OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.

 


Help: OASIS Mailing Lists Help | MarkMail Help

ws-rx message

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]


Subject: RE: RE: [ws-rx] PR007 rough proposal


Stefan,

First off, I don't think PR007 applies only to the case of a desktop client
that needs reliable responses. In a more general sense I think this issue
applies to any RMD that needs to close/terminate an incoming sequence before
the RMS sends either a CloseSequence or TerminateSequence. I'll describe
another scenario:

Suppose an RM-enabled service is contained by an app server that is hosted
on some enterprise-level box. Suppose this box has built-in management
hardware that monitor things like the internal temperature, the power
supply, disk health, etc. Suppose one of these factors goes from green to
orange (not bad enough to crash but bad enough to warrant shutting down).
The OS begins an orderly shut down of all processes. The app server starts
shutting down its services. One of those services is using an RMD that has
active incoming sequences. Ideally you would like the RMD to inform any RMS'
with which it shares an active sequence that it is terminating those
sequences.

Secondly, even in the narrow case of a desktop client that is exiting, your
idea of relating an offered sequence to the sequence that carried the offer
runs counter to the oft-stated consensus of the TC that no such relationship
exists . What ends up happening is that a client that does think the
offerer/offeree relationship exists terminates its outgoing sequence and
then waits for the server (which doesn't think the offerer/offeree
relationship exists) to terminate the incoming sequence. This, of course,
never happens and the client eventually times out with an ugly exception
which causes the user to wonder what went wrong. He or she then calls his or
her help desk which . . . . 

- gp

> -----Original Message-----
> From: Stefan Batres [mailto:stefanba@microsoft.com] 
> Sent: Thursday, November 30, 2006 3:27 PM
> To: Gilbert Pilz; Paul Fremantle; ws-rx@lists.oasis-open.org
> Subject: RE: RE: [ws-rx] PR007 rough proposal
> 
> Gil, Paul
> 
> I don't think we need to do anything here. Let me try to explain why:
> 
> Problem: How does either side of a two-way RM session 
> gracefully end the session?
> Solution: The spec allows this today, but does not prescribe 
> how to do it. This *could* be implemented by terminating the 
> outbound session when the inbound session is terminated. The 
> questions then is: how does either side know that this scheme 
> is in place? I think that is a binding/profile-specific issue 
> and outside the scope of the spec.
> 
> --Stefan
> 
> 
> -----Original Message-----
> From: Gilbert Pilz [mailto:Gilbert.Pilz@bea.com]
> Sent: Thursday, November 30, 2006 12:23 PM
> To: Paul Fremantle; ws-rx@lists.oasis-open.org
> Subject: RE: [ws-rx] PR007 rough proposal
> 
> I agree the PR007 is dependent upon PR022. That being said I 
> think the route to go with this is as follows:
> 
> 1.) Revise the definiations of CloseSequence and 
> TerminateSequence to allow them to flow from the RMD to the 
> RMS (targeted to the AcksTo EPR corresponding to the Sequence 
> that is being closed or terminated).
> 
> 2.) Add language stating that the RMD may indicate to an RMS 
> that it is closing a sequence with no errors (i.e. there is 
> nothing wrong with the RMD and the RMS didn't do anything 
> wrong either) by sending a CloseSequence accompanied by an 
> Ack+Final to the AcksTo EPR.
> 
> 3.) Add language stating that the RMD may indicate to an RMS 
> that it is terminating a sequence with no errors by sending a 
> TerminateSequence accompanied by an Ack+Final to the AcksTo EPR.
> 
> 4.) Close any gaps in fault handling. For example, if a RMD 
> closes a sequence by send a CS and then receives an new 
> Sequence Traffic Message, it should return a SequenceClosed 
> fault, etc.
> 
> - gp
> 
> > -----Original Message-----
> > From: Paul Fremantle [mailto:paul@wso2.com]
> > Sent: Thursday, November 30, 2006 11:43 AM
> > To: ws-rx@lists.oasis-open.org
> > Subject: [ws-rx] PR007 rough proposal
> >
> > Firstly, I believe the solution to PR007 relies on us 
> resolving PR022 
> > first.
> >
> > I think the solution to this is that the "client" should 
> add a Final 
> > marker onto a sequence acknowledgement for the offered sequence and 
> > add this into the TerminateSequence (or other message). 
> This signals 
> > that the offered sequence is closed. If the server receives this I 
> > think it should consider the offered sequence closed.
> >
> > Furthermore, I think that in this case we should state that in this 
> > case:
> >
> > 1) the client need make no further MakeConnection for 
> messages for the 
> > offered sequence.
> > 2) the server should not send a TerminateSequence in this 
> case (unless
> > PR022 makes it mandatory).
> >
> > Paul
> >
> > --
> > Paul Fremantle
> > VP/Technology and Partnerships, WSO2
> > OASIS WS-RX TC Co-chair
> >
> > http://bloglines.com/blog/paulfremantle
> > paul@wso2.com
> > (646) 290 8050
> >
> > "Oxygenating the Web Service Platform", www.wso2.com
> >
> >
> >
> 

smime.p7s



[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]