About
The fifth project in the 3M series.
MMR measures. MMLC records. MLF contains. MMPF chooses. MMRF is where the results are kept, and the first of them to be defined by a refusal.
The series
| Project | What it does | Site |
|---|---|---|
| MMR-Bench | benchmark and audit workpaper | mmr |
| MMLC | ledger runtime; time-indexed transaction chains | mmlc |
| MLF | matrix ledger format; byte-exact containers | mlf |
| MMPF | route-aware auditable factorization runtime | mmpf |
| MMRF | research infrastructure; this site | — |
Why the refusal is the design
MMPF factors integers and is open about it. MMRF holds a prime dataset and is equally open about the fact that a prime dataset with an unrestricted query surface is a different kind of object entirely.
The distance between the two is not a policy document. It is a guard that runs before the data layer, an operation allowlist, a field denylist, a per-session budget, and a governance process in which the safety classification is declared inside the proposal and checked by a reviewer whose only job is that. The boundary is load-bearing.
Release
- Version
- v1.0
- Release ID
MMRF-1.0.0- Licence
- Apache-2.0
- Runtime
- mmrf-runtime
- This site
- mmrf-site
- Publisher
- EveMissLab Technology Co., Ltd.
Changes made after the release was cut
Three, all recorded in the runtime's git history as a single commit against an unmodified import of the published package, so the diff is readable:
family_countsno longer reports a Sophie Germain count. It was sharing a bit mask with safe primes and no Sophie Germain bit was ever written, so the surface returned one real quantity under two names.- An empty shard selection now raises instead of returning an empty aggregate, so a missing index can no longer be mistaken for a range with no primes in it.
- A replayable baseline study was added, because the shipped baseline workflow cannot run from the shipped package.
None of them changed a shard byte, and the stable manifest hash is the same as the one the release published.