/127 between routers?

Mark Smith nanog at 85d5b20a518b8f6864949bd940457dc124746ddc.nosense.org
Fri Jan 8 01:34:43 CET 2010


On Thu, 7 Jan 2010 10:39:20 -0800
"George Bonser" <gbonser at seven.com> wrote:

> 
> > > History has shown that creating hard boundaries between the network
> > and
> > > node portion (i.e. Classful addressing) may reduce forwarding
> > > system flexibility and require mass software/hardware upgrades.
> > 
> > You can't apply a lesson learned when addresses are scarce to a
> > scenario
> > where addresses are plentiful. I bet we would hear the exact same
> > concerns if IPv6 addresses were 256-bit.
> > 
> > 
> > /Benny
> 
> I can imagine all sorts of things happening after the widespread
> deployment of v6.  For example, let's say you currently have some kind
> of application that keeps session state on a server that is in a
> cluster.  So when a request arrives, you use some sort of session ID to
> locate the server with the state information.  But let's say with v6 you
> now simply program an ip address that has the value of the session ID.
> To find the server with the state, you simply connect to the IP address
> represented by or that is derived as some function of the session ID.
> 
> We might see servers with thousands of IP addresses configured where
> those IP addresses are basically session cookies.  This is going to put
> some interesting pressure on kernel developers. So you have a subnet
> where the addresses are basically nothing more than 64-bit GUIDs
> representing client connections.  Want to find where the client state
> is?  No problem, connect to the GUID as the IP address or redirect the
> connection there.  No more cluster software needed.
> 
> 
> 

Sounds like what is in this paper -

"Transient addressing for related processes: Improved firewalling by
using IPv6 and multiple addresses per host." - Peter M. Gleitz and
Steven M. Bellovin

http://www.cs.columbia.edu/~smb/papers/tarp.pdf



More information about the ipv6-ops mailing list