>Has anyone, Kevin?, got an idea how I can get out of this ????
>
>Marcus

Marcus,

As Paul and Kevin said, this is a definite limitation. The workaround I use is to have a batch table which keeps track of batch processes. Before beginning a batch, I create a new batch record. Then, every table involved in the batch has a batch field, in which I add the key from the batch table. At the end of the batch processing, if the batch was successful, a field in the batch table (I think I call it lSuccess) is assigned .t. If it fails for some reason, I can go back and delete all records from that batch, since they have a batch field.

I'm trying to create a batch object which will manage this whole process, and also check itself at startup to find unfinished batches which need attention. I've modeled much of the behavior, but there are a few issue, such as determining whether an unfinished batch is still in progress or has failed, which have prevented me from implementing this quite yet.

There may well be a better way, but this is the workaround I've been using.

Good luck,

David