client side vs server side deliver pre-condition advice
My customer wants to implement a control on new files being created. The work occurs in the RTC client.
1/ Is there a way now to trigger processing at check-in ? That would be the best place to do it, so that the control would be performed as early as possible.
2/ the control that the customer wants to do might be expensive in some cases. I understand that precondition running on the server side run in the context of a transaction. Delaying the commit of the transaction is not good for other transactions. In this case, is it safer to run this control as a client-side deliver pre-condition (if check-in is not available) ?
Thanks,
Nicolas
One answer
as far as I know there is no extension point for check in. I have no solution to that. You would maybe want to make sure that not more than on file goes into a change set and I think you can't. All you could do is check that at deliver.
In general you want to have server based operational behavior, rather than client side, because it is much easier to deploy and maintain (only on the server).
Client side behavior is only needed if the information it acts on is only available on the client. E.g. a syntax checking program or something like that.
I don't think you gain anything in the transaction department, since the transaction spans the client as well as the server.
Comments
Custom client side operational behavior only works in the Eclipse based clients. We don't have a way to extend the other clients. So you would have to make sure it is installed everywhere and fails if not.
Thanks Ralph for the answers.
In this case this is for the Eclipse client only so that would work. And yes I was planning to make it fail if not installed.
The transaction piece is the one that I care about here. I thought that the service call was done after the client pre-condition would run, and therefore a client pre-condition would not be part of the transaction that is driven by the service call and would not block it. But maybe I'm wrong ?
The whole point of a pre condition/advisor is to be able to prevent the transaction from finishing. I don't know all the glory details, but if the advisor fails, the transaction is not performed. Again, a pre-condition is to control behavior and not to change the data of the item.
A follow up action can issue a second save after the transaction finished and is allowed to modify the item this way.