Dan Fandrich wrote:
> On Fri, Oct 12, 2007 at 10:54:42AM +0200, Daniel Stenberg wrote:
>> I can't but to feel that it seems utterly silly that this layer of code
>> would need to know about and adjust to the transfer layer's MTU etc. I mean,
>> sure 1460 bytes is for the typical TCP over IPv4 ethernet, but the
>> lib/transfer.c code really is completely unaware and non-caring about
>> that...
>
> It would be useful to test whether this is actually source of the speed
> problem.

I don't think those odd packets can cause a large speed problem. The worst
case seems to me when the useful part of the MTU is BUFSIZE -1, and it still
takes the same time for each packet: that would nearly double the time.

Ivan reported a factor of 500/70 (~7)...
Received on 2007-10-12