Four things asked for in turn, landing together because they run through the
same files.
THE MAPS. The tiles were already cached and always had been -- a second
evening on the same view was measured at nought network requests -- but the
work done on them was not. Every window open decoded forty PNGs and resampled
a megapixel and a half into this program's own projection, for an answer that
cannot have changed, because a coastline does not move. The finished map is
kept beside the tiles now: 0.61 seconds become 0.01, byte for byte identical,
two hundred kilobytes a view. Both windows and both kinds of still picture get
it, all four reaching the ground through one function. A map with squares
missing is deliberately not kept, since caching a hole would keep it for a
month and the point of calling a partial map provisional is that it is asked
for again. And because this adds a disk consumer, the whole cache is now
pruned to four hundred megabytes, least recently *used* first: a tile fetched a
year ago and looked at last night is the receiver's own neighbourhood, and
discarding that to keep last week's holiday is the wrong way round.
THE LOOKUPS. APRS and FT8 now ask who each station is licensed to. Every
other mode that hears a callsign already did -- speech transcripts, Morse
idents, the recording browser -- and all five resolve through one file, so a
callsign heard on two bands is asked about once. The rules for finding the
licensed callsign inside a heard one are now in one place rather than per band,
because getting them wrong is silent: a register asked about W1AW-9 returns
nothing, which looks exactly like a station that is not licensed. An SSID, a
rover suffix, a guest prefix, a digipeater alias and an unspelled hash all come
off or are refused.
The bug worth recording is that the first cut of this did the lookups and threw
them away. The book has to be told to save and was not, so every evening would
have asked the register about the same net again -- which is the one thing
caching them was for, and is invisible from inside a single run because the
answers are all in memory while it lasts. Caught by looking at the file on
disk rather than at the display. There is a test for each side of it now, and a
register of which modules resolve callsigns at all, which fails when a new one
starts so that somebody has to decide whether it should.
THE REGISTER. c in saunterbrowse opens everything ever looked up: sixty-odd
callsigns and sixteen hundred aircraft here, every field of each in two columns
because a licence has a dozen and a screen is wider than it is tall. tab
switches, / searches every field rather than the name -- the question is
usually "who was in Arizona" rather than "which callsign" -- and g opens the
place in a browser. Only the coordinates go into that link: a map does not
need to be told whose licence it is looking at, and the link is the one part of
this that leaves the machine. A headless box, which is the normal case for a
receiver, gets the coordinates printed instead. Callsigns no register could
place are kept rather than dropped, because "asked about, and in no register
reachable from here" is a fact about a station.
THE FRONT OF IT. A title screen for each program: five rows of blocks cut by
hand, a figlet dependency to draw eleven letters being the largest thing that
would then be in the requirements, coloured blue to red across the width, which
is the ramp every waterfall here already uses because it is what a spectrum
looks like. The interesting part is where it does not appear -- everything
here can be piped into something else and a banner in the middle of that is
corruption rather than decoration, so anything that is not a terminal gets
nothing, --help is untouched because it is drawn after parsing, --no-splash
turns it off for a run and BANDSAUNTER_NO_SPLASH=1 for good.
And the address. INSTALL.md said "git clone <the repository>" for a long time:
a placeholder in the first command anybody types, unnoticed because nothing
reads install instructions except somebody installing, who then cannot. It is
filled in, along with the readme, the metadata, both manuals and the Homepage
field of all three packages. The tile server's User-Agent pointed at a topic
listing on somebody else's site for want of an address of its own; the usage
policy of that service asks for one naming the application and giving somewhere
to look it up, so an operator with a question about the traffic has somebody to
ask, and now it gives the real one. Seven tests so the placeholder cannot come
back, verified by putting it back and watching two of them fail.
One thing forced by all this: the keys page in saunterbrowse was exactly as
tall as an eighty-by-twenty-four terminal, so the register entry pushed "q
quit" off the bottom. A test caught it. Home and End have merged into the
Page Up line, which were always the same thought.
THE WINDOWS. They open maximised now, this being a map and the thing anybody
wants more of being map; f goes to true full screen and back, and is written
along the top of the screen because a window with no frame is one somebody has
to know a key to get out of. Maximised rather than frameless by default,
because the title bar is where the band and the frequency are written.
And a map made bigger now gets a sharper map, which it did not. The window
only ever re-examined the ground when the view left the box that had been
fetched, so a window opened at its default size and taken to the whole screen
kept the map it started with until an aircraft wandered far enough to move it
-- on a quiet band, a long time to look at a blurred coastline. Measured
before it was believed: 1100 to 1920 asked for nothing and stretched a
1364-pixel map across 1920, then across 3840. It now compares map pixels per
degree in hand against what the view wants, and asks when it is being blown up
by more than fifteen per cent.
That comparison has to be per degree rather than pixel against pixel, which a
surviving mutation was what established: the fetched box is a quarter wider
than the view, so a map with exactly as many pixels as the window is wide has
only four fifths of them on the screen. The two ways of measuring agree
everywhere except a narrow band, and the realistic case sits inside it. There
is a test pinning that case now. Asking is safe at any size, because the
request is keyed on the window's dimensions: once answered, nothing more is
asked, which is what stops a window larger than the tile budget can cover from
asking all evening.
The braille, while this was open. The banners were blocks in capitals; they
are braille in mixed case, two dots wide and four tall to a character, which is
eight times the detail and is what makes room for two heights of letter at
once -- a five-row block font has no second height to spend, so the name came
out shouted. Two attempts failed first: thin one-dot strokes came out as
confetti, because braille dots render as dots and a one-dot stroke reads as a
dotted line rather than a line. What worked was cutting the font small, at cap
height seven where there is only one way to draw each letter, and doubling it.
Coloured deep blue through cyan to a cool white, which is deliberately not the
waterfall ramp the rest of the program draws in: that one has to run to red
because it stands in for a spectrum, and a title screen stands in for nothing.
One hundred and four new tests. Full suite 2892 passed. Built as
2026-09-24_01.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016PsWPTweCT6pwxKngvVxcg
A new section, alongside the aircraft, the weather sensors and APRS.
FT8 is the odd one out among the things this program listens to, and the
reason is worth stating because it shapes everything below. Every station on
the band transmits in the same quarter-minute slots, on the same dial
frequency, fifty hertz wide each, stacked across three kilohertz of audio.
One receiver parked on one frequency therefore hears the whole band's worth of
stations at once -- and hears most of them well below the noise, because half
of what is sent is error-correcting code. That is the entire trick: a rate of
about one half buys a mode that decodes twenty-odd decibels under what an
operator can hear. A receiver that took the loudest tone of each symbol and
hoped would decode almost nothing, which is why the tone detector reports how
confident it is bit by bit rather than what it thinks it heard.
Written from first principles except for two tables. The checksum, the
belief propagation over the sparse graph, the Costas sync search, the
waterfall, the soft-bit metric, and the seventy-seven bits that hold two
callsigns and a grid square are all here. The generator and the parity-check
matrix are not: they cannot be derived, being the code itself rather than
consequences of anything, so they are taken from ft8_lib under its MIT licence
with the attribution it asks for, and said so in the readme, the manual and
the file. No decoding logic came with them. That the two agree -- and they
are not derivable from one another, the generator's parity half running to
fifty-odd bits a row against the sparse matrix's six or seven -- is a test
rather than an assumption.
Tested against the air, not against itself. Eleven off-air recordings with
published decodes: ninety-seven of a hundred and fifty messages, no false
decodes, timing within a hundredth of a second, frequency within a hertz,
signal reports within half a decibel on average. The third not decoded are
the weakest in each slot; a mature decoder subtracts what it has decoded and
looks again in the remainder, and does ordered-statistics decoding where
belief propagation fails, and neither is built here. What is here decodes
nothing that other receivers did not also hear, which is the property that
matters in a log. Ten whole codewords lifted off the air are in the tests as
a permanent fixture, so the recordings can go missing and the regression
cannot.
Three things that looked like bugs and were not, and three that were. The
half-second timing discrepancy was the convention: a transmission is 12.64
seconds in a slot of fifteen and everybody starts half a second in, so
lateness is reported against that. Synthetic signals at known offsets proved
the clock self-consistent before anything was changed. The signal reports
were twenty-one decibels optimistic because those recordings have a receiver
passband above three kilohertz, putting a whole-band median twelve to sixteen
decibels below the real noise floor -- so noise is now measured beside the
signal, and in the tone that was actually sent rather than the loudest of
eight, the largest of eight noisy numbers being well above their mean even
with no signal at all. And the test transmitter was thirteen decibels
pessimistic, scaling its noise into a fifty-hertz reference instead of the
sampled bandwidth, which made the decoder look deaf when it was the test
signal that had been quietly attenuated.
The real bug the simulator caught was a one-block timestamp error: samples
were dated a block earlier than they were taken, which slid every slot slice a
second late and cut the first half-second -- three symbols, part of the
opening Costas array -- off every transmission on the band. One decode a slot
became six.
Reachable both ways, as everything here is. Twenty-one options, every one of
them a command-line flag and a line in the menu, both built from one table so
they cannot disagree -- and a test that says so, since an option in no group
would be settable from the command line and invisible in the menu. The band
list says which channels a plain dongle can reach and which need an
upconverter, because almost all the activity is on shortwave and finding that
out by listening to silence for ten minutes is the wrong way to learn it. The
default is two metres, which a plain dongle can hear.
--grid turns decodes into geography: every CQ says where it is, so each gets a
distance and a bearing and the furthest is named. --adif writes the log in
the form every amateur logging program imports, marked as heard rather than
worked, because nothing here transmits and an ADIF that let a logging program
treat these as contacts would put claims into somebody's log that they cannot
make.
One bug shipped and found by being used rather than by being tested: the
line that opens the receiver called a function this program has never had.
Every test reached it through the simulator, which takes the other branch, so
the one line that matters to somebody with an aerial was the one line never
run. There is now a check that every name these modules import actually
exists -- it names the missing one rather than failing somewhere downstream --
and two that say a receiver which cannot be opened is reported rather than
raised, and that nothing claims to be listening before there is one. It had
been announcing the frequency first, so a dongle that would not open read as
listening that had gone wrong.
Ninety-six new tests. Full suite 2781 passed. Built as 2026-09-21_04.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016PsWPTweCT6pwxKngvVxcg
The Dust Council