The IMS Log Reader Capture agent actually performs two functions: First it mines the IMS Recovery Log for the Segments explicitly specified or that are part of a DBD explicitly specified for capture and places the "99" records for only committed Units-of-Work into a z/OS System LogStream. Once the captured data has been written to the LogStream by the Capture agent, the zLog Publisher function pushes the committed data directly to Engines using TCP/IP. The Publishing function manages the captured data until it has been transmitted and consumed by Engines, ensuring that captured data is not lost until the Engines, which may operate on other platforms, signal that data has been applied to their target datastores.