Margaret Wasserman [mailto:[EMAIL PROTECTED] wrote:
> Hi Jeroen, > > >These enterprises apparently don't want/require/need global > >reachability for their hosts. Otherwise they would not NAT. > > That depends on what you mean by "global reachability". > I am writing to you from behind a NAT right now. From here, > I can reach web sites on the global Internet, etc. I can't > run servers here, so I need to depend on my friends to do > that for me. I rather meant end-to-end connectivity which makes the sentence have a completely different meaning. Note that my boxes are also living behind a NAT in IPv4 but they do have global connectivity in IPv6. I rather have e2e because then every machine can also do all the really interresting applications like H323. <SNIP> > >IMHO the real solution to this and some other problems we > >are currently seeing in IPv6 is really one thing which > >must be solved before anything else: IPv6 Multihoming > > I'm not sure how IPv6 Multihoming applies here. Could you > explain? Michel Py explained this 'solution' quite well. He's then also quite busy on the IPv6 multihoming front :) > > > So, if we don't come up with a way to allow > > > provider-independent address > > > allocation in IPv6, we will probably get IPv6<->IPv6 NAT. > > > >We don't want PI because that would also imply a routingtable > >explosion. PI thus is not the answer. > > The simplest ways to provide PI addresses imply routing table > explosion. There are people (in the IETF, IRTF and elsewhere) > working on mechanisms for provider-independent addressing that > avoid routing table explosion. I certainly hope that they will > be successful, as that would solve a lot of problems. I hope for them to be sucessful too but I have little faith in it. > >Taking a, imho, good application like [loadbalancers] in view > >NAT should not be forbidden... > > > >(Then again, the loadbalancer could just also have all the > >backends configured with the global IP and just forward the > >packets to the correct box... hmmm ;) > > I don't have any interest in eliminating load balancers, but > are you sure that this is how they work? What happens when > the server passes its IP addresses in FTP, SCTP or SIP > packets (or any other application-layer protocol)? Does > the loadbalancer also translate those addresses to point to > the loadbalancer, or is it assumed that the client node > can (and should) reach the server directly in those cases? What I have seen they do. Though if I would have to implement one I'd quite probably go for the second version which doesn't need any NATting but just some maintainance in the loadbalancer describing which host is connected to which backend thus making the loadbalancer a policy based router. Should be quite easy to built too I think. But I also think that the real builders can comment on these situations. Nevertheless it would be great if loadbalancers sported IPv6 as it would mean that they also could handle huge sites like CNN and Google which would be one way to allow them to upgrade to IPv6. Greets, Jeroen -------------------------------------------------------------------- IETF IPng Working Group Mailing List IPng Home Page: http://playground.sun.com/ipng FTP archive: ftp://playground.sun.com/pub/ipng Direct all administrative requests to [EMAIL PROTECTED] --------------------------------------------------------------------