Messages in this thread | | | Date | Fri, 24 Mar 2000 22:14:56 -0800 | From | "David S. Miller" <> | Subject | Re: Update: Subtle data corruption of TCP streams |
| |
Date: Sat, 25 Mar 2000 00:20:12 -0500 (EST) From: wietse@PORCUPINE.ORG (Wietse Venema)
Seems that routers don't do the packet rewriting that I'm observing here. It's the domain of bandwith management systems. There are at least four players. I got reactions from several.
Disabling TCP options in the Linux kernel does help, especially when you're talking to non-Linux systems.
Some of these bandwidth management boxes have workarounds for bugs various TCP implementations have had in their TCP timestamp code. In one instance, a very old version of BSDi will not adjust the TCP packet length properly for full sized frames when it attaches the TCP timestamp option. The bandwidth management box attempts to "fix" this bug when it mangles this packet, but in effect it corrupts the packet for systems which do not have this bug.
I get very angry when I see magic boxes changing the amount of data sent per-packet in a TCP session, it breaks all of the carefully implemented algorithms in TCP to avoid congestion and improve throughput on lossy links etc.
The people using these boxes are effectively killing the internet. I can't think of any better way to say it.
Later, David S. Miller davem@redhat.com
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |