[Devel] Re: [PATCH 0/9] namespaces: Introduction

[Sam Vilain]

On Fri, 2006-05-19 at 17:47 +0400, Andrey Savochkin wrote:
> We can start with presenting and merging the most interesting part, network
> containers.  We discuss details, possible approaches, and related subsystems,
> until networking is finished to its utmost detail.
> This will create an example of virtualization of a non-trivial subsystem,
> and we will have to agree on basic principles of virtualization of related
> subsystems like proc.
[...]
> What do people think about this plan?

Network is an interesting one because you have multiple solutions - the
very simple approach of network binding (as used by Jacques Gelina's
original IP vhost work from December 1997), and network virtualisation.
That virtualisation itself can be broken down into providing merely
virtual interfaces (so that, for instance, you can have independent lo
interfaces in the virtual servers) as in Vserver 2.1.x, or providing a
completely virtualised network stack, as in Vserver ngnet (and possibly
OpenVZ?).

Each solution performs the virtualisation at a different level, and has
incrementally higher orders of inefficiency and maintenance
requirements.  Yet none of them are essentially better or worse than the
others.

So, we might end up with all three eventually - but binding alone is the
simplest and still extremely useful.

Sam.