This Annex contains the rules applied by the SGSN upon PDP context activation to determine the APN and the corresponding P-GW/GGSN. The selection process used by the SGSN to select P-GW and GGSN is described in clause 5.3.7.1. The procedures specified in TS 29.303 apply to DNS-based P-GW selection. APN selection refers to the process of selection and construction of the full APN (APN-NI + APN-OI). This full APN is then employed for interrogation of the DNS server to obtain the GGSN or P-GW address.

The SGSN knows from the subscription data the following parameters (S for Subscribed): PDP type (S), PDP address (S), APN (S), and VPLMN address allowed. In addition, the S4-SGSN and MME receive a PDN subscription context that is marked as default (and associated default APN) for E-UTRAN UEs. The SGSN may know from configuration the Local APN supporting a given PDP type. This APN is called APN (SGSN) and does not include an APN Operator Identifier. APN (SGSN) shall not be an APN with LIPA permisssions set to "LIPA-only" or "LIPA-conditional". The SGSN knows the parameters requested by the MS (R for Requested): PDP type (R), PDP address (R), and APN (R). APN (R) is the APN Network Identifier requested by the MS. In case of "an APN chosen by the SGSN" the activated PDP context is always linked with a dynamic PDP address. An MS may have multiple subscription records for the same PDP type and the same PDP address, but with different APNs. An MS may have one or two subscription records with the same PDP type and the same APN: one with a static PDP address, one with a dynamic PDP address. An MS may have multiple subscription records with the same APN, but with different PDP types. When the MS is in its HPLMN, if the MS requests an APN that does not correspond to any GGSN or P-GW of its HPLMN, the request shall be rejected by the SGSN. When the MS is in a VPLMN, if the MS requests an APN that does not correspond to any GGSN or P-GW of its HPLMN nor of this VPLMN or any of its associated PLMNs when the VPLMN is a shared network, the request shall be rejected by the SGSN. If APN (S) = wild card (see TS 23.003), it means either:

• that a Local APN (a locally defined PDN) has to be chosen by the SGSN (APN (SGSN)) if no APN (R) has been provided; or
• that a PDP context with dynamic PDP address may be activated towards any APN requested by the MS.

The PDN subscription context that is marked as default for the default bearer activation, defines a Default APN that takes precedence over the locally defined APN for the S4-SGSN and MME. In order to derive APN (R) from the APN sent by the MS, the SGSN shall check if the APN sent by the user ends with ".gprs". If not, then APN (R) is equal to APN sent by the MS. If yes, then APN (R) is the APN sent by the MS without the three last labels. For deriving a GGSN by the procedure defined in the SDL Diagram, PDPtype(R) shall be assumed equal to PDPtype(S) if PDPtype(R) is IPv4 or IPv6 and PDPtype(S) is IPv4v6.

The SGSN shall select the APN to be used to derive the GGSN or P-GW address, and set the selection mode parameter according to the rules in the SDL diagrams in this clause. The following definitions apply to the SDL diagrams: AddrMode: APN OI: HPLMN AP: HPLMN OI-1: HPLMN OI-2: Number <condition>: ODB parameter: PDPaddr: SelMode: VPLMN AP: VPLMN OI: PDN-GW allocation type: +: The subscription data may contain the information whether SIPTO is allowed or prohibited for each subscribed APN or the SGSN may know from configuration whether SIPTO is allowed or prohibited for a given APN. An indication that SIPTO is allowed or prohibited for the wild card APN allows or prohibits SIPTO for any APN that is not present in the subscription data. When a PDP context for SIPTO is established, the SGSN uses either the RAI (Routing Area Identity) and/or the serving RNC identifier depending on the operator's configuration during the DNS interrogation as specified in TS 29.303. In roaming scenario the GW selection for a PDP context with SIPTO is only possible when a GGSN or a P-GW in the visited PLMN is selected. Therefore in a roaming scenario with home routed traffic, GW selection for SIPTO is not performed. When the UE is in a network with A/Gb mode and SIPTO is allowed for the given APN, the RAI may also be used for PGW/GGSN selection. In this way when the UE enters UTRAN or E-UTRAN the PDP context deactivation with reactivation request to find an optimal PGW/GGSN for SIPTO may not be needed. The subscription data may also contain the information on whether:

1. an APN is LIPA-conditional, LIPA-prohibited, or LIPA-only;
2. a CSG in the UE's CSG subscription data supports APN(s) that are LIPA-only or LIPA-conditional; and
3. LIPA is allowed in a list of VPLMNs when roaming.

When a PDP context for a LIPA APN is established, the SGSN uses the L-GW address included by HNB in RANAP messages as the GGSN/PGW address to be used, instead of DNS interrogation for GGSN/PGW selection. If no collocated L-GW address is included by the HNB and the UE requested a LIPA only APN, the request shall be rejected. If no collocated L-GW address is included by the HNB and the UE requested a LIPA conditional APN, the SGSN uses DNS interrogation for GGSN/PGW selection to establish a non-LIPA PDP context. The subscription data for an APN with LIPA permissions set to "LIPA-only" shall not contain a statically configured PDP address or a statically allocated PDN-GW. A static PDP address or a static PDN-GW address, if configured by HSS for an APN with LIPA permissions set to "LIPA-conditional", is ignored by SGSN when the APN is established as a LIPA PDP context. When establishing a PDP context for a LIPA APN, the VPLMN Address Allowed flag is not considered. In the procedure denoted "Interface and protocol selection" in Figure A.8, the SGSN shall select one of the configurations listed in Table A.1. The SGSN may use the UE capability (indicated as part of the MS Network Capability) and roaming agreement as indicated in clause 5.3.7.1 as input to select between configurations using GGSN or P-GW as well as the interface type (Gp or S4/S8). The SGSN may give priority for a configuration using P-GW for E-UTRAN capable UEs, and GGSN for non E-UTRAN capable UE. If the SGSN supports Gn/Gp only, selection between the configurations indexed 1 and 2 are applicable. If the SGSN supports both Gn/Gp and S4, any of the configurations in Table A.1 apply. In case of P-GW selection, the service parameter shall be set as given in the respective column of Table A.1 and applied as defined in TS 29.303. If the HLR/HSS provides a subscription context that allows for allocation of a GGSN/PGW from the visited PLMN for this APN and, optionally, the SGSN is configured to know that the visited VPLMN has a suitable roaming agreement with the HPLMN of the UE, the GGSN/PGW selection function derives a GGSN/PGW from the VPLMN. If a visited GGSN/PGW cannot be derived, the APN is used to derive a GGSN/PGW from the HPLMN. DNS interrogation in Figure A.8 shall be performed based on the full APN (APN-NI +APN-OI). For index of 2, 3, or 4 DNS interrogation procedure is defined by TS 29.303. For index 1 the DNS interrogation is a DNS A query and/or DNS AAAA query at the full APN exactly as in pre-Release 8 networks. Fall back to the legacy procedure (i.e. index 1) is required for indexes 2, 3, and 4 if they fail since the APN may represent a pre-Release 8 network.