RE: [wsrp-wsia] [I#110] destroyEntity failure semantics

[Gil Tayar <Gil.Tayar@webcollage.com>]

As a patch, maybe :->. As a real solution, I would hate, in these XML-ish
days, to depend on the _message_ for this information. What is the problem
we're trying to solve here? The problem pointed out by Andre, of the
difficulty of an implementation of transactional deletes, and his
proposition that a recurring delete be accepted (and I would add "in the
case of a failure").  I believe the implementation of such a solution by the
producer is acceptable.

-----Original Message-----
From: Rich Thompson [mailto:richt2@us.ibm.com]
Sent: Thu, October 24, 2002 16:31
To: wsrp-wsia@lists.oasis-open.org
Subject: Re: [wsrp-wsia] [I#110] destroyEntity failure semantics







Would a fault message that explicitly started with the number of failures
(i.e. format => [Num] + ' ' + text) be an appropriate solution here?


 

                      Rex Brooks

                      <rexb@starbourne.        To:       Gil Tayar
<Gil.Tayar@webcollage.com>,                     
                      com>
"'wsrp-wsia@lists.oasis-open.org'"                                 
 
<wsrp-wsia@lists.oasis-open.org>                                   
                      10/15/2002 07:21         cc:

                      AM                       Subject:  Re: [wsrp-wsia]
[I#110] destroyEntity failure semantics   
 

 

 




I agree a list of failed attempts is unnecessary, but the return of a
successful attempt is probably going to be mandatory eventually if it isn't
in the first version.

At 8:23 AM +0200 10/15/02, Gil Tayar wrote:
      Status: Active
      Topic: Interface
      Class: Technical
      Raised by: Andre Kramer
      Date Added: 15-Oct-2002
      Section: Interfaces/6.4
      Title: destroyEntity failure semantics
      Description:
      "If a fault message is generated in the processing of
      destroyEntities(), then the producers MUST NOT invalidate any of the
      supplied entityHandles." It would be better if destroy(handle[]) was
      the same as mulitple calls to
      destroy(handle). The producer should attempt to destroy all (even if
      one destroy "fails") and multiple destroys on the same handle should
      be safe.
      <RDT>This was a step forward from the returned list of successfully
      destroyed handles. Sounds like Andre would like the unstructured
      fault message to contain an array of failed destroys ... </RDT>
      <ak> No list need be returned. It should be safe to re-try deletes.
      Main thing is that producers attempt to delete all</ak>



--

Rex Brooks
Starbourne Communications Design
1361-A Addison, Berkeley, CA 94702 *510-849-2309
http://www.starbourne.com * rexb@starbourne.com



----------------------------------------------------------------
To subscribe or unsubscribe from this elist use the subscription
manager: <http://lists.oasis-open.org/ob/adm.pl>