2008/4/30, Daniel Stenberg wrote to Pierre Ynard:

> I can see a use for your idea, but then it should _only_ call the progress
> function when actually and truly a signal was caught. IMHO.

And I insist that calling the progress function from inside wait_ms(),
Curl_socket_ready() and Curl_poll() even _only_ when a signal has been
caught will create problems at some points, and that the library users
more likely to suffer them will be those that don't use the progress
function to abort the operation.

I really think that calling the new callback instead of the progress
function is the best option even if it is only called when a signal
has been caught. Obviously this doesn't mean that finally it _has_ to
be done in this way.

Disregarding any specific implementation details, it would be nice to
know the real demand for the fast or improved abort capability,
besides Pierre's obvious interest. But I suppose there's no good way
to get this info, and in any case its value is just relative.

Best wishes,

-- 
-=[Yang]=-