On Sep 20, 2013, at 7:42 PM, Kingsley Idehen wrote:
First step is always decoupling the data from application code.
What specifically do you mean by "decoupling the data?"
I thought I was explicit in stating that a major challenge of legacy systems is NOT the data, but the systems themselves... COBOL, JCL, CICS, DB2, IMS, DELTA, EasyTrieve, etc.
If you insist here, the "data" is source code. A load module is data.
Then map the data (via views) to an entity relationship model oriented graph comprised of domain entity types, relation types, and actual assertions that describe how entities are associated (related).
When IBM made a run at RepositoryManager in AD/Cycle ran the table count up to 1,700 (the last I saw). Great for selling IBM iron, but simply not practical.
Again there's a huge difference between databases that track data where the RDBMS model is quite useful. But we're NOT tracking classic data here. We need to track artifacts—programs, subroutines, tables, segments, COPYbooks, etc.—and their relationships to each other.
So how quickly can you do this decoupling & entity relationship modeling? Under a month right? Remember, this 826 applications is quite small. It's entry point to the Fortune 350. An outfit like IBM claimed 4,500 applications (no clear idea what an application is) a few years back.