Actions addressing identified issues with the Session Initiation Protocol's non-INVITE Transaction

There are a number of unpleasant edge conditions created by the SIP non-INVITE transaction (NIT) model's fixed duration. The negative aspects of some of these are exacerbated by the effect provisional responses have on the non-INVITE transaction state machines. These problems are documented in . In summary: A non-INVITE transaction must complete immediately or risk losing a race Losing the race will cause the requester to stop sending traffic to the responder (the responder will be temporarily blacklisted) Provisional responses can delay recovery from lost final responses The 408 response is useless for the non-INVITE transaction As non-INVITE transactions through N proxies time-out, there can be an O(N^2) storm of the useless 408 responses This draft specifies updates to RFC 3261 to improve the behavior of SIP elements when these edge conditions arise.

There are two goals to achieve when we constrain the problem to those cases where all elements are ultimately responsive and networks ultimately deliver messages: Reduce the probability of losing the race, preferably to the point that it is negligible Reduce or eliminate useless messaging

Disallow non-100 provisionals to non-INVITE requests Disallow 100 Trying to non-INVITE requests before Timer E reaches T2 (for UDP hops) Allow 100 Trying after Timer E reaches T2 (for UDP hops) Allow 100 Trying for hops over reliable transports Since non-INVITE transactions must complete rapidly (), any information beyond "I'm here" (which can be provided by a 100 Trying) can be just as usefully delayed to the final response. Sending non-100 provisionals wastes bandwidth. As shown in , sending any provisional response inside a NIT before Timer E reaches T2 damages recovery from failure of an unreliable transport. Without a provisional, a late final response is the same as no response at all and will likely result in blacklisting the late responding element (). If an element is delaying its final response at all, sending a 100 Trying after Timer E reaches T2 prevents this blacklisting without damaging recovery from unreliable transport failure. Blacklisting on a late response occurs even over reliable transports. Thus, if an element processing a request received over a reliable transport is delaying its final response at all, sending a 100 Trying well in advance of the timeout will prevent blacklisting. Sending a 100 Trying immediately will not harm the transaction as it would over UDP, but a policy of always sending such a message results in unneccessary traffic. A policy of sending a 100 Trying after the period of time in which Timer E reaches T2 had this been a UDP hop is one reasonable compromise.

Disallow 408 to non-INVITE requests Absorb stray non-INVITE responses at proxies A 408 to non-INVITE will always arrive too late to be useful (), The client already has full knowledge of the timeout. The only information this message would convey is whether or not the server believed the transaction timed out. However, with the current design of the NIT, a client can't do anything with this knowledge. Thus the 408 simply wasting network resources and contributes to the response bombardment illustrated in . Late non-INVITE responses by definition arrive after the client transaction's Timer F has fired and the client transaction has entered the Terminated state. Thus, these responses cannot be distinguished from strays. Changing the protocol behavior to prohibit forwarding non-INVITE stray responses stops the late response storm. It also improves the proxy's defenses against malicious users counting on the RFC 3261 requirement to forward such strays.

When we expand the scope of the problem to also deal with element or network failure, we have more goals to achieve: Identifying when an element is non-responsive Minimizing or eliminating falsely identifying responsive elements as non-responsive Avoiding non-responsive elements with future requests Action 1 helps with the first two goals, dramatically improving an element's ability to distinguish between failure and delayed response from the next downstream element. Some response, either provisional or final, will almost certainly be received before the transaction times out. So, an element can more safely assume that no response at all indicates the peer is not available and follow the existing requirements in and for that case. Achieving the third goal requires more agressive changes to the protocol. As noted in , future non-invite transactions are likely to fail again unless the implementation takes steps beyond what is defined in and to remember non-responsive destinations between transactions. Standardizing these extra steps is left to future work.

A SIP element MUST NOT send any provisional response with a Status-Code other than 100 to a non-INVITE request. A SIP element MUST NOT respond to a non-INVITE request with a Status-Code of 100 over any unreliable transport, such as UDP, before the amount of time it takes a client transaction's Timer E to be reset to T2. A SIP element MAY respond to a non-INVITE request with a Status-Code of 100 over a reliable transport at any time. Without regard to transport, a SIP element MUST respond to a non-INVITE request with a Status-Code of 100 if it has not otherwise responed after the amount of time it takes a client transaction's Timer E to be reset to T2.

A transaction-stateful SIP element MUST NOT send a response with Status-Code of 408 to a non-INVITE request. As a consequence, an element that can not respond before the transaction expires will not send a final response at all. A transaction-stateful SIP proxy MUST NOT send any response to a non-INVITE request unless it has a matching server transaction that is not in the Terminated state. As a consequence, this proxy will not forward any "late" non-INVITE response.

This document makes a number of small changes to the core SIP specification to improve the robustness of SIP non-INVITE transactions. Many of these actions also prevent flooding and denial-of-service attacks. One change prohibits proxies and User Agents from sending 408 responses to non-INVITE transactions. Without this change, proxies automatically generate a storm of useless responses as described in . An attacker could capitalize on this by enticing User Agents to send non-INVITE requests to a black hole (through social engineering or DNS poisoning) or by selectively dropping responses. Another change prohibits proxies from forwarding late responses. Without this change, an attacker could easily forge messages which appear to be late responses. All proxies compliant with RFC 3261 are required to forward these responses, wasting bandwidth and CPU and potentially overwhelming target User Agents (especially those with low speed connections). The remainder of these changes do not affect the security of the SIP protocol.

This document requires no action by IANA.

Rohan Mahy provided the Security Considerations section.