>>So, Knuth supports a 12%. I'm certain a 227% would be a big deal, especially adding the safety factor.

There's no doubt that if selected cursor has 254 fields, mdot makes a difference. However, if that 227% difference is measured in microseconds in the GUI of an actual app, perhaps we can agree that people are entitled to say "so what" or to refuse to remedy VFP code that frequently is years or decades old. If we are scientists then we never treat hypotheses as universal truths, and experience/evidence that contradicts our hypotheses is the most valuable, never to be personalized or scorned.

FWIW, using private variables rather than passing parameters to functions also makes a big difference, at the cost of legibility/familiarity. Do you pass parameters? At some point, perfect becomes the enemy of good.

Also FWIW, a defining benefit of VFP used to be its auto-spanning to disk, meaning you could have resultsets that exceed utilized or even available memory in days when a growing .NET dataset could bring a machine to its knees. I was taunted on this point for years by a certain fellow in the UT forums. I never took it personally- just as well, because time went by and machines had more memory than VFP could utilize anyway. Then came devices that blurred the difference between dynamic memory and storage. Along the way, "web is best" hypotheses eliminated the concept of local datasets, though it's back now on devices unlikely to be stressed by any sensible dataset.

Certainly I agree that development remains an expensive cottage artisan industry rather than the automated more efficient models that usually result from technology, but that has been to the financial benefit of many here and change is a'coming.