Re: [wsbpel] Issue 108 -- parallel compensation handler execution order

Title: Message

On the second thought. I think the current 
spec statements are perfect for default compensation handler.

 

Even though child scopes are in parallel, the 
default compensation handler should invoke the compensation handler for its 
children in the reverse order of completion (which is running them in serial in 
reverse order). if developer want to execute them in parallel, bpel developer 
could call it explicitly in parallel.

 

-muruga

  
----- Original Message ----- 

  
From: 
  Green, Alastair J. 

  
To: wsbpeltc 

  
Sent: Tuesday, March 30, 2004 1:31 
  AM

  
Subject: [wsbpel] Issue 108 -- parallel 
  compensation handler execution order

  

  

  
I believe that the 
  compensation handlers for parallel scopes should be processed in any order, by 
  default. The implication of parallel forward processing is that there is 
  unlikely to be data sharing (i.e. the services involved are unaware of each 
  other); it is reasonable to assume the same on the way out. If there is 
  sharing/awareness, then it’s a case of caveat emptor: don’t disorder the 
  necessarily ordered, or control the disorder very carefully at an application 
  level.

  
 

  
The spec as it stands 
  seems to unambiguously mandate “reverse completion order”, as a previous post 
  from Muruga Chinnananchi pointed out.

  
 

  
I suggest that after 
  the existing text in 13.3 ending:

  
 

  
“If the compensation handler 
  for a scope is absent, the default compensation handler invokes the 
  compensation handlers for the immediately enclosed scopes in the reverse order 
  of the completion of those scopes.”

  
 

  
be inserted the 
  additional text:

  
 

  
“When scopes are 
  executed in parallel, the default compensation handler of their immediately 
  enclosing scope shall invoke the compensation handler of each completed 
  enclosed scope, in an implementation-defined 
  order”

  
 

  
Alastair