[Iccrg] Meeting agenda (really! :-) )

[Michael Welzl]

> I agree that we don't need another survey! However, as one of the  

Great!


> authors of this survey, would you say that there are general  
> principles that you can state? A discussion on these in this list,  
> and in person during PFLD, would be very useful.

Very hard to do right (as shown by the example that you
quoted from the other reference), but...


> Looking forward, I think that the goal of this discussion would be to  
> use agreed-upon general principles to narrow down the field of  
> potential schemes/mechanisms to consider for a next-generation  
> congestion control mechanism. In particular, given that Linux/ 
> Microsoft are well on their way to making some schemes de facto  
> solutions, if we find that these schemes violate agreed-upon  
> principles, then we should raise this objection with due urgency.

... worthwhile, especially in the light of these recent events.

Stating principles may be easy, but agreeing on some may not -
or they become so general and vague that they are not very
useful for the above purpose (like "a congestion control
mechanism must be efficient").

What I can imagine is that we could agree on some sort of
a checklist for congestion control mechanisms. In fact,
if you look at some congestion control papers from the
last few years, you can tell that such a checklist already
exists, albeit implicitly, in people's (author's and
paper reviewer's ;-)  ) minds...

Like:

A congestion control mechanism should...

* be stable
* not experience a significant performance degradation when
  links have a large bw*delay product
* work well with not only 1 or 10 but 100 or 1000 or more users

etc. etc.

Would that make any sense - would such a list be helpful
at all? Most of the items on it seem like common knowledge
to me... but then again, some may not.

Cheers,
Michael