So let's summarize. All we need to do to use binary floating point arithmetic in place of `DECIMAL` is: 1. Fix all of the "broken" compilers and wait for them to be sufficiently common that we can assume their broad use. 2. Rebuild all of the SQLite-using tools built by those compilers. (e.g. any Windows build of Python built by VC++.) 3. Either get this `epsilon()` into SQLite proper, or make everyone needing this add the extension to their SQLite. 4. Teach all of the world's programmers currently using `DECIMAL` how to use `epsilon()` properly and safely, which as we see from the length of this thread is not trivial. Or... Or! We could just have some form of `DECIMAL` in SQLite.