Re: Network namespaces a path to mergable code.

James Morris [off-list ref] writes:

On Wed, 28 Jun 2006, Daniel Lezcano wrote:

quoted

The attached patch can have some part interesting for you for the socket
tagging. It is in the IPV4 isolation (part 5/6). With this and the private
routing table you will probably have a good IPV4 isolation.

Please send patches inline, do not attach them.

(Perhaps we should have a filter on vger which drops emails with 
attachements).

All of this needs to be done in a way where it can be entirely disabled at 
compile time, so there is zero overhead for people who don't want 
network namespaces.

I agree with the principle of no overhead.  

The goal is an implementation that has no measurable overhead when
there is only one network namespace. 

If that goal is achieved and you can compile in the network namespace
code and not measure overhead there should be no need for a compile
time option.

Eric