Skywave Linux — inventory and delta¶
Generated by scripts/gen_skywave_inventory.py. Do not edit by hand —
regenerate. The curation table lives in the generator; the versions,
install methods, and apt availability below are all read from source data.
Release inventoried: Skywave Linux 5.10.0
Base system: Debian Linux (Sid / Unstable)
Kernel: 6.18.5-x64v2-xanmod1
Applications page: https://skywavelinux.com/
Install scripts: https://github.com/AB9IL/SDR-Scripts
apt availability: measured in a debian:13 container, 2026-08-25
Generated: 2026-09-07
Skywave Linux is written and curated by Philip Collier, AB9IL. Like AHRL
and 73Linux it is an inventory source, never a base (D-001). Unlike
73Linux, its scripts are licensed: SDR-Scripts, Skywave-Linux-scripts
and Linux-Respinner each ship a GPLv3 LICENSE, and both installer
scripts carry a GPL-3.0-or-later header. The provenance objection that
applies to a .bapp does not apply here. We still take no code — the value
is the curation.
Summary¶
| Count | |
|---|---|
| Featured applications listed | 60 |
| Delta — new coverage for us | 9 |
| Overlap — already covered by another source | 29 |
| Base system / general-purpose — out of scope | 22 |
SCOPE.md sizes Skywave at "~30 apps (small delta)"; profile-sizing.md
estimated ~10 to 15 unique units. Measured: 9. Both estimates
hold, and the shape of the delta is narrower than either implied — see
the corrections below.
The delta — 9 units¶
These are the units Skywave contributes that no other source in the
five-source union covers. Every one is absent from Debian — from stable
and from unstable. Measured in debian:13 and debian:sid containers,
not assumed. That matters: absence from stable is often just release lag,
and absence from both is a real gap in the distribution.
| Unit | Skywave version | Debian pkg | trixie | sid | Skywave's install method | What it is |
|---|---|---|---|---|---|---|
| Acarsdec | 4.4.1 | acarsdec |
absent | absent | git clone --depth 1 of the default branch, no tag |
VHF ACARS decoder; aeronautical |
| Acarsserv | 4.0 | acarsserv |
absent | absent | git clone --depth 1 of the default branch, no tag |
SQLite store for acarsdec output |
| DumpHFDL | 1.7.0 | dumphfdl |
absent | absent | shipped in the ISO but absent from the published scripts | HF datalink; oceanic aeronautical |
| Kalibrate-RTL | 0.4.1-rtl | kalibrate-rtl |
absent | absent | git clone --depth 1 of the default branch, no tag |
dongle PPM calibration off GSM bursts |
| LibACARS | 2.2.1 | libacars2 |
absent | absent | git clone --depth 1 of the default branch, no tag |
shared library for the ACARS family |
| Reticulum MeshChat | 2.3.0 | — (none) | — (none) | — (none) | shipped in the ISO but absent from the published scripts | Reticulum mesh messenger |
| RTLSDR-Airband | 4.0.2 | rtlsdr-airband |
absent | absent | git clone --depth 1 of the default branch, no tag |
multi-channel AM/NFM voice, Icecast out |
| SuperSDR | 3.14 | supersdr |
absent | absent | shipped in the ISO but absent from the published scripts | KiwiSDR client with CAT sync |
| VDLM2dec | 2.3 | vdlm2dec |
absent | absent | git clone --depth 1 of the default branch, no tag |
VDL Mode 2 decoder |
Where each one actually comes from¶
| Unit | Upstream we would use | Licence | Newest tag | Last commit | State |
|---|---|---|---|---|---|
| Acarsdec | f00b4r0/acarsdec |
GPL-2.0-only | v4.6 | 2026-06-28 | successor to the archived original |
| Acarsserv | TLeconte/acarsserv |
GPL-2.0 | — | 2018-12-19 | archived; no successor found |
| DumpHFDL | szpajder/dumphfdl |
GPL-3.0 | v1.7.0 | 2025-11-02 | maintained |
| VDLM2dec | szpajder/dumpvdl2 |
GPL-3.0 | v2.7.0 | 2026-08-01 | supersedes the archived vdlm2dec |
| LibACARS | szpajder/libacars |
MIT | v2.2.1 | 2025-11-02 | dependency of the four above |
| RTLSDR-Airband | rtl-airband/RTLSDR-Airband |
GPL-2.0 | v5.3.1 | 2026-08-23 | maintained |
| Kalibrate-RTL | steve-m/kalibrate-rtl |
BSD-2-Clause | none | 2022-02-01 | dormant, and it has never cut a tag |
| SuperSDR | mcogoni/supersdr |
none — default copyright | v3.14 | 2022-12-31 | dormant |
| Reticulum MeshChat | liamcottle/reticulum-meshchat |
MIT | v2.4.0 | 2026-08-15 | maintained |
How the last-commit column was measured, and how it was measured wrongly the first time. The dates here are the newest commit on each project's default branch, read from the GitHub API on 2026-08-28.
The first version of this table used the API's updated_at field, which is not
a measure of development at all — it moves when somebody stars the
repository, forks it, or edits its description. Two projects were reported as
active on that basis and are not:
| Unit | Was published as | Last commit on the default branch |
|---|---|---|
| Kalibrate-RTL | active (2026-08-19) | 2022-02-01 |
| SuperSDR | active (2026-02-18) | 2022-12-31 |
pushed_at is the near-miss and is also wrong: it moves on a push to any
branch, including a fork's, so it overstated five of these nine. Only the
default branch's head commit answers the question this column is asked for.
Both corrections change decisions that were resting on them, and both are recorded below rather than quietly amended (D-018, D-025).
Three findings about provenance, each checked in the repository tree rather than read off a metadata field — which, as the correction above records, is the distinction that matters here.
1. SuperSDR has no licence. mcogoni/supersdr carries no LICENSE, no
COPYING, and no per-file header — checked in the repository tree and in
supersdr.py itself. Default copyright applies. This is the 73Linux situation
again (D-001) and it lands on a headline application: SuperSDR is Skywave's
KiwiSDR client and the most visible piece of the listening delta. We can install
from upstream; we cannot vendor, patch, or redistribute it. This needs a
maintainer decision — see Q-007.
The same check on the other two KiwiSDR clients in AB9IL's repository set found
jks-prv/kiwiclient with no licence statement anywhere, and llinkz/directKiwi
with a WTFPL-style grant stated in README prose but no licence file. There is no
cleanly-licensed KiwiSDR client in this set.
2. Thierry Leconte's decoder suite has been archived, and only part of it has a
successor. All three of acarsdec, vdlm2dec and acarsserv are now
read-only on GitHub:
| Original | State | Where it goes |
|---|---|---|
TLeconte/acarsdec |
archived; last activity 2025-07-31 | SUPERSEDE → f00b4r0/acarsdec, GPL-2.0-only, v4.6 |
TLeconte/vdlm2dec |
archived; last activity 2024-02-11 | SUPERSEDE → szpajder/dumpvdl2, GPL-3.0, v2.7.0 |
TLeconte/acarsserv |
archived; last activity 2018-12-19 | CARRY — no successor found; still the companion store acarsdec documents |
The archived flag is verified from the API. The dates on which each was
archived are not: GitHub does not expose archived_at for any of these three,
so an earlier version of this table stating exact archive dates was publishing
something it could not source. What is shown instead is last repository
activity, which is checkable. Note how far back acarsserv's goes — carrying it
means carrying software untouched since 2018.
Skywave 5.10 already ships Acarsdec 4.4.1, which can only have come from the
f00b4r0 continuation, so the ISO tracks the live tree. Its published
decoders.sh still points at a 2023 fork of the dead one — another sign that the
scripts and the image have diverged.
Note that Skywave ships both VDLM2dec 2.3 and DumpHFDL, and its script
carries an update_dumpvdl2 function that is one of the broken ones. Carrying
dumpvdl2 and retiring vdlm2dec is the cleaner outcome and satisfies all four
SUPERSEDE bars in PARITY-POLICY.md: same core function, actively maintained,
installs the same way, and the trade-off states in one sentence — dumpvdl2 is
the maintained VDL Mode 2 decoder; vdlm2dec is archived.
3. GitHub's licence API is not evidence of absence, and its repository names
are not stable. The API reported NONE for TLeconte/acarsdec, which states
LGPL-2 in its README body. It also answered happily for both
szpajder/RTLSDR-Airband and charlie-foxtrot/RTLSDR-Airband — both are
redirects, and the canonical repository is now rtl-airband/RTLSDR-Airband,
which has reached v5.3.0 while Skywave ships 4.0.2. Every upstream URL in the
table above was resolved to its final location before being written down.
Corrections to SCOPE.md¶
SCOPE.md describes the Skywave delta as "remote SDR clients (KiwiSDR, WebSDR,
Web-888, PhantomSDR, OpenWebRX), utility decoders (ACARS, HFDL, VDL2, AIS), and
Reticulum/MeshChat." Measured against the 5.10.0 release, three parts of that
need correcting.
1. Most of the "remote SDR clients" are not client software. KiwiSDR, WebSDR,
Web-888, PhantomSDR and OpenWebRX are receivers and server stacks you connect
to, not applications Skywave installs. The 5.10.0 list ships exactly one
dedicated remote-SDR client: SuperSDR. Everything else is reached through a
browser, or through AB9IL's site-list tooling (dyatlov, sdr-selector,
kiwisdr-helpers) that generates maps and menus of public receivers. The
January 2026 changelog entry — "Expanded the scope of the KiwiSDR list to
include OpenWebRX sites" — confirms the reading: OpenWebRX is an entry in a
directory of receivers, not a package.
That is a smaller software delta than SCOPE.md implies, and a larger data
one. The site directory is the actual asset, and it is the thing a user with no
antenna needs first.
2. AIS is not in the 5.10.0 release. rtl-ais appears in the v4-era
decoders.sh and in Skywave-Linux-scripts (ais-mapper, ais_monitor.sh,
ais-fileto-sqlite), but not in the 5.10.0 featured list. It is also already
available to us: rtl-ais is in Debian 13, and AIS-catcher already has a
manifest in this repository. AIS should be attributed to our own ADD list, not
to the Skywave delta.
3. OpenWebRX and PhantomSDR in SCOPE.md came from a stale page.
skywavelinux.com/sourcecode.html lists both — along with simonyiszk/openwebrx
(the original, long superseded), WSJT-X 1.6.0, and an Ubuntu focal base. That
page documents the v4 era and has not been updated. The versioned featured list
on the front page is the current authority and was used here instead.
What Skywave deliberately excludes¶
SCOPE.md says these are "decisions worth respecting rather than reversing by
default." Both are confirmed from the project's own release notes, and both are
better-founded than the summary suggested.
GNU Radio and gqrx, dropped for bloat. From the v5 release announcement:
"Prior releases had too many SDR interfaces. There was Gqrx, wich was easy to use and had nice features. But along with that came a lot of Gnuradio packages."
with the rationale that "you can do almost anything with CubicSDR an SDR++ and
don't need mush else for plugged in SDR devices." The 5.10.0 list bears this
out: no GNU Radio, no gr-osmosdr, no gqrx — all three of which the published
v4-era sdr-installer.sh still builds.
The consequence for us is narrower than "respect the decision." Skywave
dropped GNU Radio because it did not need it. DragonOS Tier 3 is GNU Radio
out-of-tree modules, and our sdr profile carries gqrx from the Blend. We are
not in Skywave's position and should not inherit its conclusion — but we should
inherit the observation behind it, which is that a GNU Radio dependency chain is
heavy enough that a curator with a size budget cut it. That is an argument for
keeping Tier 3 in its own opt-in profile, which SCOPE.md already requires.
The SDRplay API, excluded as non-free. From the same announcement:
"One notable absence in the updasted distro is the SDRplay API."
with three stated reasons: the proliferation of RSP clones, the API and driver being closed source, and SDRplay's own terms forbidding use with unapproved hardware. The author frames it as a consistency point — relying on "non-free and closed source drivers" would contradict the project's commitment to open software.
This is stronger than an exclusion, and it is a real gap. The v4-era
sdr-installer.sh still contains a get_sdrplay_api function, and it is
instructive: it cannot download the artifact, because SDRplay gates it behind a
web form. The script's own comment reads "Robots are people too." The user must
place SDRplay_RSP_API-Linux-3.07.1.run in /usr/local/src by hand before the
function will do anything. SoapySDRPlay3 is built in that script and is gone
from the 5.10.0 list, so the exclusion was carried through.
For us this is a system_modifications and documentation problem rather than a
backend one. An artifact behind an interactive download gate cannot be installed
non-interactively, cannot be checksummed in advance, and therefore cannot
satisfy the security requirement that we verify checksums for any non-apt
source. If we support SDRplay hardware at all it must be as a documented
manual step with the licence terms shown, never as a silent download. Recorded
in overlaps.md terms: this is a hardware gap, not a software choice.
Install methods, and what they cost us¶
Read off sdr-installer.sh and decoders.sh. These are v4-era (2022 copyright)
and no longer match the shipped ISO — dumphfdl is in 5.10.0 and in neither
script — so treat them as evidence of approach, not as a current manifest.
Every source build is unpinned, and the pinning is then actively destroyed.
All 34 git clone invocations across the two scripts use
--depth 1 on the default branch; not one names a tag. Both scripts then run a
maintenance pass over every checkout on disk:
git tag -d $(git tag -l) deletes every tag in the local repository. The
intent is plainly disk-space hygiene on a live ISO, and the effect is that the
checkout can no longer name the version it is running. This is the same
unpinned-snapshot problem PARITY-POLICY.md records for AHRL's SatDump and
SDR++, in a sharper form: AHRL merely fails to pin, this removes the ability to.
Release-tracking downloads resolve at install time. download_last shells
out to lastversion to fetch the newest release asset, so every .deb, .zip
and AppImage is whatever upstream published most recently. Reproducible installs
are impossible by construction, and no checksum is computed anywhere in either
script — consistent with SCOPE.md's finding that none of the five sources
publishes hashes we can inherit.
Backend implications, measured. The middle column counts what Skywave's scripts do; the right column is what it costs us, which is not the same thing because we resolve several of these to apt instead.
| Method | Uses in Skywave's scripts | Status for us |
|---|---|---|
| source from git | 34 | already required by AHRL (D-004) |
upstream .deb |
2 — SDR++, noaa-apt | already required |
| upstream archive | 2 — libairspy, SDRTrunk | already required |
| AppImage | 1 — CubicSDR | CubicSDR is in Debian 13; we use apt |
| manual, gated download | 1 — SDRplay API | not a backend — see above |
| PyPI | 0 in the scripts | needed for Reticulum, which postdates them |
Skywave adds no new backend requirement. Its one AppImage is CubicSDR, which Debian 13 packages, so we take the apt route and the AppImage never arises.
The one genuine movement is elsewhere: Reticulum MeshChat ships Linux only as
an AppImage. AppImage was previously justified by HAMRS alone — post-1.0, and
its upstream discovers downloads by scraping a webpage. MeshChat is a cleaner
second consumer with ordinary GitHub release assets. That does not promote
AppImage into 1.0, since MeshChat sits in the post-1.0 mesh profile, but the
backend now has two independent users rather than one awkward one, which is the
kind of measurement D-014 asks for.
Defects in the published scripts¶
Found by reading, not by running. Recorded because they are the same class of defect the declarative-catalog argument exists to prevent, and because two of them mean the affected units cannot have been installed by these scripts.
| Script | Line | Defect | Effect |
|---|---|---|---|
decoders.sh |
update_dump1090 |
cd "$working_dir/dumpdump1090-fa" — the directory cloned is dump1090-fa |
cd fails; the build runs in whatever directory the previous job left |
decoders.sh |
update_dumpvdl2 |
cd "usr/local/src/dumpvdl2/build" — relative path, missing leading / |
same |
decoders.sh |
update_kalibrate-rtl |
CXXFLAGS='-W Wall -03' — -03 is a zero, not -O3; -W Wall is two malformed flags |
assignment is a no-op before ./configure, so it is silently ignored |
sdr-installer.sh |
get_gqrx |
cd "$working_dir/gqrx/build" then cd build again |
second cd fails; cmake .. runs one level too high |
sdr-installer.sh |
GNU Radio deps | pins libgnuradio-*3.8.1 by SONAME |
unresolvable on the Debian Sid base the ISO now uses |
None of this is a criticism of the ISO, which is built and tested as an image;
it is a criticism of shell as a packaging format, which is the architecture
this project exists to replace (D-001). A manifest cannot have a typo in a
cd, because it has no cd. The build directory is derived from the checkout,
so dumpdump1090-fa is unrepresentable rather than merely wrong.
Every one of these five is invisible to the person running the script: the jobs
are dispatched through GNU sem in parallel with no error propagation, and the
script prints "End of script. Good luck / have fun." whether or not anything
built. That is precisely the failure mode D-016 (fail loudly on unresolvable
dependencies) and the transaction log are designed to make impossible.
Overlap — 29 units already covered¶
Listed so the overlap is on the record rather than rediscovered. The
Blend column is derived from docs/reference/blend-inventory.md.
| Unit | Skywave version | Debian pkg | trixie | sid | In Blend | Note |
|---|---|---|---|---|---|---|
| CubicSDR | 0.2.7 | cubicsdr |
0.2.7+dfsg-4 | 0.2.7+dfsg-5 | yes | Blend sdr |
| Fldigi | 4.2.10 | fldigi |
4.2.06-1+deb13u1 | 4.2.12-1 | yes | Blend datamodes; manifest already written |
| Gpredict | 2.4 | gpredict |
2.3-115-g0f3beb6-4 | 2.5.2-1 | yes | Blend satellite |
| Hamlib | 4.6.5 | libhamlib-utils |
4.6.2-1+b1 | 4.7.2-1+b1 | yes | Blend rigcontrol; the rig-control floor |
| JS8Call | 2.5.1 | js8call |
2.2.0+ds-5+b1 | 3.0.3+ds-1 | yes | Blend datamodes; manifest already written |
| Libairspy | 1.0.10 | libairspy0 |
1.0.10-3+b3 | 1.0.10-7 | no | hardware library |
| Libairspyhf-dev | 1.6.8 | libairspyhf1 |
1.6.8-4+b2 | 1.6.8-5+b1 | no | hardware library |
| Libhackrf | 2024.02.1 | libhackrf0 |
2024.02.1-3 | 2026.01.3-1 | no | hardware library |
| Limesuite | 23.11.0 | limesuite |
23.11.0+dfsg-3+b2 | 23.11.0+dfsg-5 | yes | hardware library |
| Multimon-ng | 1.3.1 | multimon-ng |
1.3.1+dfsg-1 | 1.3.1+dfsg-1.1 | yes | Blend packetmodes |
| SatDump | 1.2.3 | satdump |
1.2.2-1 | 1.2.2+git20260527.6e4757c31+dfsg-3+b1 | yes | AHRL + Blend; our SUPERSEDE for the APT decoders |
| SDR++ | 1.2.1-1487 | sdrpp |
absent | 0.0~git20260520.36ea9a1+dfsg-2 | yes | AHRL ships an unpinned snapshot |
| SoapyAirspy | 0.2.0 | soapysdr-module-airspy |
0.2.0-4+b2 | 0.2.0-5+b1 | yes | per-device backend |
| SoapyAudio | 0.1.1 | soapysdr-module-audio |
0.1.1-6+b1 | 0.1.1-7+b1 | yes | per-device backend |
| SoapyBladeRF | 0.4.2 | soapysdr-module-bladerf |
0.4.2-1 | 0.4.2-1.1+b1 | yes | per-device backend |
| SoapyFCDPP | 0.2.0 | — (none) | — (none) | — (none) | no | per-device backend; not in Debian |
| SoapyHackRF | 0.3.4 | soapysdr-module-hackrf |
0.3.4-2 | 0.3.4-3+b1 | yes | per-device backend |
| SoapyLMS7 | 23.11.0 | soapysdr-module-lms7 |
23.11.0+dfsg-3+b2 | 23.11.0+dfsg-5 | yes | per-device backend |
| SoapyMiri | 0.2.5 | soapysdr-module-mirisdr |
0.2.5-8+b5 | 0.2.5-11 | yes | per-device backend |
| SoapyOsmo | 0.2.5 | soapysdr-module-osmosdr |
0.2.5-8+b5 | 0.2.5-11 | yes | per-device backend |
| SoapyPlutoSDR | 0.2.2 | soapysdr-module-plutosdr |
absent | 0.2.2-1.1 | no | per-device backend |
| SoapyRedPitaya | 0.1.1 | soapysdr-module-redpitaya |
0.1.1-5+b2 | 0.1.1-6+b1 | yes | per-device backend |
| SoapyRemote | 0.5.2 | soapysdr-module-remote |
0.5.2-4+b2 | 0.5.2-4.1+b1 | yes | per-device backend; network SDR |
| SoapyRTLSDR | 0.3.3 | soapysdr-module-rtlsdr |
0.3.3-1+b4 | 0.3.3-2+b1 | yes | per-device backend |
| SoapyRTLTCP | 0.1.0 | — (none) | — (none) | — (none) | no | per-device backend; not in Debian |
| SoapyUHD | 0.4.1 | soapysdr-module-uhd |
0.4.1-5 | 0.4.1-7 | yes | per-device backend |
| SoapySDR | 0.8.1 | soapysdr-tools |
0.8.1-5+b2 | 0.8.1-7+b2 | yes | Blend sdr |
| Uhd-Host | 4.9.0.0 | uhd-host |
4.8.0.0+ds1-2 | 4.9.0.1-1.4 | yes | hardware library |
| WSJT-X | 3.0.0 | wsjtx |
2.7.0+repack-1 | 3.0.2+dfsg-2 | yes | Blend datamodes; manifest already written |
23 of 29 overlap units are in the Debian Blend, so they arrive with the cheapest coverage in the project and cost us nothing extra.
15 of the overlap units are the SoapySDR family, which is the same
observation profile-sizing.md makes about the Blend's sdr task: thirteen
soapysdr-module-* packages, of which a user with one dongle needs one. Skywave
ships the full set because a live ISO cannot know what will be plugged in. We
can — this is the clearest argument yet that hardware detection should drive
profile resolution, and it is now supported by two independent sources rather
than one.
Two Soapy modules are not in Debian: SoapyFCDPP (FUNcube Dongle Pro+) and
SoapyRTLTCP. Both are small CMake builds. Neither is needed for 1.0, and both
belong to the M4 hardware work rather than the catalog.
Base system and general-purpose software — 22 units, excluded¶
Skywave is a live ISO, so its featured list includes the desktop it boots into. We augment an existing install and do not ship a desktop, an editor, a browser, or a video editor. Recorded so the exclusion is a decision rather than an oversight.
| Unit | Version | Debian pkg | trixie |
|---|---|---|---|
| Audacity | 3.7.7 | audacity |
3.7.3+dfsg-1 |
| BleachBit | 5.0.2 | bleachbit |
4.6.2-2 |
| Brave Browser (beta) | 1.86.116 | — (none) | — (none) |
| Dynamic Window Manager (DWM) | 6.6 | — (none) | — (none) |
| Go | 1.24.0 | — (none) | — (none) |
| I2P (Purple) | 2.58.0 | i2pd |
2.56.0-1 |
| JupyterLab | 4.5.2 | jupyterlab |
4.0.11+ds1+~cs11.25.27-7 |
| Neovim | 0.12.0 | neovim |
0.10.4-8 |
| Node.js | 22.21.1 | — (none) | — (none) |
| OBS-Studio | 30.2.3 | obs-studio |
30.2.3+dfsg-3 |
| Obsidian | 1.11.4 | — (none) | — (none) |
| Openjdk-java | 21.0.10 | — (none) | — (none) |
| Pandas | 2.3.3 | — (none) | — (none) |
| Pandas-Datareader | 0.10.0 | — (none) | — (none) |
| Pandoc | 3.1.11 | pandoc |
3.1.11.1+ds-2 |
| Pipewire | 1.4.10 | pipewire |
1.4.2-1 |
| Python | 3.13.9 | — (none) | — (none) |
| Scipy | 1.16.3 | — (none) | — (none) |
| SMPlayer | 25.6.0 | smplayer |
24.5.0+ds-1 |
| Shotcut | 25.12.31 | shotcut |
25.03.29+ds-1 |
| Veracrypt | 1.26.14 | veracrypt |
absent |
| Wezterm | 20251201 | — (none) | — (none) |
One is arguable: I2P (Purple) — i2pd is in Debian 13, and an anonymity
network is closer to the RF-security profile than to a desktop. It is left
out of the delta because nothing in our scope requires it, and adding an
anonymity network to a radio tool is the maintainer's call, not the
generator's.
A finding about the Debian Blend, turned up here¶
The same container probe was pointed at the Blend's own package list as a control, and it found something worth recording before the profile work depends on it.
8 of the Blend's 152 packages do not install on Debian 13.
| Blend package | trixie | sid | Why |
|---|---|---|---|
aethersdr |
absent | 26.8.3+ds-1 | release lag |
dump1090-mutability |
absent | 1.15~20180310.4a16df3+dfsg-8.2 | release lag |
fbb |
absent | 7.011-3 | release lag |
not1mm |
absent | 26.8.10-1 | release lag |
odr-audioenc |
absent | absent | not in Debian at all |
qlog |
absent | 0.52.0-1 | release lag |
sdrangel |
absent | 7.27.2+dfsg-1 | release lag |
sdrpp |
absent | 0.0~git20260520.36ea9a1+dfsg-2 | release lag |
The Blend tracks unstable, so 7 of these are ordinary release lag and will arrive in Debian 14. That does not help a user installing today on the stable base most of our targets derive from.
Two of them matter to decisions already made:
qlogisoverlaps.md's recommended default for logging, chosen overcqrlogandxlog. It is in sid at 0.52.0-1 and not in trixie. The recommendation is still right, but on our primary targets it needs a non-apt install path or an honest "not available here" — which is exactly what the capability matrix is for.sdrppandsdrangelare listed in the Blend'ssdrtask and are likewise sid-only.PARITY-POLICY.mdalready flags SDR++ as an AHRL unpinned-snapshot fix; this says the apt route is not available on stable either, so the source or.debpath is required rather than optional.
The general lesson. "In the Blend" is not the same claim as "installable",
and blend-inventory.md records the former. SCOPE.md calls the Blend "the cheapest
coverage in the project", which remains true at 94%
on stable — but the residual is not zero, and it lands on packages we had
already chosen as defaults. Per D-005, coverage counts only where it
installs.
Recorded here rather than silently fixed: adjusting overlaps.md and
profile-sizing.md is doc-reconciliation work, not inventory work.
What this changes¶
Profile sizing. The delta lands almost entirely in listening, which
profile-sizing.md sized at ~35 with room to spare. Adding the seven
aeronautical/maritime decoders takes it to roughly 42 — still well under the
80 threshold, and coherent: a listening profile that decodes ACARS, HFDL and
VDL2 is a recognisable thing an operator can name. SuperSDR belongs there too.
Reticulum MeshChat belongs in mesh, which is post-1.0.
The ~10-15 estimate in profile-sizing.md was close; the number is now
measured and that row can be marked as such.
The utility-decoder cluster is a genuine hole in Debian. apt-cache search
found no acars*, no vdl*, no hfdl, no airband and no redsea — searched,
not merely looked up by name — and the named packages are absent from unstable
as well as stable. This is not release lag. The entire aeronautical and
maritime decoding domain is absent from the distribution, and therefore from the
Blend, which is the largest single coverage gap the Blend leaves. Seven source
builds is a real cost, but they are small C/CMake projects with active upstreams
and they buy a domain no other source in the union covers.
One thing to carry forward to the DragonOS inventory. dumphfdl,
dumpvdl2, libacars and acarsdec are named in SCOPE.md as DragonOS Tier 1
— "apt-installable or upstream .deb". Measured above, they are apt-installable
on neither trixie nor sid, so on the stated definition they are not Tier 1 at
all. Tier 1's membership has to be decided by a probe rather than by the list in
SCOPE.md, and that is the next item's job. Whether Tier 1 is then redefined to
admit small source builds, or these four move to Tier 2, is a scope question for
the maintainer.
Open question raised. Q-007 — what to do about SuperSDR, whose upstream
carries no licence at all. It is the only dedicated remote-SDR client in the
release and the other two KiwiSDR clients in the same ecosystem are no better.
What we take, in one line. The utility-decoder cluster, SuperSDR with a licence caveat, the Reticulum client, and — more valuable than any of them — the observation that the remote-receiver directory is the product for a user who owns no hardware.