[mosh-devel] Impressed

[Hari Balakrishnan]

Mosh gathers RTT samples and calculates a smoothed estimate using the same method as TCP.

In your experience how commonly deployed is DiffServ at the edge hops?

Do you know whether 3G/LTE networks honor these codepoints?

Hari

On Apr 13, 2012, at 8:05 AM, Dave Taht wrote:

> I created a tiny fork of mosh on my github repo to play with
> classification a bit.
> 
> I think that AF42 has the combination of desirable features of both
> CS4 and IMM.
> 
> https://github.com/dtaht/mosh
> 
> after doing the commit, I went and checked, and the linux kernel does
> when checking for IMM, masks out the other CS* bits, so it jumps the
> queues in pfifo_fast, and Linux wifi checks only the CS* bits, so this
> tosses mosh packets into the underused VI hardware queue.
> 
> For good measure I tossed in ECN marking... I note that there was a
> bug with ECN and pfifo_fast we fixed last year (2.6.39?) that tossed
> it into pfifo_fast's background queue.
> Various add-on shapers may twiddle differently, as well.
> 
> I'm low on time this month to actually do anything with this
> scientific myself, but the 3 line patch is out there if anyone else
> wants to try it.
> 
> 
> is there a tool to measure RTT inside of mosh?
> 
> 42 seems to be my lucky number.
> 
> -- 
> Dave Täht
> SKYPE: davetaht
> US Tel: 1-239-829-5608
> http://www.bufferbloat.net