RE: [hyades-dev] Propose changes/additions to Execution team inte rfaces

["Toomey, Joe" <jtoomey@xxxxxxxxxxxx>]

I had a good discussion with Richard about these points.  In short, we
agreed on the first four points immediately, and the last two points will be
taken back to the Execution Team for proposal.  Richard's team will then
review the execution team's proposal and verify that it satisfies their
concerns.

More on this soon.

Thanks,
--Joe


Joe Toomey 
Senior Staff Software Engineer 
Rational Software 
IBM Software Group 
tel: (781) 676-7668 
fax: (781) 676-7640 
-----Original Message-----
From: Richard Duggan [mailto:rduggan@xxxxxxxxxx] 
Sent: Wednesday, April 23, 2003 12:15 PM
To: Hyades-dev@xxxxxxxxxxx
Subject: [hyades-dev] Propose changes/additions to Execution team interfaces



Team, 

As part of the data collection teams efforts to implement the runtime side
functionality of the interfaces dropped by the Execution team we have come
across a couple of issues we would like to have changed if posible.  Here is
the list.   

***  Replacement of the IChannelReader interface with the IDataChannel
interface 
As part of our work of implementing the data channel bindings in Java and
runtime in C, we thought it would be better to make the channels use streams
instead of 'roll your own' handlers.  We came up with the
org.eclipse.hyades.collection.framework.channel. IDataChannel interface.  I
would like to have this replace the IChannelReader interface and move it to
the org.eclipse.hyades.execution.core package.  (As a side note, we intend
to change this new interface slightly so that the streams are 'getters'
instead of 'setters' as this is the wrong paradigm.  You can inspect the
interface in CVS. 

*** Addition of a IProcessConsole interface 
The name may not be correct, but we have a requirement for console
interaction with our IExecutor realizations that map to processes.  We have
defined the org.eclipse.hyades.collection.framework.executor.IProcessConsole
interface for obtaining streams for each of the stdin, stdout, and stderr
native streams.   You can inspect the interface in CVS. 

*** Provide a getter for the IProcessConsole on the IExecutor 
Once a IExecutor relaization has been launched it is ncessary to obtain its
console through some means.  It makes sence to provide a getter on the
IExecutor interface for achieving this.  This is another substantiation for
wanting to rename this. 

*** Make the IOrderedProperty interface extend the clonable interface 
When dealing with IOrderedProperties it often becomes necessary to make a
copy of the original for local modification.  Addition of the clonable
interface would standardize this and save persons from continuously writing
the same copy code. 

*** Addition of a IExecutorComponentStateChangeListener 
We need a means of knowing when IExecutionComponent's chage state.  This
could be achieved with a IExecutionComponentStateChangeListener type of
interface. In it's absence, we have opted to provide a
IExecutionCompletionListener to our Executor implementation.   

*** Removal of the createExecutor method on the IExecutionEnvironment
interface 
I figured I would throw this one in while I was at it.  I don't agree with
this method.  It tighly couples the IExecutor implementation with the
IExecutionEnvironment interface and implies a heirarchy of
IExecutionComponents.  If we do this right we should be able to swap
implementation in and out of the heirarchy. 


Thanks, 

Richard K. Duggan
Problem Determination Enablement
IBM Toronto Laboratory
External: 905-413-2396
Internal: 969-2396