Re: [TERNLI] Notes from tonight's ad hoc

[Gorry Fairhurst <[e-mail]>]

Joe Touch wrote:

> 
> Aaron Falk wrote:
> 
>>On Jul 28, 2006, at 5:56 AM, Mark Allman wrote:
>>
>>
>>>  + Giving a generic signal that *something* has changed and it likely
>>>    matters and so the transport should re-init itself (e.g., forget the
>>>    SRTT, RTTVAR, cwnd, ssthresh, MTU value, etc.).
>>
>>Perhaps the signal should be that congestion state or RTT estimate is no
>>longer valid.  I.e., make the signal have transport semantics.
> 
> 
> That is what I was afraid of - that means that the link knows what the
> network is doing, e.g., whether it is using the link for CBR or bursty
> traffic, etc.
> 
> Joe
> 

It isn't clear to me yet that the idea of transport semantics is useful...

If say, the link has a substantial change in RTT: this obviously effects 
the RTO in TCP (how much depends upon how much the total path RTT 
changes), the signal could impact what a sender stores to calculate the 
SRTT -  perhaps it also means that you may have to change other things 
(such as cwnd), but links don't really know the implications on the 
transport ...

Gorry