This technique reduces the cost of reading the next record of a sequential file in CICS, from hundreds or thousands of machine instructions down to just 3.
It also supports updates, deletions and even insertions of new records, with extraordinary efficiency. It is so efficient that IBM wanted to patent it, but were unable to do so because they only became aware of it shortly before the 12-month time limit expired.
The file can be preloaded into above-the-line storage during PLTPI processing, thus providing all the efficiency advantages right from the first application task. This enables some applications to become not only possible but extremely fast, where they would have been completely impossible using any standard language or access method.
A later enhancement provides for the file to be loaded above the bar instead and retrieves the records using a threadsafe mechanism, so that all the space overhead becomes effectively completely free, while the processing remains completely safe.
The original technique was developed to satisfy the FBI requirement of scanning a file of tens of thousands of suspect names to perform a fuzzy match on every field of every wire transfer into or out of the US after 9/11. That requirement was found to be impossible using any standard access method, so this new technique was developed to achieve it.
Even GETMAINs for Register Save Areas are avoided during calls to the functional routines, resulting in a typical total overhead of less than 40 machine instructions to delete an existing record or insert a new one. And updates incur almost no overhead as long as a couple of simple coding guidelines are followed.
The technique can be applied to any sequential file processing task with minimal modification, enabling implementation in just a single day via a turnkey approach to the file-handling.