Internet Perspectives Presentation at Memtek's seminar on the World Wide Web and Java, held at Scandic Crown Hotel, March 15th, 1996.	Peter Magnusson psm@sics.se Finger PGP public key

Each image is 256-color GIF (interlaced, format 87a), 800x600 pixels, with individual color palettes, so you should probably have a few thousand colors available for this to look good. The slides were produced using Microsoft Powerpoint 7.0, and the GIF images using Corel Draw 6.0. The slides are also available in (black&white) format inside this postscript file.

Permission is hereby granted to use this material freely, provided that proper credit is given, either to Peter Magnusson and SICS, or to the "source" as indicated in the lower left of some slides.

This version of the presentation is the one given at Memtek's seminar "WWW and Java", at Scandic Crown Hotel on March 15th. It is a snapshot and is thus likely to change, any new versions may be available at http://www.sics.se/~psm/internet_perspectives.html

Special thanks to Seif Haridi, Anders Landin, and Andreas Moestedt for providing valuable feedback on an earlier draft.

Some material on the slides are from the Internet Society, who allow free use of much of their material provided that this credit is given.

The figure of 50000 is from early 1995. (ftp://rs.internic.net/netinfo, I think, I can't get in on their ftp site at the moment).

The growth rate at the time would indicate over 70000 networks today.

The Internet is difficult to define.

General: The means by which all the world's computers are able to communicate with each other. Narrow: The number of networks passing traffic via the major backbones. Broad: Includes application level gateways (e.g., e-mail), store and forward networks, and networks using non-IP protocols

According to (ftp://ftp.nw.com/zone), in january there were 9.5 million hosts. Growth rate is 20% per quarter.

Almost universal connectivity.

Completely uniform standard, allowing products to be completely portable.

Goes against the grain of earlier standards developments, such as TV, electricity, radio frequencies, and which side of the road you drive on.

This data is from 1994, so it is ancient.

BITNET (Because It's Time NET) began in 1981, and is an academic store-and-forward network.

In case there's any doubt as to my personal opinion, I write it in stone.

Speech was our first (sophisticated) form of communication. It allowed us to communicate with one individual or a small group of individuals. The communication was limited in time (duration of talk), reach (those who could hear it), and space (those who were there).

The invention of writing, ca 6000 BC, freed us of the constraint on time and space, thereby laying the basis for civilization. The consumer no longer needed to synchronize with the producer. A message could be read later (time constraint), or sent somewhere (space constraint). We still had a restriction on reach, in that making copies was arduous.

The invention of the printing press in China removed this last fundamental restriction. When coupled with a simpler written language several centuries later, this ushered in the current era of civilization, which we might call modernism.

The 19th century saw several new technologies that multiplied the power of previous communication methods. The telephone removed spacial constraints on communication, and radio and television removed constraints on reach. Today we view these technologies as quintessential of life in the 20th century. Yet their emphasis on few senders, combined with requirements to synchronize in both space and time, left much undone.

The Internet is the next, true stepping stone. It is important to note that Speech, Writing, and Printing could all do what the previous level of communication could (though perhaps with some loss of intonation). The telephone, radio, and television cannot and are, in retrospect, not true steps of improvement. The Internet can, in principle, do everything that the other five media can. And more.

And of course, the digital nature of the internet (of computing in general) is key: any information can be transported, and the format supports meta-information.

So the perception that Internet will have a fundamental impact is based on observing that it carries a bigger change than any previous step since the transition from speech to writing.

Today, in the heady glory of ubiquitous Internet, we're left with the impression that there are no longer any restrictions communication. Of course, this is simply an effect of our limited imagination.

first experiments in 1966

ARPA requested quotations in 1968

packet switching was widely doubted, but was desired for its fault tolerance. One objective was to build a system that could withstand any systematic attack on central nodes, such as from a nuclear strike.

(the van Johnson story)

AT&T was particularly pessimistic!

4 computers on ARPANET in 1969

in 1973 an effort began to connect ARPANET with mobile networks using synchronous satellites (SATNET) and mobile packet radio (PRNET), this effort became known as "internetting"

part of this research effort resulted in Ethernet (in Hawaii)

July 1977 a four-network demonstration linked ARPANET, SATNET, and the PRNET. TCP/IP version 4 came in 1978.

the new internet protocolls, generally called TCP/IP, began to be developed in the early 70s, and ARPANET switched to them in January 1983

supercomputer centers programme in 1986 (Senator Gore's legislation) led to the NFSNET, which remained the backbone until April 1995

experimental electronic mail relay put into operation in 1989, interconnecting MCI Mail with the Internet. Compuserve, ATTMail, and Sprintmail followed shortly

The machine running Alta Vista is the largest machine Digital every built.

Alta Vista sweeps the world every 14 days, approximately.

The WWW was conceived by Tim Berners-Lee in March 1989, in a CERN proposal. It first aimed at managing information about accelerators and experiments at CERN.

The concept of "web" derived from an observation of how new staff at CERN were introduced to the work structure - by being given a few hints on who to talk to. At CERN there is and was a large turnover, several thousand people are involved yet seldom stay longer than 2 years.

Also improves on problems with hierarchical data structures (such as file systems) and keyword based systems (such as bibliographic databases).

"Hypertext" was coined in the 60s by Ted Nelson. A more accurate label today might be "Hypermedia".

Other than using TCP/IP, a key idea is a uniform naming scheme that includes the service name. Also, in HTTP the client can send a list of the representations it understands, and the server reply appropriately.

NCSA Mosaic came in 1992, and was instantly popular. Netscape Communications was formed in 1994.

There is not really any new technology in the Web concept, just a good combination of ideas.

Many claim that the Internet is not a safe place to do business - not true.

Cash are tokens that can be authenticated independently of the user. Cheques require reference to user. Card channels payment through existing card payment systems.

We don't necessarily need a single system, the "real" financial world has multiple, complementary ones. Transactions are already available on the web. The problem is not technology, but achieving critical mass despite legal and political obstacles. Not clear that it will work very well in the foreseeable future.

Credit card companies and banks are exploring the field, including making various announcements, but the fact is that an intereconomy is a huge threat to their livelyhood. Their payment systems are exepensive, with a large overhead per transaction, and preferably linked with expensive (and very profitable) credit. An internet economy with very low transaction overhead would be a huge threat. Immediately, people would get 3-5% rebates for placing mail orders on-line.

International transactions is a big profit source for banks. Digital cash and similar technologies would eliminate this completely. It would also open up completely the market for financial services, such as loans and fund managements. Already small financial institutes are beginning to look at the Internet as a way to reach a mass market without a branch in every town.

Personally, I use First Virtual to make donations, and Netscape "secure" protocols to purchase with credit card, including things like hard disks, SIMMs, etc. First Virtual's method is particularly neat, relying on the Internet's size for protection, and using telephone keypad to update credit card information.

In fact, the way I got the rights to using the Dilbert cartoon (omitted on hand-out versions) is a good example. I first emailed United Media, but, getting no response, two days before the presentation I e-mailed Scott Adams and complained. He promptly replied with a cc to a suitable person, who quickly replied. I payed with credit card (faxed via our internal intranet) and downloaded the image of the net. 45 dollars for the rights of using it once.

Sales in 1995 indicated clearly that home users preferred high-end machines, in fact high-end sales were dominated by home customers. One of the reasons for Apple having problems was that they assumed that home owners wanted small machines (and they still do).

Apparently home owners either find synergies in being able to take home work from the office, or the PC has simply become like the automobile - expensive but perceived as being worth it.

Developments in computer architecture are also fairly clear. Computing power continues to fall in price faster than bandwidth,and, in combination with the theoretical limit on latencies, this would indicate that we want to do more processing at the nodes, not less.

Finally, the market for increasingly powerful video games shows a big demand for graphics performance.

In reality, the vision of the set-top box is just a lingering desire for centralized systems, since they're easier to envision, and also easier to make money on.

When I demonstrate internet surfing, I do it on a high-end workstation that costs around 100000 crowns. This is not going to be sold for 500 dollars anytime soon.

Low end is not enough.

From a Gateway add. This type of machine is already being announced.

Basically, a 120 Mhz Pentium PC is connected to the TV and stereo. It forms a platform for home entertainment, education, communication, and work.

Problem is very much in the close association of services with infrastructure today.

If you let the same provider define services and deliver the infrastructure, they will invariably screw up the infrastructure part if it is not associated with services that they make money on.

Since digital data is at the lowest level untyped, it doesn't make sense in the long term to sell customers anything but bandwidth. The cost of routing packets is so small that to carry any accounting information along with them, in the sense that is done with telephony, is silly,

Of course, today they're hamstrung. If they sell me 10 Mbit home at-cost, I could just resell telephone lines to the other 475 flats in my area.

Information on the net today is generally untyped. For example, when I'm searching for information on the Amazon car, I can access Altavista or similar global index, but I need to define my constraints semantically. Thus, I don't want information on the djungle, so I try to include in the specification things like (not "djungle"), etc. I can't ask "who has an Amazon car for sale" or "who services Amazon cars in Stockholm".

The problem is a lack of metadata. The digital format is unique in its ability to represent information about information, to any recursion.

Middlemen will use any and all technologies available, including "agents" (a better term is robot).

Netscape product offers a possibility of an OS independent set of services.

Soon we'll see products saying "requires such-and-such a computer, memory, video, and Netscape 2.x".

It is the byte code that is the portable platform, not the Java language. There is already work on, for example, Ada-to-Java bytecode compilers.

We will *not* get a significant threat to monolithic applications from ones built up from Applets. This is still a research problems, with essentially nobody knowing how to do it. The problem is primarily in the interface design (which slows development) and the overhead of modularity. Neither is solved by Java.

We need to remember that significant technology advances in the marketplace invariable are preceded by research prototypes in academia. Take GUI and RISC as an example.

Netscape packaging is reasonable: Java, Javascript, TCL/TK, plug-ins.

Some uses of Java are just plain silly, such as for multimedia (much better to use plug ins) or scrolling text (idiotic reuse of a one-to-many metafor)

This example Java applet, from C-Net, is an excellent demonstration.

You interactively move the sliders to indicate you preferences - how you value performance, ergonomics, etc. Then you press Submit, and the applet queries a server with information on hundreds of PCs. The result is a selection of a few choices, plotted by price and "value", as defined by your preferences.

You can the click the dots on the plot and get more information on each machine.

The applet is highly interactive, with hot pop-up zones and a cursor-sensitive help box at the top right. This typ of computation is not realistic to do remotely, and it is here that Java excels.

Names, not addresses.

This is not a good group to have against you.

there used to be hype about 500 channels coming from sattelites, the problem now is that we have litterally hundreds of thousands of channels, and the content is extremely varied. This frightens people.

Singapore orders internet providers to censor materials on sex, religion, and politics

in China, all users must connect through a government ministry, and existing users must register with the police

belgium is considering mandatory key escrow

in Zambia, the only internet service provider (Zamnet) is threatened with police action unless they remove the on-line version of "The Post" (the editors were imprisoned). But not before US sites had put up copies.

in France, Dr. Claude Gubler (Pres. Mitterand's personal physician) wrote a tell-all book called "Le Grand Secret". A french court censored the book, but a scanned copy made it onto the Internet. On-line efforts quickly translated the book to English.

With single platform, I mean not only single internet, but also everybody being able to at least run Java programs (or something similar).