Dave, Marc, and all, This is Akira Nakagawa, Tokyo. Thank you for the coments.
> > (Q1) This proposal leads extending the life of IPv4, so IPv6 will not > > deploy. > > > > (A1) No, our goal is IPv6. NAT444 + Shared Address is designed for the > > co- > > existing period. The purpose of this network is to access from IPv6 > > enabled > > network to remaining IPv4 network. > > I don't see how IPv6 is used in the NAT444+shared address model. > You say "No" above to "this proposal extends the life of IPv4". > I think you mean "Yes" to that part. At the Internet-Area session in Minneapolis, I should have explained how to use shared address. I was focusing on explaining only IPv4 part of the entire network model. Just before the Internet-Area session, I explained the network structure of NAT444. See page 7 of my slide. http://www.ietf.org/proceedings/08nov/slides/behave-7.pdf As described on the slide, NAT444 = IPv6 + IPv4(444) for the co-existing period. The purpose of this model is not extending the life of IPv4. > > (Q4) Possible to use 240/4 in ISP-controllable backbone instead of > > getting > > Shared Address. > > > > (A4) Miyakawa-san asnwered on Nov.25 (JST) on this ML. > > (At least, in Japan,) now so many PCs are running as a CPE device. > > > > Also, if a PC behind CPE in home network need to contact > > any machine in a network X between LNS and CPE (see below), > > and the network X is using Class E, > > PC in the home can not send/receive any packets > > to machines in the network X. > > > > This is not good, because we should save any status on a LNS > > as much as possible.... > > > > Please understand that we must needs to have new space > > to keep existing devices in the home. > > > > > > to/from the Internet > > | > > | > > +--------+--------+ > > | Large Scale NAT | > > +--------+--------+ +------------------------+ > > | |SOME SERVICE such as DNS| > > | +------+-----------------+ > > | | > > ---------+-------------------------+--------- <- Network X > > | > > | > > +----+-----+ > > | CPE | > > +----+-----+ > > | > > | > > ---------+------------ HOME NETWORK > > | > > +----+-----+ > > | PC | > > +----------+ > > > > If you have PC's serving as a CPE then your proposal won't help them > when such PCs are multihomed since the address space is not unique. > Multihoming is much more likely when it's a PC CPE than when it's an ISP CPE. As I answered in Q5, multihome service is out of scope. See Q5/A5. If I don't get your question, let me know, please. > > (Q5) When Multihoming with two ISPs, needed to consider the path > > selection > > like RFC3484. > > > > (A5) Customers who require Multihoming Service are not the target of > > NAT444 > > + Shared Address model. They should use expensive service that uses > > Global > > Address and/or Provider Independent Address (PI Address). > > To do this service, ISP should manage the limited number of Global > > Address > > in its network. > > > > (Q6) Rewritten Source Address prevents 6to4 communication. > > > > (A6) Nothing will change as today's typical network model. > > The endures who can use 6to4 service have to use the nodes with Global > > IPv4 > > address. > > Today, we cannot use 6to4 service because we use private address in our > > local network. So nothing will change after introducing Shared Address. > > Yes it will. If a PC gets an address, it will think it's a public address > and will try using 6to4 and run into problems where it thinks it should > work and it doesn't work right. This is very different from today where > if it gets an address it thinks is public it works, and if it gets an address > it thinks is private, it doesn't try 6to4. 6to4 is out of scope, because as I explained in Q1 avobe, NAT444 model is IPv6/IPv4 dual stack. > (Q9) ISPs aren't special. Corporate VPNs have the same problem with > conflicting space, for example. (I believe there are other examples > but VPN's are the most well known.) If ISPs get a space, then so > should VPNs and various other things. <snip> > [DT] It's not covered by Q7 above. Q7 above is about whether a possible > space given to ISPs will work or not. Q9 is about whether we should > give a space to ISPs, whether we should give another space to VPNs, > whether we should give another space to every other class of network > for which there exist many instances. Q9 is saying that we should not > do so for ISPs unless we also are willing to do so for all the others. I'll explain it using the example of VPN. [Fig.A] [Fig.B] [Fig.C] [Global] +-------+ +-------+ | LSN-C +---+ LSN-D | +---+---+ +---+---+ | | +-----------------+ +-----------------+ +---+---+ +---+---+ | | | | | | | | | ISP-A | | ISP-B | | ISP-C | | ISP-D | | | | | | | | | +---+---------+---+ +---+---------+---+ +---+---+ +---+---+ | | | | | | | | | | | | | | | | | | |[P1] |[P3] |[S] |[S] |[S1] |[S2] +--+--+ +--+--+ +--+--+ +--+--+ +--+--+ +--+--+ | VPN | | VPN | | VPN | | VPN | | VPN | | VPN | +--+--+ +--+--+ +--+--+ +--+--+ +--+--+ +--+--+ |[P2] |[P4] |[P] |[P] |[P] |[P] | | | | | | Conflict OK Network model issue [Fig.A] If ISP-A uses Private address for its backbone, ISP's Private Address (P1) and customer's Private (P2) conflict. (might be defferent issue) [Fig.B] If ISP-B uses Shared Address, it works because shared address is unique in one ISP. [Fig.C] If two sites belong to deferent ISPs, Shared address S1 and Shared address S2 may conflict. If they use Class-E or Private address instead of using Shared Address, it also happens. But S1 and S2 are covered with Large Scale NATs. So it is not the address type issue. When this case happens, ISPs should persuade their customers to use IPv6 service or expensive Global IPv4 Address service. akira _______________________________________________ Int-area mailing list [email protected] https://www.ietf.org/mailman/listinfo/int-area
