A cautionary tale about flat-file (VSAM) databases on IBM mainframes and the practice of embedding padding fields to allow schema evolution without costly data migrations. When a team of ETL contractors built an extract-transform-load pipeline from a mainframe to a modern RDBMS, they misread the copybook and split padding fields at arbitrary boundaries. The mistake went undetected because they only tested against the production mainframe, missing in-flight features on the dev mainframe that consumed some of that padding. When those features shipped, reports broke. The contractors declared it 'works as designed,' leaving the mainframe team to find alternative padding fields to absorb the new data without disrupting reports.
67.2K Impressions1 Comment