David,
>>Well, the big difference in the above code is aRec is not an object, it's >>an array. An array doesn't have the same mechanism of object references >>that the SCATTER NAME object uses.

I understand it is an array but it was still created with a Scatter command and one could expect the underlying behavior to remain the same when the session object is release, regardless if it were scatter to an array or an object.

>>Do you understand that the session must exist throughout the lifetime of >>the object created inside it? Because these objects can be more complex >>than your simple SCATTER NAME object. and they probably want to commit >>changes back to their datasource at some point.

I understand this now but in light of the behavior of the scatter to an array it make less sense to me why it behaves this way. I mean the array may still need to be committed to the datasource, right? In VFP releasing a datasession behaves one way if you use the 'scatter to' and another way if you use 'scatter name'. Another thing, why does Fabio's solution produce the behavior that I would expect? It uses scatter name oEmptyObjectReference additive, does one simple clause in the scatter command make that much of a difference?

A datasession that is opened when you create a session object will go away when you release the session object Eeeeeeceeeeept when..... Where is this behavior documented? I can't find it, did I miss something somewhere? This makes a lot less sense to me now. It's like playing a game and you think sombody is changing the rules in the middle of the game when infact the rules are being told.

>>Can I ask the bigger scope issue about why this is a problem for you? What >>are you trying to do?

I'm maintaing somebody elses code and to be honest, I have no idea the the goals, responsiblities or intent of this code. But what I do know is when we run the code that is written there is a pretty big memory leak. When I step into the code and start commenting out code that doesn't cause the leak, the last thing that happens when the leak surfaces is a SCATTER NAME command. When I take this line of code out the memory leak goes away. Which lead to my testing of the session object, datasessions and scoping of reference varible and scatter objects.

>>What is the use case for wanting the same object to contain data from >>different data sessions?

To share data across datasessions

David, Thanks for you help and insight

BW