flymsg

source

Provenance

Public source and introducing commit for each idea already in flymsg. Dates are the commit dates. Sources are published papers, preprints, or public repositories. This table is built from the git history and the existing docs.

A feature enters this table when its commit is on main. The changelog is the release note. This table is the citation.

Date Idea / feature Public source Commit
2026-09-21 MaleCNS v1.0 loader (fetch / build); traced-neuron graph Berg et al., Cell 189, 5504–5526 (2026). doi:10.1016/j.cell.2026.08.015 1919375
2026-09-21 Whole-CNS current-based LIF; Poisson stimulus; info / path / sim Shiu et al., Nature 634, 210–219 (2024). doi:10.1038/s41586-024-07763-9; code philshiu/Drosophila_brain_model (MIT) 1919375
2026-09-21 Fast synapse sign from predicted transmitter (ACh +, GABA/Glu/His −, modulators 0) Shiu et al. 2024 (same as the model); GluCl / HisCl as in that paper 1919375
2026-09-21 Validation battery on known circuits with size-matched controls Shiu et al. 2024; von Reyn et al., Nat. Neurosci. 17, 962–970 (2014). doi:10.1038/nn.3741; Ache et al., Curr. Biol. 29, 1073–1081 (2019). doi:10.1016/j.cub.2019.01.079; King & Wyman, J. Neurocytol. 9, 753–770 (1980). doi:10.1007/BF01205017; von Philipsborn et al., Neuron 69, 509–522 (2011). doi:10.1016/j.neuron.2011.01.011; Gordon & Scott, Neuron 61, 373–384 (2009). doi:10.1016/j.neuron.2008.12.033 10a5afd
2026-09-21 P1 as pC1 subtypes whose synonyms include pMP4/pMP-e Yu et al., Curr. Biol. 20, 1602–1614 (2010). doi:10.1016/j.cub.2010.08.025; Cachero et al., Curr. Biol. 20, 1589–1601 (2010). doi:10.1016/j.cub.2010.07.045 10a5afd
2026-09-21 Criterion v2: reliable spikes, specificity ratio, stability thresholds stated in docs/validation.md before the run; commit that locked the ranking rule fc5c332
2026-09-21 Optional spike-triggered threshold (th_jump) standard spike-frequency adaptation; not a measured fly parameter. Used while w_syn stayed at the FAFB value b6dc609
2026-09-21 Browser view of neuron surfaces, skeletons and neuropils three.js (MIT); geometry from the public MaleCNS volumes in Berg et al. 2026 0b8f4e5
2026-09-21 Stream source-resolution meshes from the public volumes (no local geometry in the export) Berg et al. 2026 data release; three.js Draco loader 5c065c8
2026-09-21 Engine follows the published brian2 schedule; stimulated neurons fire at the input rate Shiu et al. 2024, repository model.py at commit 91bdd1e 3ca2fbe
2026-09-21 Criterion v3: latency order along a chain; dose-response feed-forward sensorimotor chains (von Reyn 2014; von Philipsborn 2011); Shiu et al. 2024 sugar dose (~80 % of peak at 100 Hz) 3ca2fbe
2026-09-21 Sugar GRNs = right LB3b + LB3c (two agreeing lines) Tastekin et al., bioRxiv (2025). doi:10.1101/2025.08.25.671814; Shiu et al. 2024 FAFB sugar set 3ca2fbe
2026-09-21 --dataset fafb: FlyWire v783 in the same schema Dorkenwald et al., Nature 634, 124–138 (2024). doi:10.1038/s41586-024-07558-y; Schlegel et al., Nature 634, 139–152 (2024). doi:10.1038/s41586-024-07686-5; Matsliah et al., Nature 634, 166–180 (2024). doi:10.1038/s41586-024-07981-1; Shiu repository commit 91bdd1e 3ca2fbe
2026-09-21 Synapse-density ratio between MaleCNS and FAFB; scale w_syn comparison of the two public reconstructions (Berg et al. 2026; Dorkenwald et al. 2024); method in docs/comparison.md 3ca2fbe
2026-09-21 Dimorphic / fru-dsx share among responders vs a superclass-matched null sex annotations in Berg et al. 2026; fruitless courtship circuit (von Philipsborn 2011) 3ca2fbe
2026-09-21 Giant Fiber chemical component compared with shak-B² following Blagburn et al., J. Comp. Neurol. 404, 449–458 (1999). PMID 9987990; Allen & Murphey, Eur. J. Neurosci. 26, 439–445 (2007). doi:10.1111/j.1460-9568.2007.05686.x 3ca2fbe
2026-09-21 Criterion v4: class-matched sensory control; bitter → MN9 negative case Shiu et al. 2024, Fig. 3 (bitter suppresses proboscis extension) 771b32a
2026-09-21 Pre-registered model selection (select-model) and 10-seed battery rule written in docs/validation.md before the run 771b32a
2026-09-21 Silencing a set of neurons vs a size-matched null leave-one-set-out / permutation test; empirical p with +1 smoothing 771b32a
2026-09-22 Default w_syn = 0.275 / 1.43, no adaptive threshold (replication of B−) Shiu et al. 2024 weight; density ratio from the MaleCNS–FAFB comparison; rule fixed before the replication 771e116
2026-09-22 Compiled LIF kernel with a NumPy oracle same equations as Shiu et al. 2024; tests/reference_sim.py 771e116
2026-09-22 FAFB LB3 split into male subtypes; MN9 per part Tastekin et al. 2025; Shiu et al. 2024 sugar set 771e116
2026-09-22 Reconstruction audit of bilateral types completeness check on the public tables in Berg et al. 2026 6414c94
2026-09-22 MIT licence, CITATION.cff (no DOI), CI MIT License; Citation File Format 771e116
2026-09-22 T1 / T2 silencing protocol (1,000 matched nulls) rule written in docs/dimorphism.md before the run bc2bc3a
2026-09-23 T1 / T2 results; v0.5 loop half rejected same protocol; findings kept in docs/validation.md f91a777
2026-09-23 compare regenerates the male–female tables Berg et al. 2026; Dorkenwald / Schlegel / Matsliah 2024; Shiu et al. 2024 208dee1
2026-09-23 T3 pre-registered (IN03B024 / vMS12 route) follow-up of T2, written before the run in docs/dimorphism.md 4c7c08a
2026-10-11 N1: one degree-preserving shuffle, criterion v4 on both graphs Maslov & Sneppen, Science 296, 910–913 (2002). doi:10.1126/science.1065103; Newman, SIAM Rev. 45, 167–256 (2003). doi:10.1137/S003614450342480; Shiu et al. 2024 ea70773
2026-10-11 T3 results (1,000 matched-responder nulls) protocol of 2026-09-23 702381a
2026-10-11 T4 pre-registered (AVLP717m), not run written in docs/dimorphism.md before any confirmatory draw 37f48a7
2026-10-11 GitHub Pages site from the README and docs contents already in the repository f886761
2026-10-11 Multi-draw null families, run manifests, versioned preregistration Brunton et al., bioRxiv (2026). doi:10.64898/2026.03.20.713233; Bargmann, BioEssays 34, 458–465 (2012). doi:10.1002/bies.201100185; Maslov & Sneppen 2002; Newman 2003; Rubinov & Sporns, NeuroImage 52, 1059–1069 (2010). doi:10.1016/j.neuroimage.2009.10.003 bf9924a
2026-10-11 T4 result: silencing AVLP717m drops pIP10 after P1 protocol written in docs/dimorphism.md before the run b07dd62
2026-10-11 Held-out manifest, frozen before the run (ORN_DM1 → DM1_lPN decides) Couto, Alenius & Dickson, Curr. Biol. 15, 1535–1547 (2005). doi:10.1016/j.cub.2005.07.034; Kazama & Wilson, Neuron 58, 401–413 (2008). doi:10.1016/j.neuron.2008.02.030; Suh et al., Nature 431, 854–859 (2004). doi:10.1038/nature02980; Jones et al., Nature 445, 86–90 (2007). doi:10.1038/nature05466. Rule in validation/benchmarks.yaml 51b5305
2026-10-11 Held-out result: ORN_DM1 → DM1_lPN fails specific and stability same protocol; finding kept in docs/benchmarks.md. The reported ORN_V case does not decide b3e1859
2026-10-11 Held-out sensitivity corners, both FAIL same protocol; finding kept in docs/benchmarks.md. Neither corner replaces the scored result 23c99df
2026-10-11 One-bit looming-jump protocol, frozen before the run rule in validation/behavior.yaml; circuit already cited in docs/behavior.md 96cee3b
2026-10-11 Closed loop: calibration feasibility FAIL on silence same protocol; finding kept in docs/behavior.md. Takeoff latency does not decide e0b794a
2026-10-11 No biological extension adopted decision in docs/model.md. The gap-junction numbers stay those already in docs/validation.md d4a2914
2026-10-11 TTMn_R spike raster for the one-bit loop times of the v0.8.6 run, drawn in docs/img/looming-jump.png 7dc7a1d
2026-10-11 Extension queue closed as blocked decision in docs/model.md. Issues #10 and #11 add no module d493c0a
2026-10-11 No eligible held-out behavior decision in docs/validation.md and docs/benchmarks.md. No inventory row is promoted 3c7bcf7
2026-10-12 Status prose aligned with the closed record wording in docs/validation.md, docs/model.md, docs/behavior.md and docs/benchmarks.md daefb45
2026-10-12 No scientific task is queued decision in docs/validation.md 35d7a7c
2026-10-12 actions/setup-node 7.1.0 pin in .github/workflows/tests.yml. Node stays 24 43bb246
2026-10-12 numba 0.68.0 and llvmlite 0.50.0 lock in uv.lock. The scored runs were not repeated; their manifests stay on numba 0.67.0 7beecec
2026-10-12 Numeric lock recorded v0.8.13 records the numba 0.68.0 lock. The scored runs were not repeated 32ba973
2026-10-12 Two documentary issues #19 records the FlyBody re-read at d015e9b. #20 keeps the residual TTMn_R partner unnamed. No run bdd7783
2026-10-12 Authored motor map, every row unmapped validation/motor_map.yaml. No published source names a MaleCNS motor neuron with a FlyBody actuator. No run bb1b7e4
2026-10-12 Residual TTMn_R scan pre-registered (version 1) validation/silence_scan.yaml; candidates by rule, screen, fresh-seed confirmation against the T2/T3 null 7bacb61
2026-10-12 Scan runner committed before the run flymsg silence-scan --run 6709f13
2026-10-12 Scan version 3 (replication on seeds 30–32) locked before its run validation/silence_scan.yaml v3; runs start from a commit on main 1bbf60e
2026-10-12 Doc figures as SVG and the scan replays on the site scripts/build_figures.py draws docs/img/*.svg from run outputs; scripts/build_live_scenes.py exports the scan arms (seed 0) v0.8.40
2026-10-12 Archived run outputs in results/; the one-bit raster redrawn as SVG by script scripts/archive_results.py; scripts/build_figures.py reads results/ v0.8.41
2026-10-12 Partner follow-up locked before its run (left mirror, direct drive) validation/partner_followup.yaml v1; the mirror check on seeds 0–2 disclosed in it 39b53a5
2026-10-13 KC → MBON11 depression locked before its run (optional rule) Hige et al. 2015, doi:10.1016/j.neuron.2015.11.003 (Fig 1, 3, S4, S6); validation/mb_plasticity.yaml v1 ce20fce
2026-10-13 Kenyon cell sparseness diagnostic locked before its run Manoim et al. 2022 (KC → KC via mAChR-B); Turner et al. 2008; Honegger et al. 2011; Campbell et al. 2013; validation/kc_sparseness.yaml v1 b096b4a
2026-10-13 Measured viewing direction for every optic-lobe column Zhao et al. 2025, doi:10.1038/s41586-025-09276-5; Nern et al. 2025, doi:10.1038/s41586-025-08746-0; reiserlab/eyemap_T4 at 99d2a43 (GPL-3.0), fetched, not redistributed 413a4c8
2026-10-13 Looming disk coverage per column von Reyn et al. 2014, doi:10.1038/nn.3741; the acceptance angle is stated by the lock 7bf55ae
2026-10-13 Graded photoreceptor input (eye.photoreceptor_kicks) gain stated by the caller; integer kicks reproduce the Poisson run spike for spike 490edb7
2026-10-13 DoOR odor as one response per MaleCNS ORN type Münch & Galizia 2016, doi:10.1038/srep21841; ropensci/DoOR.data at db323a4 (CC BY-SA), fetched, not redistributed ec15377
2026-10-13 Kenyon cell sparseness v1 result: not_recurrence protocol of b096b4a; finding kept in docs/model.md f5cd4b7
2026-10-13 Kenyon cell v2 lock: cholinergic local neuron → PN removed Yaksi & Wilson 2010, Neuron 67, 1034–1045. doi:10.1016/j.neuron.2010.08.041 374309e
2026-10-13 Kenyon cell v3 lock: Hige odors as DoOR vectors, with a lower bound Hige et al. 2015, doi:10.1016/j.neuron.2015.11.003; Münch & Galizia 2016, doi:10.1038/srep21841 c1cf896
2026-10-13 Partner follow-up v1 result: left mirror replicates; DNp06 alone suffices protocol of 39b53a5; finding kept in docs/behavior.md 0510027
2026-10-13 T4/T5 direction selectivity locked before its run Maisak et al. 2013, doi:10.1038/nature12320; Lappalainen et al. 2024, doi:10.1038/s41586-024-07939-3; validation/t4t5.yaml v1 0c574f5
2026-10-13 T4/T5 v1 result: neither same protocol; finding kept in docs/eye.md d874f81
2026-10-13 TTM twitch on NeuroMechFly, jump locked before its run Zumstein et al. 2004, doi:10.1242/jeb.01181; Allen & Murphey 2007, doi:10.1111/j.1460-9568.2007.05686.x; validation/jump3d.yaml v1 c05878d
2026-10-13 NeuroMechFly jump v1 result: short same protocol; finding kept in docs/embodiment.md 3ed5ca8
2026-10-13 Graded (non-spiking) neurons; T4/T5 v2 locked over a release envelope Juusola et al. 2016, doi:10.3791/54142; Zheng et al. 2006, doi:10.1085/jgp.200509470; validation/t4t5.yaml v2 1ea7cd9
2026-10-13 Electrical synapses; Kenyon cell v4 scans eLN–PN coupling Kazama & Wilson 2009, doi:10.1038/nn.2376; sim.Gap stays off unless a lock names it b1116af
2026-10-13 Kenyon cell sparseness v4 result: none protocol of b1116af; subthreshold coupling drains PNs; finding kept in docs/model.md 9a52fe2
2026-10-13 Connectome and NeuroMechFly in one loop; input cut at takeoff M0 clock; spikes before the cut match the open-loop run 78b6c9d
2026-10-13 Spikelets and asymmetric junctions Yaksi & Wilson 2010, Neuron 67, 1034–1045. doi:10.1016/j.neuron.2010.08.041; a spike crosses only when a lock sets spike_mv and spike_ms (one value or one per neuron; 0: that neuron's spikes do not cross). The two sides of a junction may differ 936000d
2026-10-13 Kenyon cell sparseness v5 locked before its run Yaksi & Wilson 2010 Fig. 4C/4F; Kazama & Wilson 2009 Fig. 7d; validation/kc_sparseness.yaml v5, seeds 86–88; no run 00aff9c
2026-10-13 Continuous refractory clamp for new electrical configurations None: an engineering correction under docs/adr-gap-refractory.md; registered locks keep their integrator 409bb34
2026-10-13 Per-neuron membranes (sim.Membrane), off by default None: per-neuron values come only from a lock that sources them 48c58d1
2026-10-13 Postsynaptic weight scale; neuron surface areas from the meshes MaleCNS v1.0 meshes (Berg et al. 2026); the density rule is tested by validation/weight_rule.yaml against Liu et al. 2022 a130380
2026-10-13 MANC muscle cited for each MaleCNS motor neuron Cheong, Eichler, Stürner et al. 2024, eLife 13:RP96084, doi:10.7554/eLife.96084, Supplementary file 3 (sha256 pinned) 1234d53
2026-10-13 Synapse weight rule (count against density) locked before its run Liu et al. 2022 Table S3 (57 pairs of six LHN types matched to MaleCNS) and Table S2; validation/weight_rule.yaml v1, bootstrap seed 139; no MaleCNS count or area computed 3252dbc
2026-10-13 Kenyon cell sparseness v6 locked before its run Eckstein et al. 2024 STAR Methods and Fig. S5F (lLN2T_abc as the cholinergic eLNs; lLN1_bc neighbouring); validation/kc_sparseness.yaml v6, seeds 89–91; no run dbbfcf2
2026-10-13 Kenyon cell sparseness v7 locked: v6 with the boundary reset, qualified scripts/qualify_boundary_reset.py, docs/evidence/kc-boundary-qualification-2026-10-13.json; validation/kc_sparseness.yaml v7, seeds 92–94; v6 interrupted after its uncoupled arms (not read); #155 45c0641
2026-10-13 T4/T5 v3 locked before its run: graded medulla inputs of T4 Groschner et al. 2022, doi:10.1038/s41586-022-04428-3, Fig. 3a, data Edmond doi:10.17617/3.8G v1.0 (CC BY 4.0), fetched by sha256, not redistributed; Behnia et al. 2014 Fig. 3B/3E; validation/t4t5.yaml v3, seed 94; no run 406f622
2026-10-14 Evidence records for sourced parameters (SF1) The sources each record cites; validation/evidence.yaml, flymsg.evidence 0525d93
2026-10-14 Taylor solver for the continuous clamp (flymsg.gap/continuous-clamp-taylor-v1) None: a truncated Taylor action with a float64 roundoff bound, qualified against the dense maps (docs/evidence/gap-taylor-qualification-2026-10-14.json); ADR accepted 2026-10-14 38f560f
2026-10-14 Kenyon cell sparseness v8 locked: v7 under the continuous clamp by the Taylor action scripts/qualify_taylor.py, docs/evidence/gap-taylor-qualification-2026-10-14.json; validation/kc_sparseness.yaml v8, seeds 95–97; v7 result read before the lock ad7b71e
2026-10-14 T4/T5 v4 locked before its run: dark-anchored medulla, lamina gain, measured delays Behnia et al. 2014, doi:10.1038/nature13427 (PMC4243710) Fig. 3B, 3E; Groschner et al. 2022; validation/t4t5.yaml v4, seed 98 8d0660b
2026-10-14 T4/T5 v3 result: robust neither Groschner et al. 2022 (CC BY 4.0) voltages; validation/t4t5.yaml v3, seed 94; finding kept in docs/eye.md 2d0d344
2026-10-14 Kenyon cell sparseness v8 result: v7 replicates under the exact clamp validation/kc_sparseness.yaml v8, seeds 95–97; finding kept in docs/model.md b2161be
2026-10-14 Timestep convergence locked (SF2) and run: converged at 0.1 ms None: a numerical check on the live defaults; validation/convergence.yaml v1, seeds 0–4 7c4d34d, c26f47c
2026-10-14 Pulse vs sine song benchmark locked before its run (#36) Pacheco et al. 2021, doi:10.1038/s41593-020-00743-y, Supp. Table 2, Methods; data doi:10.34770/gv6w-5351 (CC BY 4.0); MaleCNS syn-partners significant-only; Yamakoshi et al. 2025 data 10.17632/29hpr3nrrv; validation/song.yaml v1 2c86f71
2026-10-15 Looming → giant fiber timing locked before its data were downloaded (what an agent had seen recorded) (#34) Dombrovski et al. 2025, doi:10.1038/s41586-025-09037-4, Methods, Extended Data Fig. 8; data Zenodo 10.5281/zenodo.14983850 (CC BY 4.0); Jang et al. 2023, doi:10.1242/jeb.244790, Methods; validation/gf_looming.yaml v1 0579ad8
2026-10-15 Series S GPU parity of the native lane (12 version-1 fixtures, bit for bit) None: the CPU oracle; docs/evidence/series-s-gpu-parity-2026-10-15.json c7170cb
2026-10-15 Series S worker lifecycle: cancel, suspension with checkpoint, bit-for-bit resume, throughput None: the CPU oracle; docs/evidence/series-s-lifecycle-2026-10-15.json b0abffe
2026-10-15 Series S full-connectome throughput: event-driven delivery, bit for bit None: the CPU oracle; docs/evidence/series-s-full-connectome-2026-10-15.json dff96ca
2026-10-15 Series S exact float32 on the GPU (subnormals) and worker dashboard None: the CPU oracle; docs/evidence/series-s-full-connectome-2026-10-15.json (long_run) 2c07195
2026-10-15 Series S batch jobs: a whole battery bit for bit with sim.run None: the CPU oracle; docs/evidence/series-s-battery-2026-10-15.json f2f544c
2026-10-15 Looming to giant fiber timing v1 result: uninformative, the no-go (the model's giant fibers fire in every cell) validation/gf_looming.yaml v1, seeds 340–344; results/gf_looming/v1; the recordings were not opened 0579ad8
2026-10-15 Looming to giant fiber timing v2 result: uninformative (the flies' interval holds the onset slope) validation/gf_looming.yaml v2, seeds 350–354; results/gf_looming/v2; Dombrovski et al. 2025 control recordings (Zenodo 10.5281/zenodo.14983850, CC BY 4.0), opened after the model's no-go e5f6732
2026-10-15 Looming to giant fiber timing v2 locked: passive giant fibers validation/gf_looming.yaml v2; v1's result and a model-only probe (seed 9000), no data value 21ce984
2026-10-15 T4/T5 v4 result: robust neither Groschner et al. 2022, Behnia et al. 2014 (the lock's sources); validation/t4t5.yaml v4, seed 98; finding kept in docs/eye.md 111e3c5
2026-10-15 Simple baselines locked (SF2): the battery against 199 degree-preserving shuffles None: a structural baseline on the live defaults; validation/baselines.yaml v1 2f130fa
2026-10-14 Replication package: index, operator guide, claims audit (SF5) The archived manifests; docs/replication.md, docs/evidence/replication-index.json e4ca5f1
2026-10-14 Faster per-neuron membranes, bit-identical None: an engineering change, oracle-tested 111e3c5
2026-10-14 Graded potentials in the 3D replay; eye-on scene The T4/T5 lock's measured ranges (Groschner et al. 2022) set the glow scale e7355eb
2026-10-14 Sensorimotor environment plan Card & Dickinson 2008, von Reyn et al. 2014 and 2017, Ache et al. 2019 (as cited there); docs/sensorimotor-environment-plan.md 7da5e93
2026-10-14 Optional WebXR observer W3C WebXR Device API; three.js WebXR; docs/adr-virtual-experimental-environment.md 67106cc
2026-10-15 APL claw contrast locked before its run (#85) Prisco et al. 2021, doi:10.7554/eLife.74172, Figure 4B, Figure 4-figure supplement 1; validation/apl_claws.yaml v1 fdef643
2026-10-15 APL claw contrast v1 result: uninformative (the no-go: KCs below rest with APL, all spiking without) validation/apl_claws.yaml v1, seeds 420–424; results/apl-claws/v1.executions/b914035a…/output/ 07b04d7
2026-10-15 Pulse vs sine song v1 result: uninformative (the no-go; the data were never opened) validation/song.yaml v1, seeds 360–364, 99 shuffles; results/song/v1 a0e517c
2026-10-15 Simple baselines v1 result: all 7 positive responses specific against 199 shuffles validation/baselines.yaml v1, seeds 0–4, shuffles 4000–4198; results/baselines/v1.executions/f89ed1c3…/output/ 0528794
2026-10-16 Series S sparse counts: a whole battery bit for bit, 180 s end to end None: the CPU oracle; docs/evidence/series-s-battery-sparse-2026-10-16.json 21aec12
2026-10-16 Series S battery timing by phase; worker lists its batch outputs None: the CPU oracle; docs/evidence/series-s-battery-phases-2026-10-16.json ab38c57
2026-10-16 Series S GPU parity of version-2 fixtures (60, plus the 12 version-1), bit for bit None: the CPU oracle; docs/evidence/series-s-gpu-parity-v2-2026-10-16.json this PR
2026-10-16 Weak-pathway metrics and the identifiability re-read locked (SF2, #148) None: numerical checks on the live defaults; validation/path_metrics.yaml v1, seeds 5000–5004; validation/identifiability.yaml v1 re-reads results/sensitivity/v1 e6a041e
2026-10-16 Weak-pathway metrics v1: latency not converged (sugar MN9) None: the run from e6a041e; results/path-metrics/v1.executions/0c616b34…/output/. The identifiability classes are the re-read of results/sensitivity/v1, scored false e6a041e

The MaleCNS annotation and weight tables are used under CC-BY 4.0 as stated by Berg et al. 2026. A public GPL-3.0 repository publishes processing code for that dataset; flymsg cites the paper and does not copy that code.

FAFB connectivity used with --dataset fafb comes from the Shiu et al. 2024 repository (MIT) and the public FlyWire annotation table (citations above). flymsg does not redistribute those files.