What dispensing prescriptions taught me about input validation

The first time I filled a dispensing form in the lab, I thought I was learning pharmacy. I was actually learning input validation — the kind that software still struggles to match.

The validation layers on one prescription

Legibility and completeness. Does the form have every field? A prescription missing a dose isn't an edge case — it's an automatic reject. Software accepts half-empty forms daily; pharmacy refuses them at the door.

Range checking. Doses have plausible ranges. A prescription for 5000mg of a drug whose maximum is 500mg doesn't pass "the number parsed." It fails a clinical range check. My database forms learned the same lesson much later: not just type validation — plausibility validation.

Cross-field consistency. Age versus dose, weight versus volume, drug versus drug. Interactions check whether two fields, each individually valid, are poisonous together. This is the validation layer most web forms never grow, and the most important one in medicine.

The human signature. A valid form from an invalid authority is invalid. Pharmacy checks provenance; software is only now learning to (with signed requests and auth).

What I build differently now

Every form I design has three questions from the dispensing bench: what's complete, what's plausible, what's consistent. Most validation stops at the first. The second is where real data gets caught. The third is where real systems earn trust.