Interviewing my mother, a mainframe COBOL programmer (2016)
Mainframe COBOL systems that power core banking and insurance operations remain critical yet fragile infrastructure, heavily dependent on aging experts whose institutional knowledge is hard to replace. Commenters highlight how these legacy environments intertwine obscure business logic, decades of ad‑hoc integrations and dated tooling, making migration risky, underpaid, and often mishandled despite new AI‑assisted modernization tools. The thread also surfaces a broader reflection on the history of programming—especially the underrecognized role of women in early computing—and the importance of preserving both technical and personal stories before they’re lost.
Mainframes, COBOL, and critical banking systems
- Commenters stress how core COBOL/mainframe systems are existential for large banks; loss of a small team could cripple a bank and even affect national economies.
- Some note similar “hidden critical” roles for other niche languages (APL, MUMPS).
- IMS is clarified as a hierarchical (not network/graph) database, historically preceding relational DBs.
- Debate over performance: some claim IBM mainframe tech is “pure legacy” and would be faster on Postgres; others counter that z/OS + DB2 offer unmatched reliability and strong OLTP performance.
Bus factor, institutional knowledge, and outages
- Many focus on institutional knowledge: the valuable asset is not COBOL syntax but understanding decades of business logic and undocumented quirks.
- Examples include hard-coded limits and Y2K-era assumptions that only long-timers knew about.
- Several anecdotes of retirees rehired at high consulting rates when rewrites stalled.
- The TSB UK migration incident is cited as a cautionary tale, likely worsened by lost know‑how.
Compensation, careers, and talent pipeline
- Multiple posters say mainframe/COBOL roles are often paid average or below-average salaries despite their criticality, which worsens the aging‑workforce problem.
- State and bank roles are cited with relatively low pay and little remote work; some suspect real money flows via contractors and multi-layer vendors.
- There’s consensus that if younger programmers were paid “fat paychecks,” the pipeline problem would be smaller.
Legacy systems, migrations, and modernization
- Integrations are described as brittle, batch-oriented, and highly regulated; negotiating a new interface can cost years, so teams reuse existing document formats and spare bytes.
- Replacing cores is viewed as extremely hard due to decades of layered integrations and compliance constraints.
- Some firms migrate COBOL workloads from mainframes to Linux, gaining speed and maintainability without a risky full rewrite.
- New tools using generative AI are mentioned for COBOL→Java help and for auto-documenting legacy codebases; some see AI as a good fit for tasks like IMS→DB2 refactoring.
History, gender, and capturing stories
- Many share family stories of mothers/grandmothers in COBOL, FORTRAN, simulation, and Soviet space work, contrasting with today’s more male-dominated field.
- Early programming is recalled as evolving from clerical/data‑entry work that employed many women.
- Several urge readers to record interviews with older relatives; apps and question sets for oral history projects are recommended, and people regret missed chances or inaccessible old recordings.
Work environment and culture
- Hotdesking and open-plan offices are widely disliked; people want personal desks, even in hybrid setups.
- Descriptions of mainframe work include shared login to central systems, no local dev environment, and reliance on source control tools and separate LPARs for prod/test/dev.
- Older time-sharing practices (terminal messaging, chat commands) are recalled as predecessors to modern tools like Slack.