles all resided on the PC, so choices and preferences were under the complete control of the individual user. However, PCs proved to be complex to setup, configure, connect to a network, and to maintain. The initial cost of a PC turned out to be just a fraction of the total cost of maintaining the device, and administration was very difficult. Also, since the data was stored locally on each device, the process of backing up data had to be handled separately by each user, and company wide access to the data was nearly impossible. Security and administration of such disseminated data was quite difficult.
About five years ago, the client-server model of desktop networking emerged as a popular solution. In this paradigm, data is centralized in a server while applications and processing power are decentralized and remain on the desktop computer. In this way, the business still has administrative and security control over their data, yet the individual users can still benefit from customized applications, along with the better performance associated with compute cycles on the desktop device. However, the cost of these desktop devices is no less than that of a PC, and after initial purchase price, administration, and maintenance costs are taken into account, the annual cost of a so called ‘fat client’ may range from $7000 to $12,000 (7).
Enter Network Computer’s (NC’s)
With the problems associated with the three previously discussed paradigms of desktop networking, there was a clear need for a better option. Enter Network Computer’s. Companies including IBM, Microsoft, Sun, Oracle, and Intel, have recently announced their plans of developing a new breed of desktop computer, that will offer benefits of both the centralized dumb terminal model, the personal computer model, and the client-server model. The new devices offer the attractive price of NPT’s, along with the flexibility of personal computers with decentralized applications and processing power, and the benefits of administration and security of centralized data repositories. These devices are called Network Computers, or NC’s, and their concepts were first put forth by Oracle (8).
With a network computer, the ‘fat-client’ desktop PC is replaced by a ‘thin-client’ device. This means the expensive, complex hardware and software of PC’s is replaced by a simpler hardware device that gets its applications and data from the network on an as needed basis. This approach has several advantages:
Little Desktop Administration: Network computers have very little administration compared to regular personal computers. Since the hardware is simpler, and the software applications and data reside on the network, if something goes wrong, the device can simply and cheaply be replaced. Since the software is not stored locally, any administrative tasks can be performed centrally.
Improved Security: Since data and applications reside on a centralized network server, access to both can be carefully controlled.
Better Economics: The initial cost of a network computer is much less than a PC, but perhaps more importantly, the long term costs are also much less. Estimates of annual costs of ownership for PCs, including hardware and software maintenance, can approach $12,000 (7). Yet with an NC, where the software maintenance nearly disappears, costs could be as low as $2500 (7).
Simplified Application Distribution: Since software is stored centrally, and only downloaded to an NC when necessary, users will automatically get the latest version of every application once it is installed on the server. No longer will there be several different versions of one application spread throughout the network, which will reduce maintenance and support costs.
The network computer paradigm ensures that the devices are built to comply with the current Internet standards, including HTML, Java, audio and video file formats, etc. All of these standards are stored in ROM on the NC, or downloaded from the network as needed. Perhaps the most important component is the Java programming language, which was developed by Sun Microsystems. Programs written in Java can run on any computer that has a Java interpreter installed. This allows software developers to write a single version of their application that will work on all platforms. As Java becomes an Internet standard, and since the network computer will adhere to all of the current standards, developers will be able to create applications that can run on any network computer.
It is unlikely that NC’s will replace any of the devices used in the three paradigms of centralized dumb terminals, personal computers, or client-server networking. Instead it will come to be another solution in certain situations. “To us, it’s not about network computers replacing PCs, it is the right technology for the right problem,” said Phil Hester, vice president of network computing for IBM. “We have been through three phases of computing, (….) and none of them have replaced the ones that came before them. (3)” We will see when the NC is the right solution below.
Now that we have seen the general concepts of network computers, we can take a detailed look at several of the major vendor’s versions, as although each follows the general principles, each also has their own visions.
IBM’s Network Station
In October of this year, IBM introduced its version of a network PC, which it dubbed the Network Station. This device matches the general concepts of the description for NC’s above very closely. It will initially be targeted at businesses, where the current cost of installing and maintaining a regular PC over the life of the device can often range from $7000 to $12,000 (7). IBM estimates that their network PC can save 50 to 70 percent of those costs (5).
Also, applications and data are removed from the desktop and put on the server, where LAN administrators can securely control and backup the data, and easily administer the applications from one location.
The Network Station will have a full color, graphical interface. It will have a built-in full-function Web Browser, so that users can connect to the Internet, or to their own companies private Intranet. The device will come with 8 megabytes of RAM (expendable to 64), network adapter cards (either Ethernet or Token Ring at first, but eventually twinax and coax will also be available), serial and parallel ports, and a monitor port. It will also come with a keyboard and mouse.
Instead of having its own internal disk storage, it will plug right into a network and connect to one or more servers to receive the necessary applications and data. Once connected to the network and powered on, the network station will contact a designated server, from which it will download the necessary system code and a signon screen. Once the user’s access has been authenticated, the server will send the station software based on the users configured preferences.
The network station should be available during the fourth quarter of 1996, and the base configuration should cost about $700 US, with no monitor, and $1000 US with color monitor.
Sun’s JavaStation (1) (2)
Like IBM’s Network Station, Sun’s JavaStation will have not have expansion slots, and no hard, floppy, or CD-ROM drives. It will be about 8.5 by 11 inches and weigh in at four pounds. A fully configured system with 8 megabytes of memory, along with keyboard, mouse, and 14 inch display will sell for under $1000. Sun estimates that the yearly cost to own and administer their device may be as low as $2500, versus the $7000 – $12,000 of regular PCs.
Sun’s approach to the Network Computer, however, is slightly different from that of IBM. Whereas IBM is targeting a replacement of current desktop PCs, Sun will target major corporate customers with fixed-function applications, such as airline and hotel reservations, kiosks, stock trading, etc. The main difference is not necessarily the machines’ hardware configuration or how the machines will function, but how they will be marketed to corporations.
Microsoft and Intel’s NetPC (4)
Also in October of this year, Microsoft and Intel announced their plans to develop and market a network oriented PC. Their approach, however, is somewhat different from the basic concepts of a network PC first put forth by Oracle, that were discussed above, and that IBM’s Network Station and Sun’s JavaStation both closely followed. Microsoft and Intel plan to maintain the basic components of a personal computer, by using an Intel microprocessor, Microsoft’s Windows operating system, memory, and hard drives. However, they plan to eliminate the user’s ability to add software via floppy diskettes or CD-ROMs. Instead, software updates would have to come over the network from a server. By retaining hard disk drives, unlike the original NC concept, data and applications can be stored locally. The only real difference between the NetPC and a regular personal computer is that application updates will have to come from the network. However, the NetPC took it another step, by announcing “Zero Administration” software that they will build into the operating system, which will automate the process of such updates. Microsoft and Intel stated that rather than focusing on the entry price of the device, they are choosing to focus on the real savings which will be in the maintenance the device.
Their take was that the original concepts of an NC would be too limiting for many users. As Microsoft Group Vice President Paul Maritz stated, “We believe that by taking the cost out of the PC while maintaining its flexibility will reduce the cost of (managing PC networks) without taking a bet on one single platform that in the long run might prove to be more costly.” The one platform he refers to is the Java programming language, that IBM and Sun are betting on heavily.
Microsoft has yet to discuss the prices of their NetPC, but suggest that it will be slightly less than a comparable desktop PC of today, which would typically be in the $1500 range.
NetScape and Oracle (3)
NetScape and Oracle, two other major players in today’s computer and network industry, have also announced an alliance. The original concept of the NC came from Oracle, though the plans for their devices have not been announced as concretely as IBM and Sun. Though Oracle originally maintained that they are not a hardware provider and only planned on licensing their concepts (8), that has apparently changed (3). Again, the specifics are not available, but Oracle has agreed to use the NetScape Navigator as the web browser in the Intel version of its NC, which it expects to ship next year. NetScape will also offer Oracle database software for its server packages used in corporate intranets.
We can see that all of the industry leaders are involved in the network computer concepts and are planning hardware and software to facilitate their growth. But will the concept be successful?
Will Network Computers Succeed?
There are clearly problems with the Network Computer vision, and these issues will have to be overcome for their deployment to be successful. Such problems include economical issues (can the NC be developed for the low entry cost initially announced) and technical issues such as bandwidth, security and privacy concerns.
The vision of the network computer is to provide nearly the same services as a normal personal computer, without the headache of software installations and upgrades, along with a low price. Originally seen as a $500 device, we saw above that IBM’s Network Station would sell for $700 with no monitor and $1000 with a monitor, Sun’s JavaStation with a monitor for under $1000, and Microsoft and Intel’s NetPC in the range of $1500. Clearly these are well over the $500 originally planned (though they are still cheaper than regular PCs).
One of the problems with the price is the fact that the speed of the processors necessary to run today’s applications, and the quality hardware necessary for good sound and video, are both very expensive. A Pentium processor costs nearly $200 dollars just by itself. Memory prices have come way down, but still add up when many applications today require 16 megabytes or more. Also, quality monitors and audio devices are not cheap. As we have seen in the past 20 years in the computer industry, and as George Gilder’s book “Microcosm” showed, processing power continues to grow and prices continue to drop. As these trends continue, the price of the hardware necessary for a network computer of good quality will drop, so that we will see the $500 network access device.
Technical difficulties with the network computer concept must be overcome for the devices to be successful. These include bandwidth limitations, software bulk, and security and privacy issues.
Perhaps the biggest issue with Network Computers is the lack of necessary bandwidth. As Tom Halfhill, from Byte Magazine, stated, “a device that drinks most or all of it’s software from a network really needs a big straw” (9). Current software, which in the network computer paradigm will be downloaded from the network, is quite large. As just one example, Microsoft’s Internet Explorer is over ten megabytes. With today’s top dial up speeds of 33.6 kilobytes per second, it would take much too long to download such large software applications often.
For home users, until the next generation of network connections catches on, the network computer will be nearly useless. Once the telephone companies introduce ADSL, which is already in pilot in several areas of the country and provides access speeds of 6-7 megabytes per second, or once cable modems become popular, which will offer 10 megabytes per second or more, then maybe the NC will be viable in home uses. However, we know from George Gilder’s book “Telecosm,” (10) that tremendous bandwidth is on the way. For now, however, the network computer may only be useful in the corporate environment, where LAN’s provide transfer rates of 10 Mbytes and higher.
In addition to the sheer size of software, in which bandwidth will play a major role, is also the ‘bulkiness’ of most operating systems and applications. By this I mean that software that was developed to run on conventional personal computers is bulky, and expects to run from a hard drive on the local device. This paradigm must basically be thrown out the window for applications that will run on NCs. More streamlined versions of applications will be necessary in order for network computers to run them well. NetScape, via their subsidiary Navio, is already pursuing a stripped down version of their web browser, which will be more tailored to a network computer (3). Microsoft’s Windows operating system will also be streamlined to better suite the NC device (3).
Java will play a key role in many of these issues, as the applications will be written once, and will be runnable from any platform. Such programs have evolved from a networking paradigm where transmission of the software was key, so they were originally written as streamlined as possible. This will have to continue with Java applets written for network computers.
Security and Privacy Issues
As data moves from a local computer onto a network server, security and privacy issues leap to the top of the list of concerns that many people have with network computers. The benefits of putting data in one location are clearly seen when looked at from a corporate perspective, as backup and security policies can more easily be enforced and implemented. However, some data, even within a corporation, is highly sensitive and should be kept securely away from all employees except those with a need to access basis. This is difficult to enforce when the data is kept on the network.
Also when looked at from an individual home user’s perspective, their personal data is now on a network, and is maintained by a service provider. Users will have to trust their providers both with keeping the integrity of the data (i.e. not looking at it and using it in any way), and on the safeguarding of the data (i.e. keeping others from looking at it and using it).
Oracle’s vision of the network computer has security built in via the use of smart cards (8). The NC will have a smart card interface that will be used to identify and authorize the user. The smart card looks just like a credit card, but will also contain a small secure chip that stores information about the user. The smart card will employ a Personal Identification Number (PIN) security system, just like ATM cards, or some other secure method of identification. This will help prevent unauthorized use of the computer.
The card will be able to store user preferences, and information can be written to in addition to being read from the card. This enables the card to “remember” certain things about the last time the user was logged on. In the case of the home user, the name and number of their Internet Service Provider (ISP) and which applications or are used most often could be stored. For the corporate users, it can provide address information about the network servers where the user’s applications and data reside. Also, encryption keys unique to the user can be stored on the card, so that network servers can verify with certainty the identity of the user and therefore provide secure access to the data on the network.
To increase security on the card, certain strict processes can be followed. For instance, protocols of reading from and writing to the card will be strictly enforced, and if they are broken the card can be erased. Also, if the wrong PIN is entered too many times, the card can be erased. This may be extreme for the normal home user, but for corporate users who work with sensitive data, it may be necessary.
Short Term versus Long Term NC Success
Even with all of the technical and economical issues discussed above, the network computer will be successful in the long run. In the short term, the NC is already viable for many areas, including corporations, education, and developing countries, as discussed below. However, it may be several years before the NC is successful in the home, which is also discussed below.
Viable for Corporations
Corporations that have extensive LAN environments with multitudes of computers are sure to benefit from the use of network computers. Most corporate users use their workstations for tasks such as e-mail, databases, spreadsheets, word processing, and Internet access. All of these are well suited to the network PC paradigm were the device gets its data and applications from the network. In this case, all of the applications will reside on servers, and with LAN speeds capable of downloading the information quickly, the NC will be useful.
“Market research suggests that the NC can potentially replace up to 50 to 70 percent of the PCs now in place in corporations. “(8) And since personal computers cost so much to maintain because of administration, support, and training (estimated to be from $5000 to $8000 per year per PC (8)), corporations can potentially reduce such costs by 70%. Initial acquisition costs will also be much less (under $1000 vs. the typical $2500-$3500 for today’s computers). The NC will be able to peacefully coexist with today’s personal computers, so the transition will not be a one day event, and for those power users who will still need the benefits of a regular personal computer in the future, they will be able to connect to the same network as NCs.
Viable for Education
Currently, the hardest obstacle to overcome in getting students connected to the network is the initial acquisition costs of personal computers. The second most difficult task is finding the necessary skills to install and maintain complex personal computers. This is especially prevalent since most schools, in an effort to overcome high prices, use a wild mix of used computers, which make the schools face problems of inconsistency and compatibility issues. Since the network computer has a much lower entry cost and lower maintenance costs, and because problems of software upgrades and configurations nearly disappear, schools should benefit immensely from this technology.
Viable for Developing Countries
Developing countries are similar to schools in that the most difficult obstacle to overcome to get their population online is the initial cost of computers. These countries will be able to reach full participation in the world wide Internet community much more quickly by using network computers instead of personal computers.
Viable for Home Users in the Long Run
For home users, bandwidth and security will remain issues too hard to overcome in the next 3-5 years for the NC to be successful. With access limited to 33.6 Kbs right now for most users, there is no way software and data can be downloaded quickly enough to make the NC work. Unless the devices is used as a ‘dumb’ Web Browser only, with the browsing software built into ROM, the device will not be beneficial. Many individuals who do not want to spend the several thousand dollars necessary to buy a personal computer may be willing to pay several hundred dollars for such a dumb browsing device. However, this does not fit the original vision of a network computer that is flexible enough to run various applications.
In the Long Run the NC will be a Success
In the long run, as all of the issues discussed above are overcome, the NC will be a success in many ways. However, it will not replace the desktop solutions of today, including dumb terminals and personal computers. Instead, it will be a new option worth considering in particular situations. For corporate users who are looking to simplify workstation administration and support, the simplicity of a network computer will be terrific. For users who are looking to get on the Internet cheaply, such as home users without the skills necessary to run a personal computer or cost-conscious schools, the device also will be beneficial. However, there will still be those applications and those people that will require the flexibility and power of personal computers. Both personal computers and network computers will continue to be viable products for a long time to come.
Future Applications of the NC
Considering that the NC will be a viable and successful product per the discussion above, and assuming an installed base on the order of 250 million units in the year 2007 and 500 million units by the year 2012, we can discuss a few possible future applications.
For such lofty predictions on the number of installed devices, it is obvious that all of the problems considered above, especially limited bandwidth, will be overcome. Bandwidth is perhaps the most important aspect to consider because of how much audio and video sequences consume. For the following discussion, we will assume that we have enough bandwidth to basically consider it infinite, for both wire and wireless network computers. We will also assume that privacy and security issues are overcome, enough so that people will feel comfortable leaving their personal data on server which resides on the network somewhere.
With these assumptions in mind, we can consider future applications of the network computer. However, specific network applications most likely will not be much different from those offered today over regular personal computers, such as e-mail, Internet Web browsing, groupware, and so on. Yet the network computer, with its simplicity, will show up in many places computers currently are not in, and in some forms may evolve into the ‘teleputer’ that George Gilder so often refers to in his writings. So with these thoughts, lets discuss a few applications.
From the discussions above, we saw how an NC will be a great replacement for a PC in many, but not all, situations. Corporations will be able to save money on the initial cost and over the life of the device. Home users not willing to pay out major dollars will be able to obtain the benefits of getting on-line. Schools will also be able to get more students on line faster. For these situations, the NC basically replaces today’s PC, and retains the same functionality as a network access device that gives e-mail and Internet capabilities, word processing and spreadsheet applications, etc. to its user.
Since the NC is such a simple device that is easy to set up and connect to the network, it may merge with today’s television set to become a teleputer. A teleputer is a device that mixes the best of both a computer and a television. An NC as a teleputer would allow users to watch television, in today’s broadcast form, where the programs are sent to everyone at once, and tomorrow’s narrowcast form, where individuals order the programs they want to see, when they want to see them. The NC-teleputer would also allow this for radio signals, in addition to television signals.
The device would also act as an Information Highway access device, so that it could be used to read e-mail, browse the Web, etc. (Note that WebTV, announced earlier this year by the company of the same name, is offering a set-top device that allows a television to do this. We already witnessing the merging of the two technologies.) The teleputer will also allow audio or audio-video calls to take place, so it can become a video phone.
Fixed-Application NCs (FA-NC)
The recently announced Sun JavaStation is targeting fixed application devices. These will be nearly ubiquitous in the future. Currently, personal computers, dumb terminals, and cash registers are performing the tasks the FA-NC will perform in the future. Why spend the money on a full blown computer when all that is needed is a network application device? And why waste valuable processing cycles on a mainframe when they can be done on an NC. Hotel, airline, and rental car reservations are just a few areas where FA-NC’s will be better suited for the tasks than the devices being used today.
The NC can act as a smart cash register, that sends and receives information to the network, such as purchasing and inventory requests which is currently done. However, it can do much more. For instance, if a store is out of a product and a customer wants to order it, the NC could quickly download a simple application from the network that allows the clerk to fill out an order form quickly and easily.
FA-NC’s will also be useful as Kiosks, where consumers simply walk up to a booth to perform some simple task, such as ordering merchandise, voting, etc. The FA-NC will send the entered information onto the network to the application. But unlike today’s Kiosk’s where the application is generally placed on them once, and if changes are necessary someone has to go out to the device to make the changes, the FA-NC will be able to download the changes from the network at anytime.
Public Network Access NC (PNA-NC)
As people become more comfortable with the concept of “the network is the computer,” which means they will feel more at ease with their data and applications residing on the network, the Public Network Access NC will become popular. The PNA-NC is similar to today’s phone booths, that charge users for connecting to the telephone network. However, the PNA-NC will give access to the Information Super Highway.
Users will be able to signon to the device and connect to their home server to access the applications they prefer and the data they need. It will not matter what type of PNA-NC people use, since the standards are all open, you will be able to quickly download the applications and data to the public device, anywhere, anytime. The PNA-NC will likely be very successful in libraries, hotels, and airports.
It will also be important as companies send more employee’s home to telecommute because of office space limitations. Then, when a worker does have to go into the office, it won’t matter what office they are in, as the PNA-NC will always give them the same interface to the network. Or for employees that travel from one company location to another, the PNA-NC will again give the same applications to the user no matter what device they signed onto.
The network computer concept seems to be the next logical step in the evolution of desktop networking solutions. In the first stage, dumb terminals were simple yet inflexible, and processing power, data, and applications were all centralized. In the second stage, personal computers, which were much more complex than dumb terminals, decentralized processing, data, and applications. Next was client-server where much of the complexity of personal computers remained, but data was centralized and applications and processing were local. The network computer is a simple yet flexible device with the all of the benefits of keeping data and applications centralized, but with compute cycles local.
The network computer has some obstacles to overcome, such as the bandwidth limitations of today’s networks, and the privacy and security issues involved when data is stored centrally, but this will happen in the next few years. The network computer will be wildly popular in corporations, education, developing companies, and for many people who want to access the network but can’t do so because of the cost and complexity of today’s personal computers..
In the future, the network computer will be used for many applications, such as personal computer replacements, teleputers, fixed application devices, kiosks, and public network-access devices. For computer replacements and fixed application devices, the main benefits are cost savings. For teleputers, the benefits are the simplicity of getting the masses connected to a global information and entertainment network. And for public network access devices, the benefits are that a user no longer has to worry about getting to and working with their data via different interfaces, or worry about the data being accessible at all. Instead, the information is available anytime and anywhere, and the interface to that information will remain constant.
The Network Computer is the next logical step in the progression of computer networking. It will not replace the current paradigms, but instead join them, and it will bring simple, inexpensive networking to many people, places, and applications that the current solutions simply can not do.
(1) “Sun Micro Launches Its Own NC attack.” Oct. 29, 1996. URL available at http://www.yahoo.com/headlines/961029/tech/stories/java_1.html.
(2) “Sun Has Low-End NC.” Oct. 30, 1996. URL available at
(3) “The Battle Over Network Computer Gets Uglier.” Oct. 29, 1996. URL available at http://www.yahoo.com/headlines/961029/tech/stories/nc_1.html
(4) “Microsoft, Intel Announce Plan For Cheaper PC.” Oct. 29, 1996. URL available at http://www.yahoo.com/headlines/961029/tech/stories/msintel_1.html
(5) “Announcing IBM’s first network computer: IBM Network Station.” Oct. 1996. URL available at http://www.internet.ibm.com/computers/networkstation/
(6) “IBM’s new desktop network computer for business.” October 1996. URL available at http://www.internet.ibm.com/computers/networkstation/whitepaper.html
(7) “JavaStation.” White Paper. October 1996. URL available at
(8) “Oracle NC White Paper”. URL available at http://www.oracle.com/products/nc/html/wp.html
(9) Masson, Jim. “Network Computers.” URL available at http://ra.colorado.edu/advisors/masson/NetPC.html
(10) Gilder, George. Telecosm. To be published this year.