No models for traffic sharing

(How do transient flows impact large flows?)

Matt's first TCP question (1991):
 	Half T3 NSF net (22 Mb/s) with 10 Mb/s load
 	Best possible FTP was 5 Mb/s
 	Where was the missing 5 Mb/s?

No theory either
 	(the mice and the elephant problem)

Akin to turbulence

We do not even know the dimensionality of the problem space