Thanks Daniel,

We're currently investigating in these directions trying to work with a
smaller pool of handles 3, 5 or 10..

-- Peter :)

-----Original Message-----
From: curl-library-bounces_at_cool.haxx.se
[mailto:curl-library-bounces_at_cool.haxx.se] On Behalf Of Daniel Stenberg
Sent: Mittwoch, 13. August 2008 11:42
To: libcurl development
Cc: Andreas Wurf
Subject: RE: Multi Interface

On Wed, 13 Aug 2008, Peter Thiess wrote:

> I'm a bit surprised that the multi interface wouldn't be able to
enhance
> transfer rates since this could increase bandwidth.

Oh, you mean by connecting to the same server with multiple connections
and
thus avoiding/working around evil "traffic shaping" that service
providers
impose? (On a "normal" network connection do don't gain any speed by
doing
that stunt.)

Yes sure that it can do, but that's not thanks to the multi interface
but more
thanks to you doing multiple simultaneous transfers. You could easily
use
multiple threads and use the easy interface in each thread and
accomplish the
(almost) same effect. I'm not saying you shouldn't use the multi
interface,
I'm just a bit surprised about the mentioned reason.

Do you still get these problems if you only do a small amount of
connections
(like 3) to that server? If not, what amount of connections do you need
to get
the problems? Do the problems always occur (the same way) if you run the
same
68-connections-against-the-same-server test on other FTP servers?

What if you try the same code against a HTTP server, does that show the
same
problems?

What does the FTP server logs tell about the transfers? Does it think
they're
all fine and dandy?

I've not seen these problems before.

-- 
  / daniel.haxx.se