Conversation
Each report: where the idea comes from in the firmware, the change, what was predicted before running it, what happened, why it came out that way, and whether it should be adopted with the pros and cons behind that. Ideas 1 to 4. Three of the four should not be adopted as written, and the most useful result is a negative control that failed: rejecting a repeated path hash costs eighteen points of reach because path hashes are one byte, so 311 nodes share 256 values and legitimate paths collide constantly. The control arm produced 573 transmissions and 4,830 receptions in every sweep, which is what makes the deltas attributable to the firmware. Raw MeshBench exports are under evidence/.
Ideas 03, 05 and 06 produced the same five metrics to the digit from three verifiably different binaries. The harness has stopped applying the arm firmware, so those deltas measure nothing. The control did its job: it reproduced exactly in every sweep, which is what made the duplicate arm values obvious rather than plausible. A discriminating run of two known-different arms with no control is queued to confirm the cause.
Idea 08 ran last, after every suspect arm, and produced distinct numbers. A harness that had stopped applying arm firmware would have given it the stuck value as well. The honest statement is narrower: three arms coincide exactly and four do not, and three changes that truncate a flood at the same depth would produce exactly that on a deterministic simulator. The discriminating run decides it.
The series is sound. Four arms return identical numbers because three of them depend on conditions this workload never creates, not because one binary ran four times, and every run now carries the checksum of each binary it attached to prove it.
Each report: the line of firmware the idea came from, the change, what was predicted before it ran, what happened, why it came out that way, and whether it should be adopted with the reasoning on both sides. Three are worth adopting. Three changed nothing measurable, and in every case the reason is more useful than the number: two depend on a transmit budget this workload never exhausts, and one needs request and response traffic the study does not generate. One is a good mechanism at a bad constant. One cost eighteen points of reach and, in doing so, established that a one byte path hash cannot carry an identity comparison. The control produced 573 transmissions and 4,830 receptions in all eight sweeps.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Eight ideas read from MeshCore at the commit
repeater-v1.17.0points at, each proposed from a specific line of the firmware, pre-registered before running, and run against an unmodified control on the Scotland and Ireland import.Each report carries the mechanism, a diagram, what was predicted, what happened, why, and an adopt or do-not-adopt verdict with the reasoning on both sides.
Worth adopting: 8 on the evidence here, 5 and 7 on reasoning the evidence could not reach.
Not as written: 1 costs nine points of reach, 2 costs 19% more airtime, 3 is unsound at the hash width the protocol uses, 4 needs a bounded constant.
Three arms returned identical numbers. Every run records the SHA-256 prefix of each binary it attached, so that is verifiably four different builds converging rather than one build run four times - see METHOD.md. That provenance was added to the simulator during this work because its absence made the question unanswerable.
The MeshCore branches are local and are not pushed anywhere.