Tim Chown wrote: > On Mon, Oct 27, 2008 at 09:43:19AM +0100, Victor Grishchenko wrote: >> IPv4 world was built on an assumption of static IP addresses, so >> it is clearly impossible to introduce routine renumbering as an >> incremental feature because of the reasons you mentioned and >> other reasons. >> "Fluid" addresses are possible in the world of prefix bunches, >> multi-path TCP etc. Changing an underlying assumption affects >> all the stuff piled on top of it. > > There are always corner cases that either can't be covered or that have > some notable cost to work around. > > I know of someone who recently asked me about a problem they had... their > sensor network had a hard-coded IP address for the server to send data to. > Unfortunately the server was hosted in a network that was being renumbered > due to a change in ISP. > > The snag as you've probably already guessed was that these sensors were > buried some way underground and not reconfigurable. > > In every case there are things you can do to mitigate the cost of a > renumbering event, but there are most likely tradeoffs in doing so.
I think there are lots of things that could be done with design of protocols, applications and management systems to ease renumbering. Protocol design is maybe the least important IMO. The problem is that renumbering is generally not considered when designing such things. Stig _______________________________________________ rrg mailing list [email protected] https://www.irtf.org/mailman/listinfo/rrg
