[tor-talk] What are use cases made unpleasant by Tor's speed?

[Virgil Griffith]

The variability is def a big deal, from a UX perspective, maybe even larger
one than raw speed---I can relate to the occasional Pyongyang circuits.

Out of curiosity, are there stats on the average speed (error bars would be
nice) over the past few years?  Just curious.  This would be helpful.  Im
sure that overall speed is improving.  But it'd be nice to check if it's
improving at the same rate typical webpages are fattening.

For addressing latency/variability, I know the  KIST scheduler
addresses some of this.  One of the first things slated to funded is to get
the KIST socket scheduler merged.

-V

On Monday, July 28, 2014, Griffin Boyce <griffin at cryptolab.net> wrote:

> Hey Virgil,
>
>   I'd say that the issue with Tor's speed is one of inconsistency, rather
> than outright slowness.
>
>   Give you an example:  I was watching a friend set up a new laptop, and
> she elected to fetch drivers direct from the manufacturer's website.  So
> someone downloads a few dozen files with the Tor Browser.  Pretty common
> stuff.  She ran into issues of speed -- at one point all pending downloads
> failed -- but for the most part the speed ranged from 200kbps to 1mbps.
>  There are circuits where the connection slows to Pyongyang levels, but for
> the most part Tor is pretty solid.
>
>   Tor excludes relays that have less than X speed over N period of time,
> but perhaps the threshhold for X should be raised and length N should be
> lowered.  My theory is that this would solve many of the "speed issues"
> that people experience.  Perhaps try simulating these differences with
> Shadow and report back your findings?
>
> best,
> Griffin
>