Hi

This is my last week of work on the hiper side of libcurl for this time. I'm
going back to working on my "actual" job and I intend to get back to full-time
hiper work later this spring/summer.

Changes to the suggested API that I'm about to do:

Remove the timeout value from the callback prototype. We don't want to
time-out on a per-socket basis. Instead, we offer curl_multi_timeout() that
returns the maximum time the application is allowed to sleep before calling
curl_multi_socket() with CURL_EASY_TIMEOUT set.

Internally, libcurl will have a sorted list of timeout values so that when a
timeout action is called, it would only run the action(s) for the
sockets/transfers that actually are affected.

I'm not sure how well this applies to the (high performance) event systems
that might be used with this new API.

(I'm also considering a curl_multi_setopt() to set the callback function
pointer and 'userp' pointer with, to reduce the amount of arguments to the
curl_multi_socket() functions...)

More details are/will be mentioned on the hiper schedule page:

         http://curl.haxx.se/libcurl/hiper/schedule.html

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