[Iccrg] TCP requirements

[Michael Welzl]

On Wed, 2006-10-11 at 23:20, Lachlan Andrew wrote:
> Greetings Keshav,
> 
> On 11/10/06, S. Keshav <keshav at uwaterloo.ca> wrote:
> > If scheme A is fair to all users, and scheme B is unfair, but every user
> > with scheme B gets more than the same user with scheme A, would you
> > prefer B to A? I would.
> 
> Very good point.  However, most concepts of fairness (like max-min
> fairness and utility maximisation) already eliminate that possibility.
>  They define all "fair" points to be on the "Pareto-optimal" boundary,
> where it is impossible for all users to improve simultaneously.  A
> point where everyone gets equal rate isn't necessarily "fair".

But a real mechanism only approximates one of these fairness
concepts. For practical evaluation of "fairness" of a congestion
control mechanism, one tends to neglect this Pareto optimality,
i.e. the rates of flows are usually compared without taking
the "distance" from the optimum into account.

Thus, in practice, you can have the situation that Keshav
describes above...


Cheers,
Michael