From: "Björn Persson" <spam-away@nowhere.nil>
Subject: Re: Libre site down, again
Date: Tue, 17 Oct 2006 19:31:00 GMT
Date: 2006-10-17T19:31:00+00:00 [thread overview]
Message-ID: <UbaZg.20197$E02.7962@newsb.telia.net> (raw)
In-Reply-To: <1161102935.771738.42820@e3g2000cwe.googlegroups.com>
Someone claiming to be a lethal noddle wrote:
> The strange thing is, running a 2.6.16 kernel as I am now,
> tcp_window_scaling is set to 1, but I can access libre just fine.
They changed the calculation of TCP buffer sizes. I think the point was
to make better use of available memory. If there's plenty of memory,
then Linux can afford to allocate a large buffer so that the sender can
keep the pipe full of packets, and a large buffer requires a large scale
factor. But the larger the scale factor is, the more hindered the
communication is when these broken boxes mangle it. If a small scale
factor gets mangled, your buffer will seem smaller than it is but the
connection can still limp along. If a large scale factor gets mangled,
the connection freezes completely.
--
Bj�rn Persson PGP key A88682FD
omb jor ers @sv ge.
r o.b n.p son eri nu
next prev parent reply other threads:[~2006-10-17 19:31 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2006-10-14 10:07 Libre site down, again Tapani Rundgren
2006-10-14 10:19 ` Pascal Obry
2006-10-14 10:41 ` Björn Persson
2006-10-14 17:42 ` Tapani Rundgren
2006-10-14 18:02 ` Pascal Obry
2006-10-14 23:04 ` Björn Persson
2006-10-15 7:37 ` Pascal Obry
2006-10-17 12:47 ` Olivier Ramonat
2006-10-17 16:35 ` deadlyhead
2006-10-17 19:31 ` Björn Persson [this message]
2006-10-17 20:22 ` Samuel Tardieu
2006-10-18 13:45 ` Thomas Quinot
2006-10-18 19:21 ` Björn Persson
replies disabled
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox