On Fri, 18 Nov 2011, Rob Ward wrote:

> As it seems no one has any information with regards to this I am going to
> begin looking at implementing the first proposed solution as time permits in
> order to see if it is feasible and produces better results. I will post a
> patch as soon as I have any potentially working solution for your
> review/comment.

Sorry, I'm quite backlogged... I think the "issue" to solve is to have the
algorithm not use the entire transfer time as basis when it checks the
transfer speed, but instead perhaps the last N seconds. libcurl already keeps
track of the speed over the last 5 seconds and that might be a start.

-- 
  / daniel.haxx.se
-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette:  http://curl.haxx.se/mail/etiquette.html