Mohun Biswas wrote:
> I still think this would be improved by saying "libcurl will then
> transfer any data the handles in the stack are ready to deal with
> without blocking". Or something to make clear that curl_multi_perform is
> designed not to block. As I understand it avoiding blocking is the
> entire point of the multi API, so I can't see any harm in mentioning it
> even if you don't want to talk about sockets per se. Currently the only
> mention libcurl-multi.3 makes of "blocking" at all is down at the bottom
> in the caveats where it lists a few things that may block. But nothing
> is said to clarify that outside of these corner conditions, it will not.
> This is the central cause of my confusion.

Daniel,

This point seems to have gotten dropped. Do you disagree with it? If so
I'll let it go, but I'd still like a ruling on whether my concern makes
sense.

Thanks,
MB
Received on 2009-07-14