I will try when I get a chance, but this only happened about 1 out of every
200 attempts, so its little hard to reproduce.
  Lucas.

-----Original Message-----
From: Daniel Stenberg [mailto:daniel_at_haxx.se]
Sent: Sunday, October 21, 2001 11:04 PM
To: libcurl Mailing list
Cc: Lucas Adamski
Subject: RE: Missing Timeout?

On Fri, 19 Oct 2001, Lucas Adamski wrote:

> While testing HEAD requests, I would occasionally observe that libcurl
> would return a "transfer too slow" when I'd set a low speed limit=10, low
> speed time=10 even though the actual transfer took less than a second.
> I'm guessing its still checking the low speed limit after a HEAD request,
> and if there was a slight latency and the theoretical speed for the HEAD
> request is less than 1/s, it returns an error even though the transfer
> did not take 10 seconds. This never happened for a regular GET request.

This should indeed be considered and viewed as a bug.

Could you set a breakpoint in Curl_speedcheck() when you do this and try to
see which values that seem to be wrong that make that condition trigger
even
though it shouldn't?

Right, the header data is not counted as true data and thus do not count to
download speed, but that shouldn't make anything happen until after the
"low
speed time" anyway!

--
    Daniel Stenberg -- curl groks URLs -- http://curl.haxx.se/