Writing
Notes on compiling banking logic to COBOL, modelling money and failure, and evaluating the evidence.
-
A banking language that compiles to COBOL
An overview of BankLang 0.10.0, its compile-time checks, generated COBOL, local evidence, and unvalidated IBM boundary.
-
Why banks still run COBOL, and why rewriting it keeps failing
Why long-lived COBOL systems are difficult to replace, and where a compiler that still emits COBOL fits.
-
Rounding money is harder than it looks
Why decimal scale and explicit rounding matter for financial arithmetic, and how BankLang generates and tests them.
-
Compile-time checks for missing banking invariants
How compile-time checks can catch omitted invariants such as idempotency, audit events, and balanced postings.
-
Reading code you did not write, and did not want
What makes generated COBOL reviewable: stable output, useful comments, source maps, and clear ownership.
-
Testing a compiler whose target you cannot run
How to test a compiler that targets an unavailable runtime, and what local evidence can and cannot show.
These posts introduce the project. New writing will appear when there is a completed technical note; the feed is the best way to follow updates.