Below is the text for a message back in May. Is there anymore information on this problem?
The reason I ask is because of another problem I am trying to find where after a timeout, the next request appears to "fall through" (with no error), and the next request after that gets the data from the second request. Weird, I know, and it may not be in libcurl.

On Mon, 12 May 2003, Dan Halbert wrote:
> I am not sure this is a libcurl problem per se, but may be a network,
> socket, or perhaps webserver problem.
>
> This is a case where trying to reuse an old connection causes an "empty
> reply" instead of an I/O error. Have any of you seen this before?
I have not. I think it looks very odd, as even when doing the second request,
it doesn't discover anything strange.
If the second request returns no data at all, that too is supposed to cause
libcurl to return an error, since all requests to a HTTP server should cause
*some* data to be returned from the server.
> On the client: libcurl 7.10.1 on Redhat 7.2,
Can you try upgrading to a later version and see if you can repeat this
problem?

-- 
 Daniel Stenberg -- curl: been grokking URLs since 1998
-------------------------------------------------------
This SF.Net email sponsored by: Free pre-built ASP.NET sites including
Data Reports, E-commerce, Portals, and Forums are available now.
Download today and enter to win an XBOX or Visual Studio .NET.
http://aspnet.click-url.com/go/psa00100003ave/direct;at.aspnet_072303_01/01