Re: [sca-j] JAVA-143: Proposed resolution

[Mike Edwards <mike_edwards@uk.ibm.com>]

Folks,

I am OK with Simon's proposal to close
this issue with no action.

However, thinking about the scenario,
I believe that the whole thing is bogus in that it 
conflates the concepts of "implementation
instance" and "reference proxy to a service".

In reality, the second of these can
exist without any of the former for the target service.

What I am saying is that a reference
proxy to a service can be created based on static
configuration information from the SCA
Domain and does not in itself require the existence
of any implementation instances of the
component which implements the service.

Thus, such a proxy can be created and
injected at any time once the Domain configuration
is known.

An implementation instance only need
be created at the point where an operation of a service
implemented by that implementation is
actually invoked..  

This means that as long as the SCA Domain
configuration is known, a reference proxy can
*always* be injected when an implementation
instance is initialized.

Now, this does not get us away from
the horrible case of a cyclic dependency between
two implementation classes where each
one invokes operations on the other during intialization.
However, this I think is just the same
as two implementation classes that invoke operations on
each other during regular processing.
 If the code in the classes does not prevent it, then
the two will call each other ad infinitum
and the SCA runtime is going to fail.


The one other case to discuss is where
the SCA Domain configuration is NOT complete at the
point where a component implementation
is initialized.  In this case, it can be literally impossible
to inject a meaningful reference proxy
for the target of the reference since it may not exist in the
domain.  null is
the right thing to inject in this circumstance.  Reinjection comes
into play if
the SCA Domain configuration is later
modified so that the target for the reference appears.


Yours,  Mike.

Strategist - Emerging Technologies, SCA & SDO.
Co Chair OASIS SCA Assembly TC.
IBM Hursley Park, Mail Point 146, Winchester, SO21 2JN, Great Britain.
Phone & FAX: +44-1962-818014    Mobile: +44-7802-467431  
Email:  mike_edwards@uk.ibm.com

From:	Simon Nash <oasis@cjnash.com>
To:	OASIS Java <sca-j@lists.oasis-open.org>
Date:	15/06/2009 11:21
Subject:	[sca-j] JAVA-143: Proposed resolution

---

I would like to revise my proposed resolution to this
issue in the
light of the discusssion and proposed direction for resolving JAVA-131.
If the currently proposed resolution to JAVA-131 is accepted, it would
be valid to inject null for an optional reference and then re-inject
a non-null value later.  In this case, the null value would mean that
no target for the reference is currently available, and the target
would become available later as described by the rules for re-injection
in section 10.17.1 of JavaCAA.

As an example to illustrate this, consider the case of two components
A and B that have cyclic references to each other (as fields or setter
methods), where A's reference to B and B's reference to A are both
optional.  The first bulleted option of JCI50003 would require the
SCA runtime to do something like the following:

  1. create instance of A
  2. inject null value for A's reference to B
  3. drive A's @Init method (A can't invoke B because of the null
reference)
  4. create instance of B
  5. inject null value for B's reference to A
  6. drive B's @Init method (B can't invoke A because of the null
reference)
  7. inject B instance into A's reference
  8. inject A instance into B's reference
  9. make A and B available for service invocations

I think the above works OK, so I am now proposing to close JAVA-143
with no action.

   Simon



---------------------------------------------------------------------
To unsubscribe from this mail list, you must leave the OASIS TC that
generates this mail.  Follow this link to all your TCs in OASIS at:
https://www.oasis-open.org/apps/org/workgroup/portal/my_workgroups.php

---

Unless stated otherwise above:
IBM United Kingdom Limited - Registered in England and Wales with number 741598.
Registered office: PO Box 41, North Harbour, Portsmouth, Hampshire PO6 3AU