WS-I

Reliable Secure Profile Version 1.0

Working Group Draft

2009-07-20

This revision:
http://members.ws-i.org/dman/Folder.phx/Private+Folders/Community+Folder/Working+Groups/Reliable+Secure+Profile/Profile/1.0/ReliableSecureProfile-1.0-WGD.html?versionID=$Revision: 1.2 $
Latest revision:
http://members.ws-i.org/dman/Folder.phx/Private+Folders/Community+Folder/Working+Groups/Reliable+Secure+Profile/Profile/1.0/ReliableSecureProfile-1.0-WGD.html
Editors:
Jacques Durand, Fujitsu
Gilbert Pilz, Oracle
Administrative contact:
secretary@ws-i.org

Abstract

This document defines the WS-I Reliable Secure Profile 1.0, consisting of a set of non-proprietary Web services specifications, along with clarifications, refinements, interpretations and amplifications of those specifications which promote interoperability. It also contains a set of executable test assertions for assessing the conformance to the profile.

Status of this Document

This document is a Working Group Draft; it has been accepted by the Working Group as reflecting the current state of discussions. It is a work in progress, and should not be considered authoritative or final; other documents may supersede this document.

Notice

The material contained herein is not a license, either expressly or impliedly, to any intellectual property owned or controlled by any of the authors or developers of this material or WS-I. The material contained herein is provided on an "AS IS" basis and to the maximum extent permitted by applicable law, this material is provided AS IS AND WITH ALL FAULTS, and the authors and developers of this material and WS-I hereby disclaim all other warranties and conditions, either express, implied or statutory, including, but not limited to, any (if any) implied warranties, duties or conditions of merchantability, of fitness for a particular purpose, of accuracy or completeness of responses, of results, of workmanlike effort, of lack of viruses, and of lack of negligence. ALSO, THERE IS NO WARRANTY OR CONDITION OF TITLE, QUIET ENJOYMENT, QUIET POSSESSION, CORRESPONDENCE TO DESCRIPTION OR NON-INFRINGEMENT WITH REGARD TO THIS MATERIAL.

IN NO EVENT WILL ANY AUTHOR OR DEVELOPER OF THIS MATERIAL OR WS-I BE LIABLE TO ANY OTHER PARTY FOR THE COST OF PROCURING SUBSTITUTE GOODS OR SERVICES, LOST PROFITS, LOSS OF USE, LOSS OF DATA, OR ANY INCIDENTAL, CONSEQUENTIAL, DIRECT, INDIRECT, OR SPECIAL DAMAGES WHETHER UNDER CONTRACT, TORT, WARRANTY, OR OTHERWISE, ARISING IN ANY WAY OUT OF THIS OR ANY OTHER AGREEMENT RELATING TO THIS MATERIAL, WHETHER OR NOT SUCH PARTY HAD ADVANCE NOTICE OF THE POSSIBILITY OF SUCH DAMAGES.

Feedback

The Web Services-Interoperability Organization (WS-I) would like to receive input, suggestions and other feedback ("Feedback") on this work from a wide variety of industry participants to improve its quality over time.

By sending email, or otherwise communicating with WS-I, you (on behalf of yourself if you are an individual, and your company if you are providing Feedback on behalf of the company) will be deemed to have granted to WS-I, the members of WS-I, and other parties that have access to your Feedback, a non-exclusive, non-transferable, worldwide, perpetual, irrevocable, royalty-free license to use, disclose, copy, license, modify, sublicense or otherwise distribute and exploit in any manner whatsoever the Feedback you provide regarding the work. You acknowledge that you have no expectation of confidentiality with respect to any Feedback you provide. You represent and warrant that you have rights to provide this Feedback, and if you are providing Feedback on behalf of a company, you represent and warrant that you have the rights to provide Feedback on behalf of your company. You also acknowledge that WS-I is not required to review, discuss, use, consider or in any way incorporate your Feedback into future versions of its work. If WS-I does incorporate some or all of your Feedback in a future version of the work, it may, but is not obligated to include your name (or, if you are identified as acting on behalf of your company, the name of your company) on a list of contributors to the work. If the foregoing is not acceptable to you and any company on whose behalf you are acting, please do not provide any Feedback.

Feedback on this document should be directed to wsi_rsp_comment@lists.ws-i.org.


Table of Contents

1. Introduction
1.1. Relationships to Other Profiles
1.2. Guiding Principles
1.3. Test Assertions
1.4. Notational Conventions
1.5. Profile Identification and Versioning
2. Profile Conformance
2.1. Conformance Requirements
2.2. Conformance Targets
2.3. Conformance Scope
2.4. Claiming Conformance
3. Reliable Messaging
3.1. Use of Extension Elements and Attributes in Messages
3.1.1. Ignore Unknown Extension Elements
3.2. SOAP Version Considerations
3.2.1. SOAP Version Selection for Sequence Lifecycle Messages
3.3. Targeting Sequence Lifecycle Messages
3.3.1. CreateSequence Target
3.3.2. Use of the Offer Element
3.4. Sequence Identifiers
3.4.1. Duplicate Identifier in CreateSequenceResponse
3.5. Sequence Termination
3.5.1. Sequence Termination from the Destination
3.5.2. Last Message Number
3.5.3. Sequence Lifecycle Independence
3.6. Sequence Faults
3.6.1. Transmission of Sequence Faults
3.6.2. WS-ReliableMessaging Faults
3.7. Sequence Assignment
3.7.1. Sequence Assignment for Reliable Response Messages
3.7.2. Scope of an RM Node
3.8. Retransmission of Messages
3.8.1. Retransmission of Unacknowledged Messages
3.8.2. Retransmission of Sequence Lifecycle Messages
3.8.3. Message Identity
3.9. Piggybacking
3.9.1. Endpoint Comparison for Piggybacked SequenceAcknowledgment Headers
3.9.2. Treatment of ReferenceParameters in AcksTo EPRs
3.9.3. Preventing Piggybacked Acknowledgements
3.9.4. Conflicting Requirements for wsa:Action
3.9.5. Use of the mustUnderstand Attribute
4. Secure Conversation
4.1. Unsupported Context Tokens
4.1.1. Unrecognized Extensions in a Security Context Token
4.2. Demonstrating Proof of Possession
4.2.1. Amending Contexts
4.2.2. Renewing Contexts
4.2.3. Cancelling Contexts
4.3. Claims Re-Authentication
4.3.1. Re-Authenticating Claims
4.4. Referencing Security Context Tokens
4.4.1. Associating a Security Context
4.4.2. Derived Token References to Security Contexts
4.5. Addressing Headers
4.5.1. Protecting Addressing Headers
5. MakeConnection
5.1. Guidance On the Use of MakeConnection
5.1.1. Action Values
5.1.2. Binding to HTTP
5.2. MakeConnection Addressing
5.2.1. Addressing Variants
5.2.2. MakeConnection Anonymous URI
6. Secure Reliable Messaging
6.1. Initiating a Secure Sequence
6.1.1. Secure Context Identification
6.1.2. Security Token References
6.2. Signature Coverage
6.2.1. Single Signature for Sequence Header and SOAP Body
6.2.2. Signed Elements
6.2.3. Single Signature for SOAP 1.1 Fault and SequenceFault Header
6.3. Secure Use of MakeConnection
6.3.1. Security Context for MakeConnection
6.3.2. Signing the MessagePending header
6.4. Replay Detection
6.4.1. Unique Timestamp Values
Appendix A: Referenced Specifications
Appendix B: Extensibility Points
Appendix C: Normative References
Appendix D: Acknowledgements

1. Introduction

This document defines the WS-I Reliable Secure Profile 1.0 (hereafter, "Profile"), consisting of a set of non-proprietary Web services specifications, along with clarifications, refinements, interpretations and amplifications of those specifications which promote interoperability.

Section 1 introduces the Profile, and explains its relationships to other profiles.

Section 2, "Profile Conformance," explains what it means to be conformant to the Profile.

Each subsequent section addresses a component of the Profile, and consists of two parts; an overview detailing the component specifications and their extensibility points, followed by subsections that address individual parts of the component specifications. Note that there is no relationship between the section numbers in this document and those in the referenced specifications.

1.1 Relationships to Other Profiles

This Profile is intended to be composed with the WS-I Basic Profile 1.2, WS-I Basic Profile 2.0, WS-I Basic Security Profile 1.0 and WS-I Basic Security Profile 1.1. Composability of RSP with the previously mentioned profiles offers the following guarantee to users: conformance of an artifact to RSP does not prevent conformance of this artifact to these other profiles, and vice-versa.

Because the conformance targets defined for RSP may not match exactly the conformance targets for another profile, the following more precise definition of composability is assumed in this profile:

A profile P2 is said to be composable with a profile P1 if, for any respective pair of conformance targets (T2, T1) where T1 depends on T2 (see definition below), conformance of an instance of T2 to P2 does not prevent conformance of the related T1 instance(s) to P1, and vice-versa in case T2 depends on T1.

A target T1 is said to depend on a target T2 if either:

In order to conform to this profile (RSP):

1.2 Guiding Principles

The Profile was developed according to a set of principles that, together, form the philosophy of the Profile, as it relates to bringing about interoperability. This section documents these guidelines.

No guarantee of interoperability
It is impossible to completely guarantee the interoperability of a particular service. However, the Profile does address the most common problems that implementation experience has revealed to date.
Application semantics
Although communication of application semantics can be facilitated by the technologies that comprise the Profile, assuring the common understanding of those semantics is not addressed by it.
Testability
When possible, the Profile makes statements that are testable. However, such testability is not required. Preferably, testing is achieved in a non-intrusive manner (e.g., examining artifacts "on the wire").
Strength of requirements
The Profile makes strong requirements (e.g., MUST, MUST NOT) wherever feasible; if there are legitimate cases where such a requirement cannot be met, conditional requirements (e.g., SHOULD, SHOULD NOT) are used. Optional and conditional requirements introduce ambiguity and mismatches between implementations.
Restriction vs. relaxation
When amplifying the requirements of referenced specifications, the Profile may restrict them, but does not relax them (e.g., change a MUST to a MAY).
Multiple mechanisms
If a referenced specification allows multiple mechanisms to be used interchangeably, the Profile selects those that are well-understood, widely implemented and useful. Extraneous or underspecified mechanisms and extensions introduce complexity and therefore reduce interoperability.
Future compatibility
When possible, the Profile aligns its requirements with in-progress revisions to the specifications it references. This aids implementers by enabling a graceful transition, and assures that WS-I does not 'fork' from these efforts. When the Profile cannot address an issue in a specification it references, this information is communicated to the appropriate body to assure its consideration.
Compatibility with deployed services
Backwards compatibility with deployed Web services is not a goal for the Profile, but due consideration is given to it; the Profile does not introduce a change to the requirements of a referenced specification unless doing so addresses specific interoperability issues.
Focus on interoperability
Although there are potentially a number of inconsistencies and design flaws in the referenced specifications, the Profile only addresses those that affect interoperability.
Conformance targets
Where possible, the Profile places requirements on artifacts (e.g., WSDL descriptions, SOAP messages) rather than the producing or consuming software's behaviors or roles. Artifacts are concrete, making them easier to verify and therefore making conformance easier to understand and less error-prone.
Lower-layer interoperability
The Profile speaks to interoperability at the application layer; it assumes that interoperability of lower-layer protocols (e.g., TCP, IP, Ethernet) is adequate and well-understood. Similarly, statements about application-layer substrate protocols (e.g., SSL/TLS, HTTP) are only made when there is an issue affecting Web services specifically; WS-I does not attempt to assure the interoperability of these protocols as a whole. This assures that WS-I's expertise in and focus on Web services standards is used effectively.

1.3 Test Assertions

This profile document contains embedded Test Assertions (TA) that are associated with each normative profile requirement. In the HTML rendering of this document, these test assertions are accessible via a toggle link at the end of each requirement. When clicking on such a link, a table pops up that displays the TA parts. At the end of this table is another toggle link ("help-glossary") that displays an explanation glossary for the TA structure. In other formats of this document, the test assertions are grouped in an appendix not controlled by any link, in order to facilitate the printing of hard copies. The resulting set of test assertions embedded in this document represents a conformance test suite for the profile.

Release notes related to the test material included in this document are available here:

TESTING-RELEASE-NOTES

1.4 Notational Conventions

The keywords "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC2119.

Normative statements of requirements in the Profile (i.e., those impacting conformance, as outlined in " Conformance Requirements") are presented in the following manner:

RnnnnStatement text here.

where "nnnn" is replaced by a number that is unique among the requirements in the Profile, thereby forming a unique requirement identifier.

Requirement identifiers can be considered to be namespace qualified, in such a way as to be compatible with QNames from Namespaces in XML. If there is no explicit namespace prefix on a requirement's identifier (e.g., "R9999" as opposed to "bp10:R9999"), it should be interpreted as being in the namespace identified by the conformance URI of the document section it occurs in. If it is qualified, the prefix should be interpreted according to the namespace mappings in effect, as documented below.

Some requirements clarify the referenced specification(s), but do not place additional constraints upon implementations. For convenience, clarifications are annotated in the following manner: C

Some requirements are derived from ongoing standardization work on the referenced specification(s). For convenience, such forward-derived statements are annotated in the following manner: xxxx, where "xxxx" is an identifier for the specification (e.g., "WSDL20" for WSDL Version 2.0). Note that because such work was not complete when this document was published, the specification that the requirement is derived from may change; this information is included only as a convenience to implementers.

As noted above, some requirements may present compatibility issues (whether forwards or backwards) with previously published versions of the profile. For convenience, such requirements are annotated in the following manner: Compat

Extensibility points in underlying specifications (see " Conformance Scope") are presented in a similar manner:

EnnnnExtensibility Point Name - Description

where "nnnn" is replaced by a number that is unique among the extensibility points in the Profile. As with requirement statements, extensibility statements can be considered namespace-qualified.

This specification uses a number of namespace prefixes throughout; their associated URIs are listed below. Note that the choice of any namespace prefix is arbitrary and not semantically significant.

1.5 Profile Identification and Versioning

This document is identified by a name (in this case, Reliable Secure Profile) and a version number (here, 1.0). Together, they identify a particular profile instance.

Version numbers are composed of a major and minor portion, in the form "major.minor". They can be used to determine the precedence of a profile instance; a higher version number (considering both the major and minor components) indicates that an instance is more recent, and therefore supersedes earlier instances.

Instances of profiles with the same name (e.g., "Example Profile 1.1" and "Example Profile 5.0") address interoperability problems in the same general scope (although some developments may require the exact scope of a profile to change between instances).

One can also use this information to determine whether two instances of a profile are backwards-compatible; that is, whether one can assume that conformance to an earlier profile instance implies conformance to a later one. Profile instances with the same name and major version number (e.g., "Example Profile 1.0" and "Example Profile 1.1") MAY be considered compatible. Note that this does not imply anything about compatibility in the other direction; that is, one cannot assume that conformance with a later profile instance implies conformance to an earlier one.

2 Profile Conformance

Conformance to the Profile is defined by adherence to the set of requirements defined for a specific target, within the scope of the Profile. This section explains these terms and describes how conformance is defined and used.

2.1 Conformance Requirements

Requirements state the criteria for conformance to the Profile. They typically refer to an existing specification and embody refinements, amplifications, interpretations and clarifications to it in order to improve interoperability. All requirements in the Profile are considered normative, and those in the specifications it references that are in-scope (see "Conformance Scope") should likewise be considered normative. When requirements in the Profile and its referenced specifications contradict each other, the Profile's requirements take precedence for purposes of Profile conformance.

Requirement levels, using RFC2119 language (e.g., MUST, MAY, SHOULD) indicate the nature of the requirement and its impact on conformance. Each requirement is individually identified (e.g., R9999) for convenience.

For example;

R9999 Any WIDGET SHOULD be round in shape.

This requirement is identified by "R9999", applies to the target WIDGET (see below), and places a conditional requirement upon widgets; i.e., although this requirement must be met to maintain conformance in most cases, there are some situations where there may be valid reasons for it not being met (which are explained in the requirement itself, or in its accompanying text).

Each requirement statement contains exactly one requirement level keyword (e.g., "MUST") and one conformance target keyword (e.g., "MESSAGE"). The conformance target keyword appears in bold text (e.g. " MESSAGE"). Other conformance targets appearing in non-bold text are being used strictly for their definition and NOT as a conformance target. Additional text may be included to illuminate a requirement or group of requirements (e.g., rationale and examples); however, prose surrounding requirement statements must not be considered in determining conformance.

Definitions of terms in the Profile are considered authoritative for the purposes of determining conformance.

None of the requirements in the Profile, regardless of their conformance level, should be interpreted as limiting the ability of an otherwise conforming implementation to apply security countermeasures in response to a real or perceived threat (e.g., a denial of service attack).

2.2 Conformance Targets

Conformance targets identify what artifacts (e.g., SOAP message, WSDL description, UDDI registry data) or parties (e.g., SOAP processor, end user) requirements apply to.

This allows for the definition of conformance in different contexts, to assure unambiguous interpretation of the applicability of requirements, and to allow conformance testing of artifacts (e.g., SOAP messages and WSDL descriptions) and the behavior of various parties to a Web service (e.g., clients and service instances).

Requirements' conformance targets are physical artifacts wherever possible, to simplify testing and avoid ambiguity.

The following conformance targets are used in the Profile:

2.3 Conformance Scope

The scope of the Profile delineates the technologies that it addresses; in other words, the Profile only attempts to improve interoperability within its own scope. Generally, the Profile's scope is bounded by the specifications referenced by it.

The Profile's scope is further refined by extensibility points. Referenced specifications often provide extension mechanisms and unspecified or open-ended configuration parameters; when identified in the Profile as an extensibility point, such a mechanism or parameter is outside the scope of the Profile, and its use or non-use is not relevant to conformance.

Note that the Profile may still place requirements on the use of an extensibility point. Also, specific uses of extensibility points may be further restricted by other profiles, to improve interoperability when used in conjunction with the Profile.

Because the use of extensibility points may impair interoperability, their use should be negotiated or documented in some fashion by the parties to a Web service; for example, this could take the form of an out-of-band agreement.

The Profile's scope is defined by the referenced specifications in Appendix A, as refined by the extensibility points in Appendix B.

2.4 Claiming Conformance

Claims of conformance to the Profile can be made using the following mechanisms, as described in Conformance Claim Attachment Mechanisms, when the applicable Profile requirements associated with the listed targets have been met:

The conformance claim URI for this Profile is "http://ws-i.org/profiles/rsp/1.0" .

3. Reliable Messaging

This section of the Profile incorporates the following specifications by reference, and defines extensibility points within them:

3.1 Use of Extension Elements and Attributes in Messages

The protocol elements defined by WS-ReliableMessaging contain extension points wherein implementations MAY add child elements and/or attributes.

3.1.1 Ignore Unknown Extension Elements

To ensure the ability to safely extend the protocol, it is necessary that adding an extension does not create the risk of impacting interoperability with non-extended implementations.

R0001 A RECEIVER MUST NOT generate a fault as a consequence of receiving a message (e.g. wsrm:CreateSequence) that contains extension elements and/or attributes that it does not recognize. Any exceptions to this rule are clearly identified in requirements below or the specifications underlying the profile TESTABLE RSP0001

While the extensibility points of the profiled specifications can be used, per R0001 they MUST be ignored if they are not understood. However if a SENDER wishes to ensure that the RECEIVER understands and will comply with any such extensions, they need to include a SOAP header, marked with mustUnderstand="1" , in the request message that requires adherence to the semantics of those extensions.

3.2 SOAP Version Considerations

In general, it is not expected that the service descriptions for applications that use WS-ReliableMessaging will include bindings of the WS-RM protocol itself. This being the case, there is some uncertainty about which version of SOAP should be used to carry Sequence Lifecycle Messages.

3.2.1 SOAP Version Selection for Sequence Lifecycle Messages

For messages that flow from the RMS to the RMD, the version(s) of SOAP used for Sequence Lifecycle Messages are constrained to the version(s) of SOAP that are supported by the target endpoint (i.e. the endpoint to which the client is attempting to reliably communicate). For example, if a client is attempting to communicate reliably to an endpoint who's service description indicates that it only supports SOAP 1.1, the RMS should only send Sequence Lifecycle Messages using SOAP 1.1. Sequence Lifecycle Response Messages (CreateSequenceResponse, TerminateSequenceResponse, and CloseSequenceResponse) should use the version of SOAP used by their corresponding request message (CreateSequence, TerminateSequence, and CloseSequence respectively); this applies to WS-RM fault messages as well. For messages that flow from the RMD to the RMS (SequenceAck messages with an empty body, unsolicited CloseSequence messages, and unsolicited TerminateSequence messages) this profile adheres to and expands upon WS-RM's statement that "The SOAP version used for the CreateSequence message SHOULD be used for all subsequent messages in or for that Sequence, sent by either the RM Source or the RM Destination".

R0900 Unless otherwise specified (e.g. through some WSDL or WS-Policy designator), the RMD MUST send Sequence Lifecycle Messages destined to the CreateSequence/AcksTo EPR with the same SOAP version that was used in the CreateSequence message. TESTABLE RSP0900

R0901 Unless otherwise specified (e.g. through some WSDL or WS-Policy designator), the RMS of an Offered Sequence MUST send Sequence Lifecycle Messages destined to the CreateSequence/Offer/Endpoint EPR with the same SOAP version that was used in the CreateSequence message. TESTABLE RSP0901

3.3 Targeting Sequence Lifecycle Messages

WS-ReliableMessaging is silent on where certain Sequence Lifecycle Messages (such as CreateSequence) should be sent.

3.3.1 CreateSequence Target

The WS-RM specification is silent on exactly where an RMS should send a CreateSequence message to establish a Sequence. This is true for the case of a client-side RMS creating a Sequence to carry request messages as well as the case of a server-side RMS creating a Sequence to carry response messages. This is an interoperability issue because, unless the respective RMS and RMD implementations agree on the expected target for CreateSequence messages, the intended recipient may not configure the necessary infrastructure (WS-RM message handlers, etc.) and the CreateSequence message may either cause a fault or be ignored.

R0800 Baring some out of band agreement, an ENVELOPE carrying a CreateSequence message MUST be addressed to the same destination as one of the Sequence Traffic Message for that Sequence. TESTABLE RSP0800

This requirement applies equally to cases in which the first Sequence Traffic Message is addressed to a URI (as may happen when the target endpoint is retrieved from a WSDL document) or to an EPR (as may happen when the target endpoint is the wsa:ReplyTo address of the corresponding request message).

3.3.2 Use of the Offer Element

The use of the Offer element within a CreateSequence message is an optional feature of WS-ReliableMessaging. Using Offer avoids the exchange of CreateSequence and CreateSequenceResponse messages to establish a new sequence for response messages. However, WS-RM does not define a mechanism by which an RMS can determine if an Offer is desired by the RMD. This creates a potential interoperability issue in cases where an RMS that either doesn't wish to use or cannot support the use of Offer attempts to create a Sequence with an RMD that requires the use of Offer. To ensure interoperability, the Offer feature must be optional for both the initiator of the Sequence (the RMS) as well as the RMD.

Conversely, when an RMS includes an Offer within a CreateSeqence and the RMD rejects that Offer (e.g. if it only has input-only operations and concludes it has no need for the offered Sequence), if the RMD indicates this choice by faulting the CreateSequence the RMS has no programmatic means of determining that the fault was due to the presence of an Offer. To ensure interoperatbility in these cases, the RMD, rather than faulting the CreateSequence, must instead simply not accept the offered Sequence by not including an Accept element in the CreateSequenceResponse.

R0010 An RMD MUST NOT fault a CreateSequence due to the absence of the Offer element. TESTABLE RSP0010

R0011 An RMD MUST NOT fault a CreateSequence due to the presence of the Offer element. TESTABLE RSP0011

3.4 Sequence Identifiers

Under certain conditions it is possible for the CreateSequence or CreateSequenceResponse messages to be lost or delayed. Depending upon the timing of the attempts to resend such messages, it is possible to receive duplicate CreateSequence or CreateSequenceResponse messages (in fact, it is possible to receive duplicate messages even without retries). This creates the potential for CreateSequence and CreateSequenceResponse messages that contain duplicate Sequence Identifiers. Furthermore there are situations in which one party (RMS or RMD) may erroneously send a CreateSequence or CreateSequenceResponse message with a duplicate Sequence Identifier. Due to the crucial role of Sequence Identifiers in the WS-RM protocol, the handling of duplicate Sequence Identifiers needs to be further refined to prevent interoperability problems.

3.4.1 Duplicate Identifier in CreateSequenceResponse

Regardless of the causative circumstances, the existence of two, non-terminated Sequences with the same Identifier makes it difficult for the RMS to correctly function, therefore the RMS should take steps to prevent this condition.

R0700 The RMS MUST generate a fault when it receives a CreateSequenceResponse that contains a Sequence Identifier that is the same as the Identifier of a non-terminated Sequence. NOT_TESTABLE COM0700

Note that this requirement does not differentiate between duplicate Identifiers created by "the same" RMD or "different" RMDs; the simple fact that the RMS already has an active Sequence with the same Identifier is enough to trigger this requirement.

3.5 Sequence Termination

Termination of sequences must be done in a way to ensure that both the RMS and RMD share a common understanding of the final status of the sequence. The Profile places the following requirements on termination procedures:

3.5.1 Sequence Termination from the Destination

An RMS may need to get a final sequence acknowledgment, for supporting a particular delivery assurance. This is only possible after the sequence is closed and before it is terminated. When the termination is decided by the RMD, the RMS must also be made aware of this closure so that it can request a final acknowledgement.

R0200 In the case where an RMD decides to discontinue a sequence, it MUST close the Sequence and MUST attempt to send a wsrm:CloseSequence message to the AcksTo EPR. NOT_TESTABLE COM0200

3.5.2 Last Message Number

Among other benefits, the use of Sequence Message Numbers makes an RMD aware of gaps - messages it has not received - in a sequence. For this awareness to apply to messages missing from the end of a sequence the RMD must be aware of the highest message number sent.

R0210 Any ENVELOPE containing either a wsrm:CloseSequence or a wsrm:TerminateSequence element MUST also contain a wsrm:LastMsgNumber element if the Sequence in question contains at least one Sequence Traffic Message. TESTABLE RSP0210

There is a corner case for sequences in which no messages have been sent (i.e. empty sequences). In these cases it is permissable to omit wsrm:LastMsgNumber since there is no valid value for this element.

3.5.3 Sequence Lifecycle Independence

WS-ReliableMessaging is unclear about the relationship, if any, between the lifecycles of a Sequence and its corresponding Offered Sequence. Considering that such a relationship is not necessary for the proper functioning of the WS-RM protocol and that the existence of a such a relationship would create unnecessary and undesirable interdependencies between the RMS and the RMD, this profile makes the clarifying requirement that no such relationship exists.

R0220 An RM-NODE (RMD or RMS) MUST NOT assume that the termination (or closure) of a Sequence implicitly terminates (or closes) any other Sequence. TBD

3.6 Sequence Faults

This Profile adds the following requirement to the handling of faults that are generated as the result of processing WS-RM Sequence Lifecycle messages.

3.6.1 Transmission of Sequence Faults

In Section 4, "Faults" WS-ReliableMessaging states that a receiver that generates a fault related to a known sequence SHOULD transmit that fault. However, the WS-I Basic Profile 1.2 states, in requirement R1029, that, under certain circumstances, the receiver must transmit the fault. Mapping the specifics of the BP 1.2 requirement onto the details of the WS-RM specification results in the following requirement:

R0400 If a fault is generated while processing a wsrm:CreateSequence, wsrm:CloseSequence, or wsrm:TerminateSequence message, or a message containing a wsrm:AckRequested header, the RECEIVER MUST transmit the fault. TESTABLE RSP0400a RSP0400b

3.6.2 WS-ReliableMessaging Faults

The use of WS-ReliableMessaging for faults that are themselves related to the WS-RM protocol is undefined and unlikely to be interoperable. Accordingly this profile prohibits the assignment of WS-RM fault messages to a WS-RM Sequence.

R0620 An ENVELOPE that has wsrm:SequenceTerminated, wsrm:UnknownSequence, wsrm:InvalidAcknowledgement, wsrm:MessageNumberRollover, wsrm:CreateSequenceRefused, wsrm:SequenceClosed, or wsrm:WSRMRequired as the value of either the SOAP 1.2 /S:Fault/S:Code/S:Subcode/S:Value element or the /wsrm:SequenceFault/wsrm:FaultCode element MUST NOT contain a wsrm:Sequence header block. TESTABLE RSP0620

3.7 Sequence Assignment

WS-ReliableMessaging is silent on the mechanism for assigning messages (either request messages or response messages) to a particular Sequence. While this flexibility is beneficial from a general web services specification perspective, it creates some interoperability issues.

3.7.1 Sequence Assignment for Reliable Response Messages

Given a scenario in which a consumer and a provider engage in a series of reliable request/response exchanges, it is important for the consumer and provider to have a common understanding of the Sequence assignment mechanism for reliable response messages. Without such an understanding it is impossible, for example, to implement in-order delivery for response messages.

R0600 Any two ENVELOPEs that are sent reliably by an RMS in response - either as a replies or as faults - to two request messages that were sent within the same Sequence, SHOULD contain the same wsrm:Identifier (that is, share the same Sequence). TESTABLE RSP0600

Note that the RMS referred to above is a "server-side RMS" (i.e. the RMS responsible for transmitting response messages from the producer to the consumer in a reliable fashion).

3.7.2 Scope of an RM Node

WS-ReliableMessaging does not define the scope of an RM node other that to say that the scope is not restricted. For example, with respect to R0600 above, an Offered Sequence should be used to carry the response to any message sent over the Sequence corresponding to the CreateSequence request that included the Offer. However, it should not be assumed that the Sequence Traffic Messages carried over the Offered Sequence must be addressed to a particular response endpoint.

R0610 The scope of an RM Node is an implementation choice that MUST NOT be constrained by the remote RM-NODE. For example, the RMD MUST NOT constrain the values used by the RMS in the wsa:ReplyTo EPRs used by the RMS to be the same for all request messages (Sequence and Lifecycle messages). TESTABLE RSP0610

Within this context the phrase "scope of an RM node" is defined as "the set of all EPRs that address a given RM node".

3.8 Retransmission of Messages

WS-ReliableMessaging protocol requires retransmission of messages. The Profile places the following restrictions and refinements on such retransmissions:

3.8.1 Retransmission of Unacknowledged Messages

To ensure reliable delivery of messages within a Sequence, it is necessary for the RMS to retransmit unacknowledged messages and for the RMD to accept them.

R0101 An RMS MUST continue to retransmit unacknowledged messages until the Sequence is closed or terminated. TESTABLE RSP0101

R0102 An RMD MUST accept unacknowledged message until the Sequence is closed or terminated. TESTABLE RSP0102

Note: there are cases where it may be obvious that retransmitting a message is unlikey to result in an outcome that is any different from the previous, failed transmission(s). For example, in the case of HTTP, a 401 status code may indicate that access to an endpoint has been refused for the credentials that accompanied the request. Unless some action is taken to grant access to those credentials, retransmitting the request is likely to result in the same error and may cause negative side-effects such as the locking of an account due to "excessive failed login attempts".

3.8.2 Retransmission of Sequence Lifecycle Messages

WS-ReliableMessaging Section 2.1 defines the messages that affect the created/closing/closed/terminating state of a Sequence as "Sequence Lifecycle Messages". WS-RM is silent on what a SENDER (RMS or RMD) is expected to do when it either fails to send one of the messages or does not receive the corresponding response message (e.g. an RMS sends a CreateSequence message but does not receive a CreateSequenceResponse message).

R0110 When a SENDER fails to successfully send a Sequence Lifecycle Message or it does not receive the corresponding response message (if one exists), it is RECOMMENDED that the SENDER attempt to resend the message. The frequency and number of these retries are implementation dependent. TBD

3.8.3 Message Identity

In cases where wsa:MessageID is being used, retransmission must not alter its value, because other headers (possibly occuring in other messages - such as wsa:RelatesTo ) may rely on it for message correlation.

R0120 For any two ENVELOPES that contain WS-RM Sequence headers in which the value of their wsrm:Identifier and wsrm:MessageNumber elements are equal, it MUST be true that neither of the envelopes contains a wsa:MessageID or that both messages contain a wsa:MessageID and the value of the wsa:MessageID elements are equal. TESTABLE RSP0120

3.9 Piggybacking

WS-ReliableMessaging allows for the addition of some WS-RM-defined headers to messages that are targeted to the same endpoint to which those headers are to be sent; a concept it refers to as "piggybacking". There are a number of interoperability issues with the practice of piggybacking.

3.9.1 Endpoint Comparison for Piggybacked SequenceAcknowledgment Headers

Because there is no standard mechanism for comparing EPRs, it is possible for different implementations to have dissimilar assumptions about which messages are and are not valid carriers for piggybacked SequenceAcknowledgement headers. For example, an implementation of the RMS may assume that the ReferenceParameters (if any) of the EPRs will be compared as part of the determination of whether a message is targeted to "the same" endpoint as the AcksTo endpoint. Meanwhile an implementation of the RMD may assume that a simple comparison of the Address IRIs is sufficient for making this determination. This creates the possibility for misdirected, dropped, and otherwise lost acknowledgements to the detriment and possible malfunctioning of the WS-RM protocol.

R0500 An RMD MUST, at a minimum, perform a simple string comparison algorithm, as indicated in the RFC 3987 section 5.3.1, of the respective wsa:Address IRIs before piggybacking a SequenceAcknowledgement Header onto another message. NOT_TESTABLE COM0500

R0501 In cases where the AcksTo EPR of a Sequence has an Address value equal to the WS-Addressing 1.0 Anonymous URI, the RMD MUST also limit piggybacking as described in section 3.9 of the WS-ReliableMessaging specification. TESTABLE RSP0501

These requirements establish a minimum baseline for an RMD to correctly piggyback SequenceAcknowledgement headers. Individual RMD implementations may choose to consider and/or compare additional elements of the EndpointReference (e.g. the value of any ReferenceParameters elements).

3.9.2 Treatment of ReferenceParameters in AcksTo EPRs

There exists an interoperability problem for Sequences in which the AcksTo EPR contains ReferenceParameters. According to the processing rules defined by Web Services Addressing 1.0 - SOAP Binding , the RMS should expect that any acknowledgements for the Sequence will be accompanied by the contents of the wsrm:AcksTo/wsa:ReferenceParameters promoted as headers in the message carrying that acknowledgement. However, in the case of piggybacked acknowledgments, the carrier message's [destination] EPR may contain Reference Parameters that conflict in some way with the wsrm:AcksTo/ReferenceParameters.

R0510 If the algorithm used by the RMD to determine if a SequenceAcknowledgment can be piggybacked onto another message does not include a comparison of the value of the ReferenceParameters element (when present), then the RMD MUST NOT piggyback SequenceAcknowledgement headers for Sequences in which the AcksTo EPR contains ReferenceParameters. NOT_TESTABLE COM0500

This requirement ensures any RMS implementation that includes ReferenceParameters in its AckTo EPRs of the following invariant: regardless of whether or not the acknowledgments for such Sequences are piggybacked, any message containing the SequenceAcknowledgement header(s) for such Sequences will also contain the AcksTo/wsa:ReferenceParameters in its SOAP headers. Note, this requirement applies equally to Sequences for which AcksTo/wsa:Address is anonymous and Sequences for which AcksTo/wsa:Address is not anonymous.

3.9.3 Preventing Piggybacked Acknowledgements

In situations where an RMD exercises the opportunity to piggyback most or all of the wsrm:SequenceAcknowledgement headers for a particular Sequence to an RMS which does not support the processing of piggybacked acknowledgments, it is likely that the operation of the WS-RM protocol will be severely impacted. This situation can be avoided if the RMS takes steps to ensure that the AcksTo EPRs for any Sequence's it creates are sufficiently unique as to cause the RMD to rule out the possibility of piggybacking acknowledgments for these Sequences.

R0520 An RMS that does not support the processing of piggybacked SequenceAcknowledgement headers MUST differentiate the AcksTo EPRs for any Sequence's it creates from other EPRs. NOT_TESTABLE

The term "differentiate" in the above requirement refers to the process of altering the information in the EPR in such a way as to cause the RMD to rule out the possibility of piggybacking acknowledgments for these Sequences while preserving the RMDs ability to connect to the proper transport endpoint. For example, suppose a particular instance of a web services stack maintains a generic, asynchronous callback facility at http://b2b.foo.com/async/AsyncResponseService. In general, all the EPRs minted by this instance for the purpose of servicing callbacks will have this URI as the value of their wsa:Address element. However, if this web services stack does not support the processing piggybacked acknowledgements, the use this value in the AcksTo EPR creates the potential for the problem described above. The RMS implementation of this web services stack could fulfill this requirement by specifying http://b2b.foo.com/async/AsyncResponseService?p={ unique value } as the address of the AcksTo EPR for any sequences it creates. Since each sequence has a "different" AcksTo EPR (as defined by R0500) from all the other services listening for callbacks, no RSP 1.0 compliant RMD will piggyback acknowledgements for these sequences, though each RMD (in the case of SOAP/HTTP) will correctly connect to http://b2b.foo.com and POST to /async/AsyncResponseService.

3.9.4 Conflicting Requirements for wsa:Action

Points (2) and (3) of Section 3.3 of the WS-ReliableMessaging state that:

2. When an Endpoint generates an Acknowledgement Message that has no element content in the SOAP body, then the value of the wsa:Action IRI MUST be: http://docs.oasis-open.org/ws-rx/wsrm/200702/SequenceAcknowledgement

3. When an Endpoint generates an Acknowledgement Request that has no element content in the SOAP body, then the value of the wsa:Action IRI MUST be: http://docs.oasis-open.org/ws-rx/wsrm/200702/AckRequested
However, this text does not take into account the possibility of piggybacking either of the above RM headers on messages with empty SOAP Bodys that contain wsa:Action values necessary to the proper processing of those messages. Such Envelopes could be the result of a WSDL that contains a doc-literal description where the value of the parts attribute of soap:body is an empty string. To clarify the expected behavior of WS-RM nodes under these circumstances, this profile makes the following requirement:

R0530 In cases where the SequenceAcknowledgement or AckRequested header is piggybacked, then the wsa:Action value of the ENVELOPE MUST be as defined by Section 3.3 of the WS-ReliableMessaging specification if, and only if, the wsa:Action value has not been agreed upon by some other mechanism (e.g. WSDL). TESTABLE RSP0530

3.9.5 Use of the mustUnderstand Attribute

Since they are not allowed to interfere with the processing of messages, piggybacked SequenceAcknowledgement and AckRequested SOAP header blocks must not have the mustUnderstand attribute set to a value of true. However, when the SequenceAcknowledgement and AckRequested SOAP header blocks are sent on messages with an empty SOAP body element and a wsa:Action SOAP header block with a corresponding value of http://docs.oasis-open.org/ws-rx/wsrm/200702/SequenceAcknowledgement or http://docs.oasis-open.org/ws-rx/wsrm/200702/AckRequested (i.e. not piggybacked), implementations are advised to set the mustUnderstand attribute on the SequenceAcknowledgement and AckRequested SOAP header blocks to a value of true. This ensures that these headers are not ignored and avoids the resulting unnecessary retransmissions.

R0540 SENDERs MUST NOT set the value of mustUnderstand attribute on AckRequested and SequenceAcknowledgement SOAP header blocks to true ("1") when those headers are piggy-backed on outgoing MESSAGEs. TBD

4. Secure Conversation

The Profile includes the use of WS-SecureConversation to request and issue security tokens and to broker trust relationship.

This section of the Profile incorporates the following specifications by reference, and defines extensibility points within them:

All requirements in Section 4 apply only when WS-Security is used to secure a message. All requirements in Sections 4.1 through 4.4 apply only when WS-SecureConversation is used to secure a message.

4.1 Unsupported Context Tokens

4.1.1 Unrecognized Extensions in a Security Context Token

During the establishment of a security context, it is possible for a participant to obtain an SCT that, for some reason, it chooses not to accept. One such possible reason is the presence of unrecognized extensions which, by definition, may indicate unknown and possibly harmful semantics. If the RECEIVER chooses to accept such an SCT, however, it must preserve this unrecognized content or nodes that understand and depend on this content may break.

R1000 A RECEIVER MAY not accept an SCT due to unrecognized extensions in exception to R0001. TESTABLE RSP1000

R1001 If a RECEIVER obtains an SCT containing content it does not recognize, the RECEIVER MUST preserve this unrecognized content in all subsequent use of the token. TBD

R1001 goes beyond R0001 (which does not require preservation of unrecognized content) to bring forward requirements from the WS-SecureConversation specification. R1000 has precedence over R1001 since an INSTANCE which faulted due to unrecognized content would not subsequently use the relevant token.

4.2 Demonstrating Proof of Possession

The following requirements describe how, for ENVELOPEs carrying a wst:RequestSecurityToken, the SOAP Body and crucial headers, specified in Section 4.5 and Section 6, must be signed.

4.2.1 Amending Contexts

R1100 An ENVELOPE containing a wst:RequestSecurityToken in the SOAP Body and an action URI of http://docs.oasis-open.org/ws-sx/ws-trust/200512/RST/SCT/Amend, MUST also contain a wsse:Security header with a ds:Signature child element that covers the SOAP Body and crucial headers as specified in Sections 4.5 and 6. TESTABLE RSP1100

R1101 In an ENVELOPE, the signature referred to in R1100 MUST be created using the key associated with the security context that is being amended. NOT_TESTABLE COM1101

4.2.2 Renewing Contexts

R1110 An ENVELOPE containing a wst:RequestSecurityToken in the SOAP Body and an action URI of http://docs.oasis-open.org/ws-sx/ws-trust/200512/RST/SCT/Renew, MUST also contain a wsse:Security header with a ds:Signature child element that covers the SOAP Body and crucial headers as specified in Sections 4.5 and 6. TESTABLE RSP1110

R1111 In an ENVELOPE, the signature referred to in R1110 MUST be created using the key associated with the security context that is being renewed. NOT_TESTABLE

4.2.3 Cancelling Contexts

R1120 An ENVELOPE containing a wst:RequestSecurityToken in the SOAP Body and an action URI of http://docs.oasis-open.org/ws-sx/ws-trust/200512/RST/SCT/Cancel, MUST also contain a wsse:Security header with a ds:Signature child element that covers the SOAP Body and crucial headers as specified in Sections 4.5 and 6. TESTABLE RSP1120

R1121 In an ENVELOPE, the signature referred to in R1120 MUST be created using the key associated with the security context that is being canceled. TBD

4.3 Claims Re-Authentication

4.3.1 Re-Authenticating Claims

As per section 5 of the WS-SecureConversation specification, the request to renew a security context must include the re-authentication of the context's original claims. It is recommended, but not required, that the claims re-authentication be done in the same manner as the original token issuance request. This creates the potential for some implementations of WS-SecureConversation to attempt claims re-authentication in a manner different than the original token issuance request, to the obvious detriment of both interoperability and security.

R1200 When a SENDER makes a request to renew a security context, it MUST re-authenticate the original claims in the same way as in the original token issuance request. TESTABLE RSP1200

4.4 Referencing Security Context Tokens

4.4.1 Associating a Security Context

Section 8 of WS-SecureConversation states that references to an SCT from within a wsse:Security header, a wst:RequestSecurityToken element, or a wst:RequestSecurityTokenReponse element may be either message dependent or message independent. However, references to SCTs from outside a wsse:Security header (or an RST, or an RSTR) must be message independent. Since message independent references provide a superset of the functionality of message dependent references, and it is simpler to support one mechanism for referencing SCTs than two, this profile includes the following requirement:

R1300 In an ENVELOPE that contains either a wsse:Security header, a wst:RequestSecurityToken element, or a wst:RequestSecurityTokenReponse element in which there are references to wssc:SecurityContextToken elements, such references MUST be message independent (i.e. MUST use a wsse:Reference to the wssc:Identifier element). TESTABLE RSP1300

4.4.2 Derived Token References to Security Contexts

Section 7 of the WS-SecureConversation specification describes a mechanism for using keys derived from a shared secret for signing and encrypting the messages associated with a security context. The wssc:DerivedKeyToken element is used to express these derived keys. WS-SC states that the /wssc:DerivedKeyToken/wsse:SecurityTokenReference element SHOULD be used to reference the wssc:SecurityContextToken of the security context who's shared secret was used to derive the key. This creates an interoperability issue because it leaves open the possibility for a derived key to either lack any relationship between the shared secret or for this relationship to be expressed by some mechanism other than a wsse:SecurityTokenReference.

R1310 When an ENVELOPE contains a wssc:DerivedKeyToken, the wsse:SecurityTokenReference element MUST be used to reference the wssc:SecurityContextToken of the security context from which they key is derived. TESTABLE RSP1310

To properly and interoperably process derived keys it is necessary to relate the key to the shared secret from which it is derived. There are no alternatives to using wsse:SecurityTokenReference's that are consistent with WS-Security.

4.5 Addressing Headers

4.5.1 Protecting Addressing Headers

Since the semantics of the WS-SecureConversation protocol are dependent upon the value of various WS-Addressing headers, ensuring the proper functioning of WS-SecureConversation requires protecting the integrity of these headers. These requirements are not specific to the use of WS-SecureConversation. They also apply whenever WS-Security is being used in conjunction with WS-Addressing.

R1400 When present in an ENVELOPE in which the SOAP Body in that ENVELOPE is signed, each of the following SOAP header blocks MUST be included in a signature: wsa:To, wsa:From, wsa:ReplyTo, wsa:Action, wsa:FaultTo, wsa:MessageId, wsa:RelatesTo. TESTABLE RSP1400

R1401 In an ENVELOPE, the signature(s) referred to in R1400 MUST be coupled cryptographically (e.g. share a common signature) with the message body. TESTABLE RSP1401

R1402 When present in an ENVELOPE in which the SOAP Body in that ENVELOPE is signed, SOAP Header blocks with the wsa:isReferenceParameter attribute MUST be included in a signature for their designated SOAP role. TESTABLE RSP1402

R1403 In an ENVELOPE, the signature(s) referred to in R1402 MUST be coupled cryptographically (e.g. share a common signature) with the message body. TESTABLE RSP1403

5. MakeConnection

The Profile includes the use of WS-MakeConnection to transfer messages using a transport-specific back-channel.

This section of the Profile incorporates the following specifications by reference, and defines extensibility points within them:

The requirements and supporting text in this section make use of the following terms:

5.1 Guidance On the Use of MakeConnection

This section describes how, when wsmc:MakeConnection is used, WSDL input and output messages correspond to SOAP envelopes containing a request or a response sent over HTTP.

5.1.1 Action Values

The WS-MakeConnection specification, while not formally requiring the use of WS-Addressing headers, neglects to mention what the wsa:Action and soapAction URIs should be - when needed.

R2030 If an ENVELOPE contains a wsmc:MakeConnection element the child of the SOAP Body, the wsa:Action header, if present, MUST contain the value "http://docs.oasis-open.org/ws-rx/wsmc/200702/MakeConnection". TESTABLE RSP2030

R2031 If a MESSAGE contains a SOAP 1.1 envelope with the wsmc:MakeConnection element as the child of the Body, the HTTP SOAPAction header, if present and not equal to the value of "" (empty string), MUST contain the value "http://docs.oasis-open.org/ws-rx/wsmc/200702/MakeConnection". TESTABLE RSP2031

R2032 If a MESSAGE contains a SOAP 1.2 envelope with the wsmc:MakeConnection element as the child of the Body, the action parameter of the HTTP Content-Type header, if present, MUST contain the value "http://docs.oasis-open.org/ws-rx/wsmc/200702/MakeConnection". TESTABLE RSP2032

5.1.2 Binding to HTTP

Consider the case of a non-addressable client NA invoking an addressable service B that supports a WSDL request-response operation. A possible request-response exchange involving wsmc:MakeConnection might take the following form:

  1. Through some mechanism, NA is provided the EPR for B .
  2. NA sends a request message (SOAP envelope included in the HTTP entity body of the HTTP request) to B . The request message corresponds to and is described by the input message in the WSDL request-response operation supported by B .
  3. If B chooses not to send the application response on the current back channel, then B sends an HTTP response (via the HTTP back channel) with a status code of "202 Accepted". No SOAP envelope is included as part of this HTTP response.
  4. NA sends a MakeConnection request message (SOAP envelope included in the HTTP entity body of the HTTP request) to B containing a wsmc:MakeConnection element containing the same MakeConnection anonymous URI identifying NA as in the wsa:ReplyTo addressing property contained in the SOAP envelope of the original request message.
  5. B sends a response message (SOAP envelope included in the HTTP entity body of the HTTP response) via the HTTP back channel. The response message corresponds to and is described by the output message in the WSDL request-response operation supported by B . If the HTTP response in step 5 does not contain a SOAP envelope, and if there is no failure, then the HTTP response must not contain an entity body and the status code must be 202.

Note: If the HTTP response from step 5 above does not contain a response message corresponding to the output message in the WSDL request-response operation, NA repeats step 3 above until a response message corresponding to the output message in the WSDL request-response operation is retrieved from B , as described in step 4.

Now consider the case of an addressable client A invoking a non-addressable service NB . Another possible message exchange involving wsmc:MakeConnection might take the following form:

  1. Through some mechanism, A is provided an EPR for NB - this EPR uses an instance of a MakeConnection anonymous URI that identifies NB , and NB is provided the EPR for A .
  2. NB sends a MakeConnection request message (SOAP envelope included in the HTTP entity body of the HTTP request) to A containing a wsmc:MakeConnection element containing the same MakeConnection anonymous URI identifying NB as in the EPR for NB .
  3. A sends a response message (SOAP envelope included in the HTTP entity body of the HTTP response) via the back channel. The response message corresponds to and is described by the input message in the WSDL operation supported by NB .
  4. In the case of a request-response operation, NB sends a request message (SOAP envelope included in the HTTP entity body of the HTTP request) to A . The request message corresponds to and is described by the output message in the WSDL request-response operation supported by NB .
  5. A sends an HTTP response via the back channel with a status code of 202 Accepted. No SOAP envelope is included as part of the HTTP response.

Note: In step 3 above, if NB encounters an infrastructure level fault resulting from the processing the response message (that corresponds to and is described by the input message in the WSDL operation supported by NB ), NB will send the fault to A via a separate HTTP request. Notice, this fault message replaces the output message in the WSDL request-response operation, if any, supported by NB .

A non-addressable endpoint may use wsmc:MakeConnection in a SOAP envelope to obtain any pending messages from an endpoint.

The MakeConnection specification does not mandate how long an MCReceiver needs to wait for an outgoing message to be generated - this is left as an implementation choice. For example, in some environments if there is no message ready to be sent back to the MCSender then returning an HTTP 202 immediately might be appropriate, while in some other cases waiting a certain period of time might improve performance with respect to network traffic. Either case could occur.

When the SOAP request-response MEP is in use and the client is non-addressable the general rules for binding SOAP envelopes to HTTP requests messages (as described by the Basic Profile) apply. SOAP envelopes, that are described by the input message of the WSDL operations supported by a service, are bound to HTTP request messages. SOAP envelopes, that are described by the output message of the WSDL operations supported by a service, are bound to HTTP response messages. For non-addressable services the situation is reversed; the SOAP envelopes, that are described by the input message of the WSDL operations supported by the service, are bound to HTTP response messages and SOAP envelopes, that are described by the output message of the WSDL operations supported by the service, are bound to HTTP request messages.

The following requirements extend the requirements defined in Basic Profile:

R2004 When the wsa:ReplyTo addressing property of a request message (SOAP envelope included in the HTTP entity body of the HTTP request) described by the input message of a WSDL request-response operation is set to a MakeConnection anonymous URI, the corresponding response MESSAGE (SOAP envelope included in the HTTP entity body of the HTTP response) described by the WSDL output message of the same WSDL request-response operation MUST be sent as an HTTP response to either the HTTP request that carried the WSDL input message, or to the HTTP request that carried a wsmc:MakeConnection message with the correct MakeConnection anonymous URI. TESTABLE RSP2004a RSP2004b

R2005 Any MESSAGE resulting from the processing of a SOAP envelope included in the HTTP entity body of the HTTP response, if transmitted, MUST be sent via a new HTTP request. TESTABLE RSP2005

5.2 MakeConnection Addressing

In section 3.1 of the WS-MakeConnection specification the WS-MC Anonymous URI is defined to uniquely identity anonymous endpoints and to signal the intention to use the MakeConnection protocol to transfer messages between the endpoints. The WS-MakeConnection protocol uses the receipt of the MakeConnection message at an endpoint as the mechanism by which the back-channel of that connection can be uniquely identified. Once identified, the MC Receiver is then free to use that back-channel to send any pending message targeted to the URI specified within the MakeConnection message.

5.2.1 Addressing Variants

The WS-MakeConnection specification defines two distinct ways for the MC-Sender to indicate its messages of interest. One of these mechanisms uses the wsmc:MakeConnection Anonymous URI, the other uses a WS-RM Sequence ID. However, the WS-MakeConnection specification doesn't define any way of advertising or agreeing upon which variant of the MakeConnection protocol is supported or required by an endpoint. This creates the potential for different, incompatible implementations of WS-MakeConnection. To promote interoperability this Profile refines the WS-MakeConnection specification with additional requirements to mandate the use of a single, consistent addressing variant. Since the URI variant of WS-MakeConnection is a superset of the functionality of the Sequence-ID variant, use of the URI variant is mandated by this Profile.

R2100 If an ENVELOPE contains a wsmc:MakeConnection element as a child of the SOAP Body, the wsmc:MakeConnection element MUST contain a wsmc:Address child element. TESTABLE RSP2100

R2101 If an ENVELOPE contains a wsmc:MakeConnection element as a child of the SOAP Body, the wsmc:MakeConnection element MUST NOT contain a wsrm:Identifier child element. TESTABLE RSP2101

5.2.2 MakeConnection Anonymous URI

The following requirements describe how the MakeConnection anonymous URI is used in the various addressing properties and within RM protocol elements transmitted on SOAP messages.

R2110 When present in a SOAP ENVELOPE, the /wsmc:MakeConnection/wsmc:Address element MUST be set to a MakeConnection anonymous URI that identifies the MC-SENDER. TESTABLE RSP2110

R2111 Once the MakeConnection protocol is established through the exchange of an EPR that contains the wsmc:MakeConnection Anonymous URI as its [address] property, the MC-RECEIVER MUST make use of the MakeConnection response channel to transfer messages targeted to that EPR. TESTABLE RSP2111

R2112 A MESSAGE sent to a non-addressable endpoint MUST have the wsa:To addressing property set to an instance of the MakeConnection anonymous URI that identifies that endpoint, except in the following situation where this is [permitted but] not required (a) the message (that is not a WS-RM lifecycle message) is sent non-reliably over the back-channel of an underlying protocol connection initiated by the non-addressable endpoint. TESTABLE RSP2112

R2113 When referring to a non-addressable endpoint, and if present in a SOAP ENVELOPE, the /wsrm:CreateSequence/wsrm:Offer/wsrm:Endpoint element MUST be set to an instance of the WS-MakeConnection anonymous URI. TESTABLE RSP2113

6. Secure Reliable Messaging

This section of the Profile contains requirements that address the composition of reliable and secure messaging.

This section of the Profile incorporates the following specifications by reference:

6.1 Initiating a Secure Sequence

6.1.1 Secure Context Identification

Section 5.2.2.1 of the WS-ReliableMessaging specification states that "During the CreateSequence exchange, the RM Source SHOULD explicitly identify the security context that will be used to protect the Sequence". This leaves open the possibility for RMS implementations that, for some reason, attempt to use WS-SC to secure their Sequences in some manner that does not explicitly identify the security context that will be used to protect the Sequence (e.g. by some out of band understanding of an inferred security context). This possibility creates an obvious operational and interoperability issues since (a) point-to-point, out-of-band configuration creates unscalable operational overhead and (b) not all WS-RM implementations may be capable of supporting such understandings.

Within Section 6, the phrase "secure Sequence" is defined as "a Sequence beginning with an exchange in which the wsrm:CreateSequence element has been extended with a wsse:SecurityTokenReference element." This profile does not cover the out-of-band understandings mentioned just above.

6.1.2 Security Token References

When initiating a secure Sequence, an RMS must ensure that the RMD both understands and will conform to the requirements listed above.

R3010 If an ENVELOPE contains a wsrm:CreateSequence element as a child of the SOAP Body, and the proposed Sequence is to be secured, the ENVELOPE MUST also include the wsrm:UsesSequenceSTR element as a SOAP header block. TESTABLE RSP3010

6.2 Signature Coverage

In a secure Sequence there exists both security and interoperability issues around the inclusion of SOAP message elements within signatures.

6.2.1 Single Signature for Sequence Header and SOAP Body

As discussed in Section 5.1.1 of WS-ReliableMessaging, any mechanism which allows an attacker to alter the information in a Sequence Traffic Message or break the linkage between a wsrm:Sequence header block and its assigned message, represents a threat to the WS-RM protocol.

R3100 When present in an ENVELOPE in a secure Sequence, the wsrm:Sequence header block MUST be included in a signature. TESTABLE RSP3100

R3101 In an ENVELOPE, the signature referred to in R3100 MUST be coupled cryptographically (e.g. share a common signature) with the message body. TESTABLE RSP3101

R3102 In an ENVELOPE, the signature referred to in R3100 MUST be created using the key(s) associated with the security context that protects the applicable Sequence. TBD

6.2.2 Signed Elements

As discussed in Section 5.1.1 of WS-ReliableMessaging, any mechanism which allows an attacker to alter the information in a Sequence Lifecycle Message, Acknowledgement Messages, Acknowledgement Request, or Sequence-related fault represents a threat to the WS-RM protocol.

R3110 If a wsrm:CreateSequence, wsrm:CreateSequenceResponse, wsrm:CloseSequence, wsrm:CloseSequenceResponse, wsrm:TerminateSequence, or wsrm:TerminateSequenceResponse element appears in the body of an ENVELOPE in a secure Sequence, that body MUST be included in a signature. TESTABLE RSP3110

R3111 In an ENVELOPE, the signature referred to in R3110 MUST be created using the key(s) associated with the security context, that protects the applicable Sequence. TBD

R3114 If a wsrm:AckRequested, or wsrm:SequenceAcknowledgement element appears in the header of an ENVELOPE and that element refers to a secure Sequence, that element MUST be included in a signature. TESTABLE RSP3114

R3115 In an ENVELOPE, the signature referred to in R3114 MUST be created using the key(s) associated with the security context that protects the applicable Sequence. TBD

R3117 When using SOAP 1.2, if a soap12:Fault element appears as the body of an ENVELOPE and the fault relates to a known secure Sequence, the soap12:Body MUST be included in a signature. TESTABLE RSP3117

R3118 In an ENVELOPE, the signature referred to in R3117 MUST be created using the key(s) associated with the security context that protects the applicable Sequence. TBD

6.2.3 Single Signature for SOAP 1.1 Fault and SequenceFault Header

As described in Section 4.1 of WS-ReliableMessaging, the wsrm:SequenceFault element is used to carry the specific details any SOAP 1.1 faults generated during the WS-RM-specific processing of a message. As with SOAP 1.2, the integrity of fault information needs to be protected. In addition to this, it is necessary to ensure that the linkage between a wsrm:SequenceFault header and the soap11:Fault body is preserved.

R3120 When using SOAP 1.1, if a wsrm:SequenceFault appears in the header of an ENVELOPE and the fault relates to a known secure Sequence, the wsrm:SequenceFault header MUST be included in a signature. TESTABLE RSP3120

R3121 In an ENVELOPE, the signature referred to in R3120 MUST be coupled cryptographically (e.g. share a common signature) with the message body. TESTABLE RSP3121

R3122 In an ENVELOPE, the signature referred to in R3120 MUST be created using the key(s) associated with the security context that protects the applicable Sequence. TBD

6.3 Secure Use of MakeConnection

This Profile places additional requirements on the composition of MakeConnection, WS-SecureConversation, and WS-ReliableMessaging.

6.3.1 Security Context for MakeConnection

From a security standpoint, it will be commonly desired that the security context of the message sent on the backchannel established by a MakeConnection and that of the MakeConnection message itself be the same. However, it is important to keep in mind that the WS-MakeConnection protocol is independent of the application protocol(s) flowing over it, thus there will be cases in which the MC-SENDER has no knowledge of the security context (if any) of the backchannel messages. For example, the WS-MakeConnection specification details a scenario in which MakeConnection is used to deliver Notifications from an Event Source. The Event Source may have a variety of different security contexts that it uses depending on the type of Notification being delivered. In this case the MC-SENDER has no way of knowing which security context, if any, should to be used. In such situations, the MC-RECEIVER needs to simply ensure that the MC-SENDER is authenticated. It would still be the MC-SENDER's responsibility to ensure that any message sent on the backchannel has the correct security context - just as would any endpoint receiving a message over a new connection.

6.3.2 Signing the MessagePending header

Since the value of the wsmc:MessagePending header effects the operation of the MakeConnection protocol, it must be protected to ensure the proper functioning of that protocol.

R3201 If a wsmc:MessagePending element appears as a header block in an ENVELOPE, that element MUST be signed using the key(s) associated with security context, if any, that protects the SOAP Body of the ENVELOPE. TESTABLE RSP3201

6.4 Replay Detection

As mentioned in Section 5 of WS-ReliableMessaging, there is a potential tension between certain aspects of security and reliable messaging; a security implementation may attempt to detect and prevent message replay attacks, but one of the invariants of the WS-RM protocol is to resend messages until they are acknowledged. Implementations must have the information necessary to distinguish between a valid retransmission of an unacknowledged message and a replayed message.

6.4.1 Unique Timestamp Values

R3300 In the absence of WS-SecurityPolicy assertions that indicate otherwise, an ENVELOPE in a secure Sequence that contains a wsrm:Sequence header MUST contain a wsu:Timestamp as a sub-element of the wsse:Security header. TESTABLE RSP3300

R3301 For any two ENVELOPEs in a particular secure Sequence that contain WS-RM Sequence headers in which the value of their wsrm:MessageNumber elements are equal, it MUST be true that neither of the envelopes contains a wsu:Timestamp as a child element of wsse:Security header, OR that both messages contain a wsu:Timestamp as child elements of their wsse:Security headers and the value of these wsu:Timestamp elements are NOT equal. TESTABLE RSP3301

Appendix A: Referenced Specifications

The following specifications' requirements are incorporated into the Profile by reference, except where superseded by the Profile:

Appendix B: Extensibility Points

This section identifies extensibility points, as defined in "Scope of the Profile," for the Profile's component specifications.

These mechanisms are out of the scope of the Profile; their use may affect interoperability, and may require private agreement between the parties to a Web service.

In Web Services Reliable Messaging 1.2:

In WS-SecureConversation 1.4:

In Web Services Make Connection 1.1:

Appendix C: Normative References

In addition to all of the profiled specifications listed in Appendix A, the following specifications are normatively referenced:

RFC2119 http://ietf.org/rfc/rfc2119 Key words for use in RFCs to Indicate Requirement Levels, S. Bradner March 1997

WS-I Basic Profile 1.0 http://www.ws-i.org/Profiles/BasicProfile-1.0-2004-04-16.html, K. Ballinger et al April 2004

Namespaces in XML 1.0 (Second Edition) http://www.w3.org/TR/REC-xml-names/ T. Bray et al August 2006

WS-I Conformance Claim Attachment Mechanisms Version 1.0 http://www.ws-i.org/Profiles/ConformanceClaims-1.0-2004-11-15.html, M. Nottingham et al November 2004

Appendix D: Acknowledgements

This document is the work of the WS-I Reliable Secure Profile Working Group, whose members have included:

Incomplete List; Doug Bunting (Microsoft) Doug Davis (IBM) Jacques Durand (Fujitsu) Leonid Felikson (Freddie Mac) Chris Ferris (IBM) Robert Freund (Hitachi Ltd.). Marc Goodner (Microsoft) Frederick Hirsch (Nokia) Ram Jeyaraman (Microsoft) Anish Karmarkar (Oracle) Charles LeVay (IBM) Rich Levinson (Oracle) Hal Lockhart (Oracle) Monica Martin (Microsoft) Arnaud Meyniel (Axway) Dale Moberg (Axway) Sanjay Patil (SAP) Gilbert Pilz (Oracle) Tom Rutt (Fujitsu) Daniel Toth (Ford Motor Company) Faisal Waris (Ford Motor Company) Eric Wells (Hitachi Ltd.).