Subject: Re: Enhanced buffering made faster uploads!

Re: Enhanced buffering made faster uploads!

From: Daniel Stenberg <daniel_at_haxx.se>
Date: Sun, 24 Oct 2010 23:27:27 +0200 (CEST)

On Sun, 24 Oct 2010, Aris Adamantiadis wrote:

> Your implementation gave good results (only 50% slower than openssh) because
> you were running in on localhost. Doing it on a high-latency connection
> (like a 100mbps path between europe and US) would give catastrophic results.

I'm fully aware that more round-trips makes it worse, and I've repeatedly
mentioned my thoughts around further improvements regarding that - even in
this very mail you responded to now. My changes now didn't change anything
about that but greatly improves the speed for low latency situations and that
was certainly needed as well. If we can't keep up at almost-zero latency, then
we certainly can't at high latency.

> You can have a look at the current API being used here:

Yes, the API really is the key here. That's what sets the restrictions or at
least sets the challenge level for how to do the implementation of this. I
need/want the API to remain like today as it's simple and elegant, but yet I
want high speed. I would be sad to sacrifize it.

-- 
  / daniel.haxx.se
_______________________________________________
libssh2-devel http://cool.haxx.se/cgi-bin/mailman/listinfo/libssh2-devel
Received on 2010-10-24