Re: New I-D revision: TFRC with sender-RTT estimate

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Lars, -
| > It is a chicken-and-egg problem. In its current form, the TFRC implementation buys
| > no compelling performance advantage over using UDP. 
| 
| Performance will never be the reason for people to chose TFRC or DCCP over UDP
| - having congestion control by definition means that there are situations in
| which you will loose performance. The reason for choosing TFRC or DCCP will be
| the desire for a well-reviewed rate control scheme to avoid having your app
| cook up its own home-grown scheme.
|
I see your point. Application developers perhaps think in a different dimension than
network engineers, but if the performance of the scheme in the real world interacts
negatively with application performance or leads to a deterioration of perceived
performance, the acceptance will be low or not existing. 
Your point is good, to me it seems however that both requirements (well-reviewed by
network engineers and well-reviewed by application developers or actual testers) are
necessary.


| > There are other problems in using TFRC, the present problem is that the receiver
| > often gets the estimated RTT wrong.
|
| Understood. But would it really matter to you whether you took these fixes to
| the TSVWG rather than to the DCCP WG?

Not at all. You comments have helped to clarify the situation, thanks a lot indeed.


[Index of Archives]     [Linux Kernel Development]     [Linux DCCP]     [IETF Annouce]     [Linux Networking]     [Git]     [Security]     [Linux Assembly]     [Bugtraq]     [Yosemite]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Linux SCSI]     [DDR & Rambus]

  Powered by Linux