RE: [Lynx-dev] lynx via specific network interface

[Pawel Chwalinski]

Hi all,
Thank you for your prompt responses.
Yes, I know this problem sounds odd and believe me it is. It is part of my 
research and I have been looking for so many different solutions but still have 
not found the right one.
I am trying to perform some optimalisation of network performance. For that 
purpose, I need to implement theory that already exists. However, I need to 
test it against some real-world settings. Some of the network simulators my 
University provides (Opnet, ns etc.) have complicated (and sometimes limited) 
application when it comes to code modification. Thus I decided to create my own 
network environment. I want to link up few computers and configure on them few 
virtual ip addresses. Therefore I will end up having many network subnets with 
just few machines. 
As I mentioned before - I managed to test it with ping. it works fine and I am 
able to decide which interface to use.
I believe that the problem lies in the protocol usage - i.e. ping uses ICMP so 
it works fine, and presumable it is not straightforward with TCP based 
applications. The ease I am looking for is to define the rate of packet 
arrivals and frequency of visits for clients so that I can test the theory 
against multiple different scenario. With lynx these problems are solved 
easily, but the interface specification.

If you think it is possible - please let me know. I was thinking about whether 
or not it is possible to achieve by code modification. I thought that maybe the 
library lynx is based on uses similar settings to, say lpcap, where you can 
manually specify the network device you listen to.

Thank you all for your time and suggestions.

Cheers,

p.

________________________________________
From: David Woolley address@hidden
Sent: 02 March 2011 19:50
To: Pawel Chwalinski; lynx-dev listserv
Subject: Re: [Lynx-dev] lynx via specific network interface

Pawel Chwalinski wrote:

> Suppose we have 3 different interface installed on a computer, say,
> eth1,eth2 and eth3.
> Would it be possible to run lynx on any of them, given its name is
> specified with a starting-up command?

Probably, if you changed the source code to do this, but it is not a
normal thing to do in an application program.

It would probably be much more useful if you told us what your top level
objective was.  Then we might be able to suggest solutions that don't
break the layering between system and applications.

If you need to do this and can't achieve it by appropriate routing
tables, you are trying to abuse the system in some way.

--
David Woolley
Emails are not formal business letters, whatever businesses may want.
RFC1855 says there should be an address here, but, in a world of spam,
that is no longer good advice, as archive address hiding may not work.