Collaborations‎ > ‎

EScience Lecture

Supervisory Control, Ecommerce and XML

May, 2001
Web Robot Demonstration

A lecture presented in May 2001 for the ANU eScience diploma. I present some experiences with internet services and observations of the internet industry. The story provides a context for a more thorough investigation of some aspects of Ecommerce, human behaviour on the internet and web technology. The aspect of web technology given the greatest emphasis is XML.

Links to other EScience Lecture pages

In The Beginning

Ken Taylor May 2001

     There was no web...

The story starts with the beginning of the web in 1994.

  • 1994 internet largely unheard of.
  • Mosaic early 94.
Mosaic what's new button showed almost every new web page worldwide for a while.
  • Almost exclusively static sites. One only in the University of Western Australia.
  • Perhaps a half dozen database backed sites. A single live web site - the Cambridge coffee pot.

  • September 94 it's working.

  • A team in the US puts a web robot up also.
  • The world hears of the internet.
  • Almost anything "internet" gets in the news. - A cybercafe starting in London is news worldwide.
  • Telerobot features in Saturdays West Australian, appears on television here, ABC network in the US and lots on the radio in Australia and overseas. A paper presented in Singapore wins a major robotics award. It's even still remembered in the Bangkok Post in 2000.
  • An unreal and heady time when a few weeks work can produce something that attracts major attention.

Things Take Off

Ken Taylor May 2001

     And people turn their minds to money...

The initial euphoria attracts all sorts of people and the take off begins.

  • Over the next few years there is rapid growth, everything is new.
  • The variety of ideas astounds everyone.
  • Even search engines didn not exist for quite a while.
  • Some of the Mosaic team start up Netscape and after a public float are suddenly very wealthy.
  • Virtually all commercial internet sites lose money but rapid growth hides this.
  • Capital flows freely into the internet companies.
  • There is a feeling that with a large number of people using your service there must be a way to extract revenue from them.
  • Advertising is the obvious choice. It is the revenue source in many a business model promising to generate revenue without the unpleasant task of having to extract money from your visitors.

I do the sums on the robot. For the first eight months of 1998 the telerobot delivered an average of 32,439 operator and observer pages per month. This usage could generate an income of US$243 per month from advertising. Our single robot could only be used by one person at a time so at flat out could have produced no more than US$385 per month. More trouble than its worth.

  • I discover Zipfs distribution and give up on advertising.
  • Zipfs law also applies to referring sites.

As seen above, the Perth telerobot did show an approximate Zipf distribution of referrers but the discrepancy between the sample and the Zipf distribution is not due to sampling error. This is shown by the chi squared statistic of 5707 with 774 degrees of freedom which indicates that the probability of the difference being due to random sampling error is almost zero. Apart from the problem of classifying referrers, it was also observed that referral numbers changed over time. The telerobot was used by some school groups and was mentioned in the press from time to time. This causes short-term changes in the number of referrals from a particular address. This effect was strongest when the telerobot featured in radio shows Net Talk Live (1995) in the United States and Safari (Heldal 1998) in Norway. The largest number of requests to the ABB1400 telerobot recorded in a day (1782) occurred on the day the telerobot was featured on Net Talk Live. Referring sites also gain and fade in popularity over time which will affect the number of referrals they generate.

The uncertainty of categorising referrers and the change in referrer numbers over time suggests that Zipf's law does not provide an adequate explanation for referral data however, it does seem to provide a useful approximation.

Boom On

Ken Taylor May 2001

     Investigating behaviour...

The commitment of most people is low and their behaviour is largely unpredictable.

  • Registered operators only controlled the telerobot for 8% of sessions or 15% of the time.
  • Very few people will wait 3 minutes.

  • From 31 January 1997 to 15 April 1998, 1,295 different registered operators used the telerobot. How often did they come back?

  • Demographics 95% male and young.
  • Watching people learn.

  • Weibull Distribution
  • Behaviour can't be predicted.

Reality Bites

Ken Taylor May 2001

     There was no web...

Wwile trying to work out how to generate income from web services and examining user behaviour I was watching the explosion of commercial offerings and wondering what they knew that I did not. Then...

  • Some internet companies start to run into trouble.
  • Burn Rate How I survived the Gold Rush Years on The Internet was published in mid 1998.
  • "The money gets bigger. The schmoozing doesn't stop. Today's height of success might be tomorrow's valley. The landscape changes even as you climb. And the precipitate is always narrow, and the ground far below. We were on the brink of disaster, just as we stood on the brink of success..." (Michael Wolff, Burn Rate).
  • "I used all the tricks:... Nothing worked."
  • 'Burn Rate' author sees severe Web turbulence ahead.
  • Market Capitalization with Poor Fundamentals.
  • Little connection between internet the world and the world of finance.
  • What they knew that I did not was that the real customer was the investor.
  • What had been required was a plausible business plan and ever larger capital raisings.
  • However now that we all knew the reality the game would end so back to looking at the fundamentals.


Ken Taylor May 2001

    Seperating Information from It's Presentation...

Telerobot operators could design their own interface.

Presenting information differently.

Madness Sets In

Ken Taylor May 2001

     No connection between dot coms and reality...

Wwile the reality of the dot com world had been well documented the boom proceeds to new heights.

  • Technology was irrelevant from this point.
  • Only marketing mattered.
  • Took to reading prospectuses and marvelling.
  • 4 Corners covered this well in This Time it's Different
  • Then it ended.

Web Services

Ken Taylor May 2001

     It is still changing the world...

A return to rationality is a great thing as there is no room for technologists in a pure marketing game.

  • Consumers have captured most of the benefits.
  • Good for humanity but bad for your company.
  • Micropayments. Why haven't they happened?
  • Teleopereation and shared control has a place when a problem is to hard to fully automate which is most of the time with robotics.
  • Proposed Helpmate robot.

  • Actual Helpmate robot.

  • Will reproduce for the physical services what call centres are doing for information only services.
  • Sharing equipment.
  • Ultaspecialisation. XML is an important enabling technology.
  • Application service provider. Again XML is an important enabling technology.
  • Doesn't need big organisations. Empowers individuals and enables atomisation.

More XML

More XML Excercises

Ken Taylor May 2001

     Some more things to try...

  • Schemas and DTDs.
  • Write a translation for converting these lecture notes in wml for viewing on a WAP device
  • Extract SMS messages sent from the database. Translate them into the format required for sending over the air.
  • Look at the translation for plotting SVG bar graphs. Then devise a transation for xy graphs.
  • Define a schema for over the air updates of a phone book.
  • Prepare a translation that maps this schema to the base schema.