Re: Surely the SSA doesn't need "retired" COBOL programmers
I'm aware of two huge system suites that are running on out of support hardware and costing HMG a fortune to keep the hardware supported - the DWP pensions/payments systems and the MoJ mags court systems.
The DWP systems are the costliest (>30yr old) and the ones they have tried several times, at great cost, to renew/migrate/convert - no its not possible to run in cobol emulation - they've tried to migrate to a newer cobol that could be emulated and failed dismally in that. Amongst other reasons - too much coupling between the code and the underlying machine architecture even though the suite itself was well designed.
IMHO the basic issue is that there is no-one left who understands what the systems actually do - there isn't a comprehensive test suite so the smallest changes take silly lengths of time - and how do you create a test suite for migrated code from scratch?
Yes the only viable approach is to build from scratch but after a reverse engineering of current business to understand what the systems do and confirming thats what they need to do. That's at least a 2 year exercise and will be expensive - the key is deciding to do it - the alternative is to keep paying fujitsu an ever increasing fee to keep ancient circuit boards under 'we'll try' support levels.
The joke is that the underlying systems are the equivalent of a small-mid-range server today.
The mags system took 15 years to build and is a humungous badly designed set of cumulative changes - given its history there is no alternative to scrap and rebuild - a hugely expensive (multi billion) programme was in place in the MoJ to replace all systems - but they focussed on disconnected 'agile' delivery which of course spent lots of time and money on front ends - ignoring architecture and the hard work of understanding what the back ends and hence the business actually did.
So yes 3 years won't hack it - but they have to start sometime.
Biting the hand that feeds IT
