[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: Prelim WSRM Minutes for Wednesday F2f
Attached are the prelim minutes for Wed F2F meeting -- ---------------------------------------------------- Tom Rutt email: tom@coastin.com; trutt@fsw.fujitsu.com Tel: +1 732 801 5744 Fax: +1 732 774 5133Title: Preliminary Agenda WSRM Face To Face Meeting
Preliminary Minutes WSRM Face To Face Meeting Wed May 28 1 Introductions1.1 Proposed AgendaWed May 28 ( (
Dial-In number: 877.302.8255
International Dial-In number: 303.928.2609 Conference ID: 4541308 Wednesday: AM: 32 voting members - 16 present, Rel 8 Intermediary
is defined in Soap. Clearer in Soap 1.2 Active vs
passive Intemediary Sunil states we do not have an
issue here. IM changing MessageID
is considered harmful. Is IM changing any WSRM Header field is
considered harmful? Payrits keeping
silent is not acceptable. Need to define
how an intermediary node can behave, in processing headers. Discussion on
ensuring that a participating intermediary is used for both the request and ack path. Message Exchange pattern vs. Header
fields. NAT at TCP layer is transparent to soap
layer, thus it is a non-participating intermediary. PM Thursday: AM: PM Friday: AM: Sunil pointed out that Lunch will
not be served on Friday, and that the afternoon breaks need to be at 1.2 Roll Call
Voting members 31, Quorum 16. Reached Quorum, even without Phone attendees. 1.3 Approval of Prior Minutes<tbd on Thursday or Friday> 2
Wednesday Requirements Issues
Discussion
2.1 Rel 8 - IntermediariesIntermediary is defined in Soap. Clearer in Soap 1.2 Message Exchange pattern vs. Header fields. NAT at TCP layer is transparent to soap layer, thus it is a non-participating intermediary. Are SOAP intermediaries store and forward entries that also implement WSRM over SOAP? Or are they strictly soap entities that are unaware of wsrm Discussion: SOAP 1.2 has a better definition of intermediary. Definitions for participating intermediary participant in the formal message exchange pattern And non-participating intermediary transparent to the soap message exchange pattern If the sender wants an intermediary to do something the sender will have to define it. If the intermediary is active, does it send an id, does it send an ack, Proposed by nokia. Do we want to deal with intermediaries that change the message exchange pattern? The intermediary cannot change the message header field especially the message id. IM changing message id is considered harmful. Is IM changing any wsrm header field considered harmful? Payrits - keeping silent is not acceptable. Need to define how an intermediary node can behave in processing headers. We need to clearly specify the behaviours of the intermediary with respect to the soap header. If you expect the intermediary to impact flow or fault management then you need to specify behaviours of the intermediary. Either you send a message to the intermed which returns an ack and then forwards to the next node. Other solution an intermediary forwards the message to the next node and then. NAT at TCP layer is transparent to soap layer thus it is non-participating intermediary. Paolo: Two concepts end to end relaibity vs point to point both are possible end user defines what type of reliability is supported. Borrowed from ebxml spec. Alan if you allow intermediaries to change the messages then you have to specify the limits on what they can do. Dock: How does the presence of an intermediary impact generation of fault messaged and acks. Only an issue if participating intermediary i.e. can impact M.E.P. Paolo could express intermediaries directly in the Header Scott stay away from routing! Message Exchange Pattern is defined in SOAP 1.2 Could we allow extensions in wsrm for other routing protocols? Sunil says no! Dock start with intermediaries which do nothing at soap layer add capabilities required one by one. Pete one pattern for actor next receiver rm header targeted at ultimate reciever One pattern for both next receiver and ultimate receiver. WS-Rel does not state anything about actor protocol does not use actor. Original author was end to end protocol between ulitimate sender and ultimate receiver;. Intermediaries were transparent. Discussion on what ws rel spec assumes about soap actor. Sysbolcs state from/to attributes overlaps with soap actor. This causes a problem. Could instead use soap actor. Other alternative is to not use it. If an intermediary is not acting in the role specified it cannot touch that header. Soap defined intermediary was intended as pass thru only. Soap intermediary processes only those headers for which it is the actor. Alan rechanging the destination back up receiver when primary path has failed. An intermediary does backup rerouting. In this case, the intermediary would change the path and destination by changing the header to field to another destination. Doug URI message is sent to Doug WS Rel schema does not use soap actor therefore the recipient is the ultimate destination of the soap message. Doug intermediaries are out of scope vs intermediaries must not process the wsrm headers. Suggestion to make use of intermediaries outside of scope of spec. We need to agree on what architectures and what technical models we are supporting. Question on whether the decision on soap actor is precluded by this decision. Pete the ws rel spec disallows intermediaries to ack since we do not allow soap actor attribute. Attempt by Sunil at straw proposal: a) intermediaries out of wsrm scope, soap:actor excluded in schema b) intermediaries out of wsrm scope, soap:actor optional c) describe intermediary behaviour in detail Doug has different straw poll: Schema: Allow optional soap actor vs. Disallow soap actor (as in WS-Rel V1.0) Text: Describe whatever we do in schema Be strong about excluding soap actor use Describe Reliable delivery with intermediaries Describe Reliable delivery with actor to actor Describe Reliable Delivery with end to end (as in WS-Rel V1.0) Sunil dealing with this extra stuff is out of scope for WS-RM Paolo since there is no Routing standard, we could only allow end to end. However this is a big limitation. Suggest limit scope to end to end reliability, and any intermediaries must be transparent to end to end view of protocol. Payrits: Support reliability for end to end. Where one end is ultimate destination, other end is sender. Dock: this leads to intermediaries not being able to change the behaviour of the protocol. Support architecture of end-to end, thus any intermediaries may not change messages. Sunil: this could be an extensibility. Proposed to resolve issue with a new Requirements: · WSRM must only support end to end reliable messages, where one end is the sender, and the other end is the ultimate destination. Because of this there is no need to define intermediaries. Payrits stated another potential requirement: · Spec needs to cover the fault case when the processor of the header is not the ultimate destination. This is not defined in the soap processing model. There were several concerns expressed regarding this additional requirement. Since the system receiving this message is not participating in the protocol, how can we make statement about what that system should do. Sunil moved, Dock Seconded. Resolve rel 8 with following requirement: WSRM must only support end to end reliable messages, where one end is the sender, and the other end is the ultimate destination. Because of this there is no need to define intermediaries. § No opposition, motion to resolve rel 8 passes. 2.2 Rel 004:From Dave Ingham: Synchronous SOAP HTTP binding: We say that a synchronous SOAP HTTP binding is in use if the outbound Reliable Message is sent in the HTTP POST request and the Acknowledgment Message is contained in the corresponding HTTP response message. We refer to this binding as "synchronous" in order to imply the WS-RM exchange inherits the blocking behaviour of the HTTP exchange. Asynchronous SOAP HTTP binding: We say that an asynchronous SOAP HTTP binding is in use if the Acknowledgement Message is sent in a separate HTTP POST request/response exchange from the outbound Reliable Message. We refer to this HTTP binding as "asynchronous" in order to emphasize that the WS-RM message exchanges are not tied to the blocking request/response behaviour of the HTTP transport. Two alternate usage patterns are possible. We say that the "callback" pattern is being used if the Acknowledgement Message is contained in the HTTP POST request of a second HTTP exchange operating in the opposite direction to the one containing the outbound Reliable Message. We say that the "polling" pattern is being used if a second HTTP POST request is issued in the same direction as the one containing the outbound Reliable Message to act as a request for acknowledgement. The Acknowledgement Message is contained in the HTTP response to this request. This polling pattern is expected to be used in situations where it is inappropriate for the sender of reliable messages to receive HTTP requests. Chris Ferris, through Mark Little, suggested using new terms (since http post does not imply blocking if HTTP pipelining is implemented) such as: Http Response Acknowledgement Pattern Callback Acknowledgement Pattern
This led to discussion of new terms for binding (i.e, mapping) soap/wsrm layer to the Transport layer: Response Acknowledgement Pattern Callback Acknowledgement Pattern Polling Acknowledgement Pattern
Each of these involves a different binding of soap/wsrm layer to the underlying transport. Payrits drew the following diagrams on the whiteboard, to illustrate his point about bindings between layers: Figure One way MEP interlayer binding examples Figure Request response MEP inter layer Binding Example 2.3 Rel 09 Pull modelIwasa presented his two slides. Figure Firewall use case for Pull capability Figure Limited system use case for Pull capability It was pointed out that these figures do not show the ack phase of reliable message protocol. Iwasa agreed that that would be required, but was not shown for simplicity. Discussion turned to how to realize the requirement (receiver of Reliable message cannot receive underlying protocol request). Tom R stated this could be treated as a special case of Request/Response MEP at WSRM user level, where the request has a empty body. Payrits stated that another solution could be a special request. Marc G stated that there is more than one way to meet the requirement. Payrits presented a use case for mobile phones: He presented two solutions, one requiring storage by the WSRM user, the other mechanism the storage is done by the WSRM layer. Marc G stated that is we agree to this as a requirement, we could defer on the particular solutions. Possible Requirement: WSRM protocol must allow Reliable messages to be received by an endpoint which cannot accept an underlying protocol request message or underlying protocol one-way message. Question is whether we need a separate requirement from the already accepted requirement for supporting reliable request response MEP. Iwasa asked that we leave this Issue open, to allow time for him to consult his developers. 2.4 Rel 13-22 as Spec Issues:Iwasa suggested moving issues Rel 13 thru rel 22 as spec issues: 2.4.1 Rel 13Agreed to move Rel 13 as spec issue 2.4.2 Rel 14 regarding wsdl definitionsTom Rutt suggested that we either make this a requirement or request a charter clarification: Could change, in the charter: The specification to be created will provide WSDL definitions for reliable messaging and the message formats will be specified as SOAP headers and/or body content. To: The specification to be created will provide message formats specified as SOAP headers and/or body content. Tom Rutt stated that the WSS security people have the same problem. This is not specific to WSRM. Doug Bunting stated that WSDL 1.2 features might be able to do this. Alan suggested leaving it in charter, not including it as a requirements, and marking it in the spec as an item for further study in the specification. Potential resolution: Move this as a Spec level issue, and close with agreement that this will be marked in the spec as an item for further study, (i.e not solved in the first version of the spec. Sunil suggested to leave it open as a spec Issue for now, and not make it a requirement. Agreed to make this a spec issue, and leave open. 2.4.3 Rel 15Alan moved, Iwasa seconded, to resolve this issue with a new requirement: WSRM spec must identify fault cases and WSRM protocol must support the reporting of these identified faults. § No opposition, motion to resolve Rel 15 passes, agree to add new requirement. 2.4.4 Rel 16 -Agreed to move to spec issue 2.4.5 Rel 17 Persistence requirementsAgreed to move configurability of persistence requirements as a spec issue. 2.4.6 Rel 19 -We are no longer using terms synch/asynch. We have defined binding patterns, but have not yet decided which will be included in the specs protocol. Dock moves to make two new requirement. Scott seconded. requirement
for Spec having a solution for response Ack Binding
pattern, for both one-way and request/response MEP. requirement for Spec having a solution for Callback Ack binding pattern, for both one-way and request/response MEP. § No opposition to adding two new requirements Motion passes. Issue still open for discussion of additional requirements. Sunil suggested we add a third requirement. Proposal to have requirement for spec having a solution for Polling Ack Binding pattern, for both one-way and request/response MEP. Pete stated this might be difficult for request/response MEP. Payrits stated he does not need the polling Ack binding pattern for any MEP. No time for further discussion, will continue on Thursday PM. 2.5 Homework:Tom told everyone to read the WS-rel spec, so we can walk thru the doc Thursday AM to find new Requiements and Spec Issues to add to Issue list. Payrits will work on making a new requirements doc. Meeting closed for the day at |
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]