+----------------------------+
                 |                            |
                 V                            |
              +------+   Client               |
              |CLOSED|-----SYN------+         |
              +------+              |         |
                  ^                 |         |
                  |TCP_TRANS T.O.   |         |
                  |                 V         |
              +-------+          +-------+    |
              | TRANS |          |  INIT |    |
              +-------+          +-------+    |
                |    ^               |        |
          data pkt   |               |        |
                | Server/Client RST  |        |
                |  TCP_EST T.O.      |        |
                V    |           Server SYN   |
           +--------------+          |        |
           | ESTABLISHED  |<---------+        |
           +--------------+                   |
            |           |                     |
      Client FIN    Server FIN                |
            |           |                     |
            V           V                     |
     +---------+   +----------+               |
     |  C FIN  |   |  S FIN   |               |
     |   RCV   |   |    RCV   |               |
     +---------+   +----------+               |
         |             |                      |
     Server FIN      Client FIN            TCP_TRANS
         |             |                    T.O.
         V             V                      |
     +----------------------+                 |
     |   C FIN + S FIN RCV  |-----------------+
     +----------------------+

 Legend:
   * Messages sent to (resp. received from) the server
     are prefixed with "Server".
   * Messages sent to (resp. received from) the client
     are prefixed with "Client".
   * "C" means "Client-side"
   * "S" means "Server-side".
   * TCP_EST T.O: refers to the established connection 
     idle timeout as defined in [RFC5382].
   * TCP_TRANS T.O: refers to the transitory connection  
     idle timeout as defined in [RFC5382].

Figure 1: State Machine

[RFC5382] specifies TCP timers associated with various connection states but does not specify the TCP state machine a NAT44 should follow as a basis to apply such timers.

Update:

The TCP state machine depicted in Figure 1, adapted from [RFC6146], SHOULD be implemented by a NAT for TCP session tracking purposes.

2.1. TCP Transitory Connection Idle-Timeout

The transitory connection idle-timeout is defined as the minimum time a TCP connection in the partially open or closing phases must remain idle before the NAT considers the associated session a candidate for removal (REQ-5 of [RFC5382]). But [RFC5382] does not clearly state whether these can be configured separately.

Clarification:

This document clarifies that a NAT SHOULD provide different configurable parameters for configuring the open and closing idle timeouts.

To accommodate deployments that consider a partially open timeout of 4 minutes as being excessive from a security standpoint, a NAT MAY allow to configure the timeout to be less than 4 minutes. Still, this specification recommends the default "transitory connection idle-timeout" minimum value to be set to 4 minutes.

2.2. TCP RST

[RFC5382] leaves the handling of TCP RST packets unspecified.

Update:

This document adopts a similar default behavior as in [RFC6146]. Concretely, when the NAT receives a TCP RST matching an existing mapping, it MUST translate the packet according the NAT mapping entry. Moreover, the NAT SHOULD wait for 4 minutes before deleting the session and removing any state associate with it if no packets are received during that 4 minutes timeout.

Admittedly, the NAT has to verify whether received TCP RST packets belong to a connection. These verification checks are required to avoid off-path attacks.

If the NAT removes immediately the NAT mapping upon receipt of a TCP RST message, stale connections may be maintained by endpoints if the first RST message is lost between the NAT and the recipient.

3. Port Overlapping Behavior

REQ-1 from [RFC4787] and REQ-1 from [RFC5382] specify a specific port overlapping behavior; that is the external IP address and port can be reused for connections originating from the same internal source IP address and port irrespective of the destination. This is known as endpoint-independent mapping (EIM).

Update:

This document clarifies that this port overlapping behavior may be extended to connections originating from different internal source IP addresses and ports as long as their destinations are different.

The following mechanism MAY be implemented by a NAT:

• If destination addresses and ports are different for outgoing connections started by local clients, a NAT MAY assign the same external port as the source ports for the connections. The port overlapping mechanism manages mappings between external packets and internal packets by looking at and storing their 5-tuple (protocol, source address, source port, destination address, destination port).

This enables concurrent use of a single NAT external port for multiple transport sessions, which allows a NAT to successfully process packets in an IP address resource limited network (e.g., deployment with high address space multiplicative factor (refer to Appendix B. of [RFC6269])).

4. Address Pooling Paired (APP)

The Address Pooling Paired (APP) behavior for a NAT was recommended in REQ-2 from [RFC4787], but the behavior when a public IPv4 runs out of ports was left undefined.

Clarification:

This document clarifies that if APP is enabled, new sessions from a host that already has a mapping associated with an external IP that ran out of ports SHOULD be dropped.

The administrator MAY provide a configurable parameter that allows a NAT to starting using ports from another external IP address when the one that anchored the APP mapping ran out of ports. This is a trade-off between service continuity and APP strict enforcement. (Note, this behavior is sometimes referred as 'soft-APP'.)

Update:

This behavior SHOULD apply also for TCP.

5. EIF Protocol Independence

REQ-8 from [RFC4787] and REQ-3 from [RFC5382] do not specify whether EIF mappings are protocol-independent. In other words, if an outbound TCP SYN creates a mapping, it is left undefined whether inbound UDP packets destined to that mapping should be forwarded.

Update:

This document specifies that EIF mappings SHOULD be protocol-independent in order allow inbound packets for protocols that multiplex TCP and UDP over the same IP address and port through the NAT and also maintain compatibility with stateful NAT64 . The administrator MAY provide a configuration parameter to make it protocol-dependent. The default value of this configuration parameter is to allow for protocol-independent EIF.

Applications that can be transported over a variety of transport protocols and/or support transport fall back schemes won't experience connectivity failures as a function of the underlying transport protocol or the filtering mode enabled at the NAT.

6. EIF Mapping Refresh

The NAT mapping Refresh direction may have a "NAT Inbound refresh behavior" of "True" according to REQ-6 from [RFC4787], but [RFC4787] does not clarify how this behavior applies to EIF mappings. The issue in question is whether inbound packets that match an EIF mapping but do not create a new session due to a security policy should refresh the mapping timer.

Clarification:

This document clarifies that even when a NAT has an inbound refresh behavior set to 'TRUE', such packets SHOULD NOT refresh the mapping. Otherwise a simple attack of a packet every 2 minutes can keep the mapping indefinitely.

Update:

This behavior SHOULD apply also for TCP.

6.1. Outbound Mapping Refresh and Error Packets

Update:

In the case of NAT outbound refresh behavior there are certain types of packets that should not refresh the mapping even if their direction is outbound. For example, if the mapping is kept alive by ICMP Errors or TCP RST outbound packets sent as response to inbound packets, these SHOULD NOT refresh the mapping.

7. EIM Protocol Independence

REQ-1 from [RFC4787] and REQ-1 from [RFC5382] do not specify whether EIM are protocol-independent. In other words, if a outbound TCP SYN creates a mapping it is left undefined whether outbound UDP can reuse such mapping and create session. On the other hand, stateful NAT64 [RFC6146] clearly specifies three binding information bases (TCP, UDP, ICMP).

Update:

EIM mappings SHOULD be protocol-dependent. A configuration parameter MAY be provided in order allow protocols that multiplex TCP and UDP over the same source IP address and port number to use a single mapping.

8. Port Parity

Update:

A NAT MAY disable port parity preservation for all dynamic mappings. Nevertheless, A NAT SHOULD support means to explicitly request to preserve port parity (e.g., [I-D.ietf-pcp-port-set]).

• Note: According to [RFC6887], dynamic mappings are said to be dynamic in the sense that they are created on demand, either implicitly or explicitly:
  1. Implicit dynamic mappings refer to mappings that are created as a side effect of traffic such as an outgoing TCP SYN or outgoing UDP packet. Implicit dynamic mappings usually have a finite lifetime, though this lifetime is generally not known to the client using them.
  2. Explicit dynamic mappings refer to mappings that are created as a result, for example, of explicit PCP MAP and PEER requests. Explicit dynamic mappings have a finite lifetime, and this lifetime is communicated to the client.

9. Port Randomization

Update:

A NAT SHOULD follow the recommendations specified in Section 4 of [RFC6056], especially:

• "A NAPT that does not implement port preservation [RFC4787] [RFC5382] SHOULD obfuscate selection of the ephemeral port of a packet when it is changed during translation of that packet. A NAPT that does implement port preservation SHOULD obfuscate the ephemeral port of a packet only if the port must be changed as a result of the port being already in use for some other session. A NAPT that performs parity preservation and that must change the ephemeral port during translation of a packet SHOULD obfuscate the ephemeral ports. The algorithms described in this document could be easily adapted such that the parity is preserved (i.e., force the lowest order bit of the resulting port number to 0 or 1 according to whether even or odd parity is desired)."

10. IP Identification (IP ID)

Update:

A NAT SHOULD handle the Identification field of translated IPv4 packets as specified in Section 5.3.1 of [RFC6864].

Discussion:

This recommendation may have undesired effects on the performance of the NAT in environments in which fragmentation is massively experienced. Such issue can be used as an attack vector against NATs.

11. ICMP Query Mappings Timeout

Section 3.1 of [RFC5508] precises that ICMP Query Mappings are to be maintained by a NAT. However, the specification doesn't discuss Query Mapping timeout values. Section 3.2 of [RFC5508] only discusses ICMP Query Session Timeouts.

Update:

ICMP Query Mappings MAY be deleted once the last the session using the mapping is deleted.

12. Hairpinning Support for ICMP Packets

REQ-7 from [RFC5508] specifies that a NAT enforcing 'Basic NAT' must support traversal of hairpinned ICMP Query sessions.

[RFC5508] specifies that all NATs must support the traversal of hairpinned ICMP Error messages.

Clarification:

This implicitly means that address mappings from external address to internal address (similar to Endpoint Independent Filters) must be maintained to allow inbound ICMP Query sessions. If an ICMP Query is received on an external address, a NAT can then translate to an internal IP.

REQ-7 from

Clarification:

This behavior requires a NAT to maintain address mappings from external IP address to internal IP address in addition to the ICMP Query Mappings described in Section 3.1 of [RFC5508].

13. IANA Considerations

This document does not require any IANA action.

14. Security Considerations

NAT behavioral considerations are discussed in [RFC4787].

Security considerations discussed in Section 5 of [RFC6146] apply also fro NAT44.

In the case of EIF mappings due to high risk of resource crunch, a NAT MAY provide a configurable parameter to limit the number of inbound sessions spawned from a EIF mapping.

15. References

15.1. Normative References

[RFC2119]	Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, March 1997.
[RFC4787]	Audet, F. and C. Jennings, "Network Address Translation (NAT) Behavioral Requirements for Unicast UDP", BCP 127, RFC 4787, DOI 10.17487/RFC4787, January 2007.
[RFC5382]	Guha, S., Biswas, K., Ford, B., Sivakumar, S. and P. Srisuresh, "NAT Behavioral Requirements for TCP", BCP 142, RFC 5382, DOI 10.17487/RFC5382, October 2008.
[RFC5508]	Srisuresh, P., Ford, B., Sivakumar, S. and S. Guha, "NAT Behavioral Requirements for ICMP", BCP 148, RFC 5508, DOI 10.17487/RFC5508, April 2009.
[RFC6056]	Larsen, M. and F. Gont, "Recommendations for Transport-Protocol Port Randomization", BCP 156, RFC 6056, DOI 10.17487/RFC6056, January 2011.
[RFC6146]	Bagnulo, M., Matthews, P. and I. van Beijnum, "Stateful NAT64: Network Address and Protocol Translation from IPv6 Clients to IPv4 Servers", RFC 6146, DOI 10.17487/RFC6146, April 2011.
[RFC6864]	Touch, J., "Updated Specification of the IPv4 ID Field", RFC 6864, DOI 10.17487/RFC6864, February 2013.
[RFC6888]	Perreault, S., Yamagata, I., Miyakawa, S., Nakagawa, A. and H. Ashida, "Common Requirements for Carrier-Grade NATs (CGNs)", BCP 127, RFC 6888, DOI 10.17487/RFC6888, April 2013.

15.2. Informative References

[I-D.ietf-pcp-port-set]	Qiong, Q., Boucadair, M., Sivakumar, S., Zhou, C., Tsou, T. and S. Perreault, "Port Control Protocol (PCP) Extension for Port Set Allocation", Internet-Draft draft-ietf-pcp-port-set-09, May 2015.
[RFC2663]	Srisuresh, P. and M. Holdrege, "IP Network Address Translator (NAT) Terminology and Considerations", RFC 2663, DOI 10.17487/RFC2663, August 1999.
[RFC3022]	Srisuresh, P. and K. Egevang, "Traditional IP Network Address Translator (Traditional NAT)", RFC 3022, DOI 10.17487/RFC3022, January 2001.
[RFC6269]	Ford, M., Boucadair, M., Durand, A., Levis, P. and P. Roberts, "Issues with IP Address Sharing", RFC 6269, DOI 10.17487/RFC6269, June 2011.
[RFC6887]	Wing, D., Cheshire, S., Boucadair, M., Penno, R. and P. Selkirk, "Port Control Protocol (PCP)", RFC 6887, DOI 10.17487/RFC6887, April 2013.

Acknowledgements

Thanks to Dan Wing, Suresh Kumar, Mayuresh Bakshi, Rajesh Mohan, Lars Eggert, and Gorry Fairhurst for their review and discussion.

Contributors

The following individual contributed text to the document:

• Sarat Kamiset, Insieme Networks, United States

Authors' Addresses