Re: RPGLE I/O opcodes with commit/rollback

On 18-Nov-2015 14:21 -0600, John R. Smith, Jr. wrote:

<<SNIP>>
Just out of curiosity, was the use of this very rare or have I just
not stumbled into the right companies that use it.

Sadly the utilization of Commitment Control (CMTCTL) [in any language] long had been quite rare; more rare than should be, given how many times I had helped customers that would have benefited greatly. More often than not it seemed, most programmers new little about the use of journaling and even less about isolation. And for those who were aware, often their unsubstantiated rationale [i.e. their decision was based on some "they say..." vs any based on actual testing and experience] was that the I/O would perform too slow to justify the benefits :-( I suspect more often, that as with many things, the cost to change is not perceived as worth the effort despite any potential [even overwhelming] value from incurring the cost of making that change; i.e. the core application(s) have long existed and sufficed without, and the tooling has long been built [or exists as templates] to help cleanup the messes that journal and commit either could have helped to prevent or could have provided easier corrective, thus just easier to keep doing the same thing than to try to do things differently even if potentially better.