On Wed, 15 Nov 2000, John Casu wrote:

CC'ing this to the curl mailing list for information.

> i) Is it possible to set the connections to be non-blocking, so that I
> can increase the rate at which I am sending request packets, and reading
> requests ?
>
> When I compare the results for ab to my benchmark for http (non ssl)
> requests, I get considerably higher transaction rates (%20 faster) for ab
> than for mine.
>
> I guess I *should* do this myself, but I figured it might be a
> generally useful feature.

I've planned to add non-blocking connections. I even have a mail in my inbox
with an example how non-blocking connections are made in windows, which would
also introduce better connect timeouts on that platform.

It is only that same old problem: I'm loaded with work and ideas and I would
need help to speed up things.

I'd welcome a patch that introduces this, any time. I just don't know when
I'll be able to write it myself.

> ii) curl_easy_setopt(handle, CURLOPT_STDERR, null_file), where null_file
> points to /dev/null doesn't work. I have to freopen("/dev/null", "a+",
> stderr) to get the desired effect.

... hm, and this is because:

> after each curl_easy_perform(), a crlf is sent to stderr, and this is
> kind of annoying. CURLOPT_MUTE doesn't fix this either.

Right?

It is an obvious error, I just committed a fix to lib/progress.c that removes
that extra newline.

Thanks for your feedback!

-- 
      Daniel Stenberg - http://daniel.haxx.se - +46-705-44 31 77
   ech`echo xiun|tr nu oc|sed 'sx\([sx]\)\([xoi]\)xo un\2\1 is xg'`ol