Transactions and Components

David Chappell

A traditional transactional application tells a transaction server that it should begin a transaction, makes changes, then tells the transaction server to commit or abort all of those changes. But this traditional approach doesn't work when the transactions are being performed by components. Why? Remember that the primary goal of component-based development is to allow building applications from independently created parts. If each component were always used alone, the traditional "Begin Transaction, Do Work, Commit or Abort" structure would work just fine. But if you want to combine multiple components into a single transaction, each component cannot contain its own Begin Transaction request.

Microsoft Transaction Server's solution is to disallow a component to determine when a transaction begins. Instead, each component can be administratively configured to require a transaction. When a client creates a transaction-required component, MTS automatically starts a transaction. If that component then commits or aborts the transaction, MTS carries out the component's request. If the component creates another component, and the new component also requires a transaction, MTS can automatically include the changes made by the new component in the transaction. When this second component commits or aborts its work, MTS takes note but doesn't end the transaction. Not until the parent component commits or aborts does MTS end the transaction. This approach allows the same component binary to be used in its own transaction or combined with others into a single transaction.