In many subjects that have had to deal with complex systems, the traditional and usually successful means of response has been to make and use predictive models of the systems. These subjects have included computer science.
The temptation when dealing with the Internet is therefore to try the same approach. But while it is possible to make models, the evolution of the Internet is so fast - including evolution driven by the consequences of people's actions that are prompted by what their models say - that modelling has no interesting predictive value.
The basic argument of the seminar is that the Internet is so unruly that it has to be tamed by measurement. It is also big, so making the right amounts of the right kinds of measurement is hard. What to measure, how to measure it, and what results exist and may exist in the future, are the expected outputs for the seminar audience.