120 likes | 131 Views
Base Protocol Errata and Pending Issues – for RFC3588bis Victor Fajardo (draft-fajardo-dime-diameter-errata-00.txt). Overview All contents of the current issues tracker in http://www.tschofenig.com:8080/diameter-interop/ has been transferred into the errata draft Includes open issues
E N D
Base Protocol Errata and Pending Issues – for RFC3588bisVictor Fajardo(draft-fajardo-dime-diameter-errata-00.txt) IETF67 DIME WG
Overview • All contents of the current issues tracker in http://www.tschofenig.com:8080/diameter-interop/ has been transferred into the errata draft • Includes open issues • With proposed solutions: Total of 10 issues • Without proposed solutions: Total of 2 issues • Includes useful optimization • Features: Total of 1 • Not disruptive to existing implementations IETF67 DIME WG
Issue Format • Each Issue defined in its own section with the following format • Description of the problem • Diff between text of proposed solution and corresponding text in RFC3588 • Description of the solution IETF67 DIME WG
Since last IETF … • Unresolved Issues • Issue 21 (Sec 2.7) • Issue 5 (Sec 2.8) • Issues with Proposed Solutions • Issue 1, 13, 14, 8, 9, 24, 16, 10, 17 and 18 • Closed Issues • Issue 2, 3, 11, 12, 19, 7 and 6 • Features • Issue 23 IETF67 DIME WG
Un-Resolved Issues Issue 21 (Sec 2.7): DiameterURI Schemes • Details: DiameterURI scheme does not fit models found in RFC3986, http://tools.ietf.org/wg/dime/draft-garcia-dime-aaa-uri-00.txt • Status: No activity Issue 5 (Sec 2.8): Application ID to be used by ASR/ASA, RAR/RAA, STR/STA • Details: Conflicting models on the relationship between base protocol and applications • Component vs. Non-component model • Component Model: Use AppId of application • Non-Component Model: Use AppId of zero (0) • Status: Very rough consensus - Use “Single Entity” model since this is the original intent of the authors and been propagated in other specifications IETF67 DIME WG
Resolved Issues Issue 14 (Sec 2.1): Usage of the E-bit and Error Answer • Details: Editorial – Clarify text on which error category should use the E-bit • Solution: Add text in each error category regarding E-bit Issue 8 (Sec 2.2): Usage of E-bit in the CEA message • Details: Un-necessary to set the E-bit in CEA because the Result-Code combined with a well defined the CER/CEA exchange is enough • Solution: Move DIAMETER_UKNOWN_PEER from protocol error to permanent error category IETF67 DIME WG
Resolved Issues Issue 9 (Sec 2.3): Inconsistent Result-code Error classification • Details: Several error codes are either duplicated or in the wrong category • Solution: • Result-codes that have been moved from protocol error to permanent errors: • DIAMETER_COMMAND_UNSUPPORTED • DIAMETER_INVALID_HDR_BITS • DIAMETER_INVALID_AVP_BITS • Result-codes that have been moved from permanent error to protocol errors: • DIAMETER_INVALID_BIT_IN_HEADER • DIAMETER_INVALID_MESSAGE_LENGTH • Result-codes that have been deprecated • DIAMETER_INVALID_AVP_BIT_COMBO – have same meaning as DIAMETER_INVALID_AVP_BITS • For backward compatibility: Result-codes which has been move from one category to another will be assigned new numeric values IETF67 DIME WG
Resolved Issues Issue 24 (Sec 2.4): Composing Failed AVP for Invalid AVP length • Details: It’s not possible to copy an offending AVP to the Failed AVP when the offending AVP reports a length that is greater than the message length or less than the AVP header length • Solution: Allow Failed AVP to carry only the offending AVPs header with a zero-filled payload with the minimum required length Issue 16 (Sec 2.5): Capabilities Exchange in the Open State • Details: Peer FSM implies CER/CEA exchange in the OPEN state but such behavior is not specified anywhere • Solution: • Add a new sub-section to fully describe the behavior • CER/CEA exchange is backward compatible and non-disruptive • Allow only application support changes and nothing else (i.e. security schemes) IETF67 DIME WG
Resolved Issues Issue 10 (Sec 2.6): ABNF for Vendor-Specific-Application-Id allows multiple Vendor-Ids • Details: Does not make sense that the ABNF allows multiple Vendor-Id in Vendor-Specific-Application-Id AVP • Solution: Modify the ABNF to require one instance of Vendor-Id AVP Issue 17 (Sec 2.9): Removal of Trailing [fixed*] avp in Command Code ABNF specification • Details: diameter-message ABNF specifies the use of trailing [fixed*]. This implies additional parsing requirements even if it is not used or there is no need for it. • Solution: Remove trailing [fixed*] from diameter-message ABNF IETF67 DIME WG
Resolved Issues Issue 18 (Sec 2.10): Mapping between Disconnect-Cause AVP and the Reconnect Behavior • Details: Clarifications are needed regarding when a peers should attempt to re-connect after a DPR/DPA exchange based on the result-code provided by the Disconnect-Cause AVP • Solution: • REBOOTING: Peer MAY attempt re-connection • BUSY and DO_NOT_WANT_TO_TALK: Peer SHOULD NOT attempt re-connection Issue 1 (Sec 2.11): Advertising of Relay application-id and the Election • Details: Which Application-Id AVP should be used when advertising yourself as a Relay agent, AppId = 0xffffffff • Solution: Clearly specify that either Auth-Application-Id or Acct-Application-Id can carry a relay AppId IETF67 DIME WG
Resolved Issues Issue 13 (Sec 2.12): Duplication of Application ID in the Header and Application Id AVPs • Details: Clarification on why we have multiple ways of carrying AppId information, i.e. Diameter Header and Application-Id AVPs and how do they relate to each other • Solution: • Clearly state that the AppId in the diameter header MUST be the primary method of obtaining the application identifier of a message • Further re-enforce that AppId in the diameter header MUST match any Application-Id AVPs present in the message • Clarify how Vendor-Specific-Application-Id must be treated IETF67 DIME WG
Optimization Issue 23 (Sec 3.1): Predictive Loop Avoidance • Details: A routing loop that occurs in a diameter node can already be detected and possibly avoided by its previous-hop • Solution: The previous-hop can check to see if the next-hop for a message is already present in the messages’ Route-record AVP. If so, it can attempt to forward to an alternate route or send an DIAMETER_UNABLE_TO_DELIVER IETF67 DIME WG