Re: AW: [ws-rx-implement] WS-RM/RX question...

[Doug Davis <dug@us.ibm.com>]

I can go for that - perhaps Stefan would
like to open the issue?  We can use an easy one  :-)
-Doug

Paul Fremantle <paul@wso2.com>

03/01/2006 02:49 PM

To	Doug Davis/Raleigh/IBM@IBMUS
cc	"Rossmanith, Stefan" <stefan.rossmanith@sap.com>, ws-rx-implement@lists.oasis-open.org
Subject	Re: AW: [ws-rx-implement] WS-RM/RX question...

Doug

What about changing the wording from MAY use close to RECOMMEND use 
close? I agree it can't be a MUST.

Paul



Doug Davis wrote:
> Stefan,
>  see comment inline in this pen.
> -Doug
>
>
>
>
> "Rossmanith, Stefan" <stefan.rossmanith@sap.com> 
> 03/01/2006 10:24 AM
>
> To
> Doug Davis/Raleigh/IBM@IBMUS, "Paul Fremantle" <paul@wso2.com>
> cc
> <ws-rx-implement@lists.oasis-open.org>
> Subject
> AW: [ws-rx-implement] WS-RM/RX question...
>
>
>
>
>
>
> Hi Doug, hi Paul,
> thank you for your fast responses!!
> Okay, this sounds good for current version of WS-RX.
>  
> Of course the old WS-RM spec is not subject of this SC, but you were
> involved creating it ;-) and therefore I am interested in your
> answer if you relate my question to WS-RM 2005/02.
>  
> WS-RM 2005/02 spec is the basis of current interoperable RM 
> implementations until WS-RX is ready for beeing the basis for "product"
> implementation.
> Is my interpretation of WS-RM 2005/02 concerning my mentioned topic
> correct? 
>  
> Concerning WS-RX: Your answer
> So good behaviour would be for the RMD to close the sequence not terminate
> it.
>  
> is hopefully written more precise in the spec. If the RMD can decide
by 
> itself to
> terminate or close the sequence, then the problem I mentioned is not
> really solved...
> To my opinion the closing instead of terminating should be a MUST
for 
> RMD?!
> <dug> under normal conditions I agree - a Close would be better
than a 
> Terminate, but I
> don't think we can force it.  There will be times when an RMD
must 
> terminate the sequence
> and not allow any more processing happen related to that sequence.
 I 
> would imagine
> those would only be under the most serious conditions though - normally,
i 
> would hope
> that people would use Close instead.
> </dug>
>  
> But in current WS-RX spec I found:
> 3.2.: To ensure that the Sequence ends with a known final state both
the 
> RM
> Source and RM Destination may choose to 'close' the Sequence before
> terminating it.
> 4.2.:
> This fault is sent by either the RM Source or the RM Destination to
> indicate that it has either encountered
> an unrecoverable condition, or has detected a violation of the protocol
> and as a consequence, has chosen
> to terminate the sequence. The endpoint that generates this fault
should 
> make every reasonable effort to
> notify the corresponding endpoint of this decision.
> What are the arguments to make only such a soft statement in 3.2.
> Why do we really need 4.2. and cannot always use close instead of
> terminate.
> Do you have examples when a closing is not possible but only a 
> termination?
> <dug> perhaps some serious system error prevent us from maintain
state any 
> more? </dug>
> From the very first time I start thinking about WS-RM spec I worried
about 
> the statement in 4.2. but maybe
> this was because I always thought that a terminated sequence must
somehow 
> be continued using a
> new subsequent sequenceID (think about offering the real application
a 
> stable serialization context using
> a sequence of underlying WS-RM sequences which are related. But there
is 
> no relation between WS-RM
> sequences and so this issue cannot be solved on WS-RM level. So in
case of 
> a termination or closing, the
> in-order processing is interrupted and can only be handled using a
> protocol on a higher level or by application
> (e.g. a rollback work).
> Do you think it would be a good proposal to define subsequent WS-RM
> sequences on WS-RM level 
> so that one can continue a sequence (by using a subsequent sequence)
after 
> termination or closing (else only 
> a kind of rollback is possible, isn't it?)? Or is this proposal out
of 
> scope of
> WS-RM in your opinion? (Background: how to offer the real application
a 
> stable serialization context hiding
> sequence terminations and sequence closing by just creating a subsequent
> WS-RM sequence for a given
> closed/terminated WS-RM sequence?)
> <dug> up to now I think people have decided that things like
the linking 
> of sequences or the 
> continuing of sequences is something that would be done at a level
above 
> RM itself.
> If you want you should continue this discussion on the main TC mailing
> list since this isn't really
> about interop anymore but more about the spec itself - and if you
believe 
> that the spec needs
> to be changed in some way I think it makes sense to see how others
in the 
> TC feel or even
> open a new issue to address them.  The worst that can happen
is the issue 
> is closed w/no action
> :-)  </dug>
>
> With kind regards,
> Stefan
>
> Von: Doug Davis [mailto:dug@us.ibm.com] 
> Gesendet: Mittwoch, 1. März 2006 14:15
> An: Paul Fremantle
> Cc: Rossmanith, Stefan; ws-rx-implement@lists.oasis-open.org
> Betreff: Re: [ws-rx-implement] WS-RM/RX question...
>
>
> Also, MaxMessageNumber has been removed from the spec. 
> -Doug 
>
>
>
> Paul Fremantle <paul@wso2.com> 
> 03/01/2006 06:52 AM 
>
>
> To
> "Rossmanith, Stefan" <stefan.rossmanith@sap.com> 
> cc
> Doug Davis/Raleigh/IBM@IBMUS, ws-rx-implement@lists.oasis-open.org
> Subject
> Re: [ws-rx-implement] WS-RM/RX question...
>
>
>
>
>
>
>
>
> Stefan
>
> This is the usage for the CloseSequence, which still allows the RMS
to 
> find out the final state of the sequence. So good behaviour would
be for 
> the RMD to close the sequence not terminate it.
>
> Paul
>
> Rossmanith, Stefan wrote:
>   
>> Hi Doug,
>> I have another question concerning WS-RM/RX.
>> Scenario: The provider decides to terminate a sequence because
of
>> timeout (or other error situations).
>>
>> Is my understanding correct that the RMS in this case does not
know
>> which already sent messages were received successfully by RMD,
because
>> maybe some acks were lost and the RMS cannot request acks after
RMD has
>> terminated the sequence?
>>
>> If this is true, WS-RM only could be used for real applications
if we
>> have in addition a higher level protocol (e.g. a kind of transaction
>> protocol or a mechanism on application level) which can handle
this
>> situation, i.e can proceed after this error situation in an appropriate
>> way.
>>
>> I only want to know if my understanding fits to WS-RM intention
;-)
>>
>> One note to WS-RM 2005/02: In the AckRequest element the spec
tells to
>> use MessageNumber element but the schema tells to use
>> MaxMessageNumberUsed Element...
>> (I looked at the documents at
>> http://msdn.microsoft.com/webservices/webservices/understanding/specs/de
>> fault.aspx?pull=/library/en-us/dnglobspec/html/wsrmspecindex.asp
)
>>
>> With kind regards,
>> Stefan
>>
>>
>> PS: unfortunately I cannot participate tomorrow and next week
at WS-RX
>> SC/TC, because of other dates.
>>
>>
>>
>>
>>     
>
>   

-- 

Paul Fremantle
VP/Technology, WSO2 and OASIS WS-RX TC Co-chair

http://feeds.feedburner.com/bloglines/pzf
paul@wso2.com

"Oxygenating the Web Service Platform", www.wso2.com