On Tue, 7 Feb 2006, Sun Yi Ming wrote:

> My basic point is that we can have a way to run those handles which don't
> wait events without change the current multi_runsingle() or
> multi_perform()'s implementation, but as you (and as i) said before, making
> multi_perform() perform as much as possible until it does block on sockets
> is also a good idea, it's up to your decision.

For some reason I like my idea better! ;-)

But don't expect *me* to implement something like that anytime soon. I have
work way up above my ears now and for the forseeable future.

-- 
  Commercial curl and libcurl Technical Support: http://haxx.se/curl.html