There are certain truths to the statements you are making,
but the conclusions you arrive at from those threads of
fact are incorrect/flawed. I am still doubtful about its
theoretical correctness leave alone its
* complexity
* lack of requirements
* duplicating existing functionalities(e.g.ToS routing.)
Just go through *http://www.cs.bu.edu/fac/matta/Papers/jsac95.ps*.
They do for two TOS classes: delay and throughput in a non-FCFS
scheduling environment. you are suggesting for n- classes.
The paper concludes based on simulations and analytical results
using liapunov functions that there is some reduction in end-to-end
delay.
So for just coming to one conclusion, the authors have done lot
of simulation and analytical work. But it is unfortunate that you
are making sweeping conclusions without a valid proof.
I will reply detaily later today.But before that clarify
these points,
- you all accept that diffrout is not going
to solve any problem.
>-> Thus said, coming with a new scheme is not a big task,
>-> problem space and applicability statement is very
>-> important.
> I agree.
- It is complex.
>HH:> There is never a free lunch. But what are a view routing
>tables compared to the probably much bigger number
>of p2p LSPs?
This archive was generated by hypermail 2b29 : Mon Aug 04 2003 - 04:10:04 EDT