Hi,
Regarding use-cases below, the operator will likely rely on a centralized policy server; actually the 3GPP/ANDSF has been specified for that purpose. Pierrick ________________________________ De : mif-boun...@ietf.org [mailto:mif-boun...@ietf.org] De la part de Tao Sun Envoyé : lundi 5 décembre 2011 20:25 À : Lorenzo Colitti Cc : mif@ietf.org Objet : Re: [mif] Use cases for draft-ietf-mif-dhcpv6-route-option needed Ok. I reiterate the use cases and the justification of using DHCPv6. For the cellular network, I mainly focus on two use cases: 1) A host is connected to the cellular network and WiFi at the same time. Some traffic need to go cellular network, others go to WiFi. 2) A mobile host in the cellular network simultaneously connects to multiple APNs (taking APN1 and APN2 for example) . Some traffics go to APN1, others got to APN2. The justification of adopt DHCPv6 is the following: 1) DHCPv6 is anyway shall be implemented and is suitable for parameter configuration. This is due to prefix delegation for tethering, network parameter configuration such as DNS, network server informaiton of IMS, etc. 2) DHCPv6 is more management/operation friendly due to the central control mechanism. 3) DHCP has been already used for paramter configuration, while using RA will cause significant modification and enhancement for network elements and hosts, especially for per-subscriber purpose (e.g., using VLAN). 4) Using RA to configure the route may cause potentially some messy. In 4G cellular network, both GTP and PMIP protocal are used for mobility management today. In GTP, it is a point-to-point link between mobile host and PGW. However, in PMIP, PGW send /64 prefix to SGW through PBA. The SGW sends RA to mobile host. If RA is also used for route configuration, how SGW obtains the specific route information? how about the complexity due to the large number of SGW as compared with PGW? how the SGW combine the multiple RA route informatio nreceived from sources in use case 2 since there is always only one SGW at a time? ... Based on above, using RA for SLAAC and DHCPv6 for route paramter configuration is a more effective way. In my view, the discussion is adequate enough to move the draft forward to version 4 and we shall not continue to stay on the necessity analysis. Tao Sun On Sat, Dec 3, 2011 at 3:21 AM, Lorenzo Colitti <lore...@google.com> wrote: On Fri, Dec 2, 2011 at 10:15, Tao Sun <hisun...@gmail.com> wrote: I'm not talking about mobile networks specifically here. I think this particular event (turning off RAs) is more likely that to happen on a wifi network or an enterprise network. As I pointed out in my previous email, WiFi network is already part of mobile network. China Mobile deployed about more than one hundred of thousand hotspots (over 1 million WiFi AP). Users use the same smart phone to visit cellular network and WiFi network. Ok, so then let's take a look at that use case and see why it requires a DHCPv6 route option to be defined. I suspect that you can do what you want using RIO.
_______________________________________________ mif mailing list mif@ietf.org https://www.ietf.org/mailman/listinfo/mif
