Re: [Tsvwg] Input for draft-paxson-tcp-rto-00.txt

[Reiner Ludwig <[e-mail]>]

Scott, thanks for your reply. That makes it clearer to me.

In that case I have two questions. (I assume 4.4BSD as the de facto 
standard for TCP implementations.)

- (2.3) Why is RTTVAR computed before SRTT? Although, I don't have a 
problem with that, the 4.4BSD code computes SRTT first. I assume there is 
good reason but I think it should be written down.

- (2.4) Although I read the TSVWG minutes, I still wonder why there should 
be a *fixed* RTO minimum of one second? One second is way conservative on a 
path with RTTs << 100 ms assuming the sender can benefit from a timer 
granularity of less than 500 ms (e.g., 10ms as in some OSs or even 1ms).
I would have expected a minimum of 2 * G (timer granularity) which is also 
what 4.4BSD does because of the heartbeat timer running out of phase. On 
the other hand, not every OS might implement a heartbeat timer. Beyond 
that, I like what FreeBSD does: MIN-RTO = R' + (2 * G), i.e., having the 
last RTT measurement be part of the lower limit for the RTO.

///Reiner