Ubuntu Desktop dns binding help

Kevin Fries kfries6 at gmail.com
Tue Feb 15 12:54:09 MST 2011


On 02/15/2011 11:51 AM, Joseph Sinclair wrote:
> I would assume that he's talking about broad testing within a local network, rather than testing against localhost directly.
> I often do this because I can insert firewalls, routers, etc... as/where desired to emulate probable scenarios.  It's particularly helpful to emulate 4in6 or 6in4 connections when using external providers that do not provide sufficient IPv6 support.
>
> It's just easier to create a hostfile entry on the test client(s) than to create or modify public DNS (sometimes that's not even possible).  This is particularly true when the service you're testing is already live and you need to black-box test a component of an interconnected SOA system.
Yes, I understand that.  My point is to question if this is wise at 
all.  I have seen far too many times where a computer sends traffic out 
to its public address, and still does not respond the same way it does 
in production.  The reason is one NIC.  You are routing from yourself to 
yourself, and it will get turned around at the NIC.  I have worked R&D 
for the past 4 1/2 years, and trust me, this happens far more than most 
people think.  You would be better off bouncing off another computer, 
that redirects the traffic truly from another machine.

Example 1: Use NX Free edition to get a desktop on another computer, and 
then run your browser on that.

Example 2: Use a *Nix machine for which you have root access to create a 
forwarded port (ssh -R 80:mypublicip:80 root at server).  This makes the IP 
address on the foreign machine tunnel back to yourself, and cuts out 
optimizations at both the NIC and the switch and gives you a true 
experience as to what your clients will see.

Example 3: Have a second NIC.  Force traffic out through NIC-1 to the 
public IP on NIC-2.  The switch and NIC have no idea that the machine 
sending and responding are one in the same.  Therefore, once again, you 
can eliminate any ability of the devices to optimize.

As I said in my original comment.  The goal is to avoid the "It works on 
my machine" situation.

I hope my comments made more sense this time.

Kevin


More information about the PLUG-discuss mailing list