In 1993, when I was still blissfully unaware of the internet, Amsterdam based hacker club Hack-Tic founded the (locally) legendary internet service provider XS4ALL. They were late to the game. Five others had beaten them to it, but since their founding, XS4ALL were always fighting for things such as internet privacy, net neutrality and free (internet) speech, in partnership with NGO Bits of Freedom. They also sponsored projects like Python, Debian and SquirrelMail. XS4ALL had ipv6 before anyone else and static IPs were standard. I wanted to be their customer, but they were also a bit more expensive than others.
What does this have to do with my email workflow?
Well, in 1998, years before I started using Gmail, XS4ALL was sold to KPN, and in 2019, when I wasn’t using Gmail any more, KPN decided to finally integrate XS4ALL. By that time I was in fact a customer and KPN was a company that tried very hard how to suck at internet the American way, so I wasn’t the only one unhappy with this situation. A group of customers actually founded a new ISP, Freedom internet, that has become the moral successor to XS4ALL. Obviously, I was one of their original customers and still am. I also invested financially in them.
When you run Internet facing server software at home, like this blog, and you do a small but slightly too careless migration and your ISP tells you your git(1) history is world readable, you figure they are looking out for you. They confirm your trust in them. This is why they do my email. They have web clients, but I use Thunderbird, which is the Firefox of email clients. Perhaps I should give Emacs a try. As for my inbox, I do the same thing that Wouter does: I archive what makes me smile and delete anything else when I’m done with it.
I used Gmail for over a decade. The emails stored there are so old now, they don’t contain anything that could be considered a privacy leak any more, so I’m not deleting it. The auto reply doesn’t even mention one of my new email addresses.
When I say Freedom does my email, I mean most of it, because the email I use for this blog I receive at home. This is also the address I use to subscribe to mailing lists. Because they go through my own server, I can do a rather nifty thing with them. I take them apart and convert them to RSS feeds, so that I can import them in FreshRSS, my feed reader. I find FreshRSS more convenient to consume blog posts and articles than an email client, which is, after all, meant for conversations. It use a script that you can find here, with instructions, if you want to do the same.
Let’s say, you’ve been daily driving Ubuntu Touch for a few weeks and it kind of seems to work for now and around the same time on your PC you really miss grabbing windows easily so you’re returning from Gnome to KDE, which apparently also means going back from pass(1) and git(1) to KeePassXC and Syncthing.
Funny how one choice leads to a lot of others, sometimes.
Switching from KDE to Gnome early last year wasn’t because I found Gnome that much better. It was quite different from KDE, but it was the little annoyances, like not easily switching between audio outputs, that was so elegantly solved in Gnome, that made me bite the bullet and switch.
Selecting audio outputs in KDE
Using Gnome for about a year made its own little annoyances surface. I was moving files by cutting and pasting instead of dragging them, and at some point I actually copied a file while trying to move a window. This madness has to stop, I heard myself thinking, and I installed KDE, re-logged and noticed that switching audio outputs was taken care of.
Pass is a command line password manager that uses git(1) to synchronise between devices. It stores each password in its own GPG encrypted file and uses a folder hierarchy for storage and fast and easy retrieval, especially with excellent fish (and bash) tab completion. For Android, there’s at least an unmaintained app that works. For Ubuntu Touch, there is one as well, UTPass. It is maintained, but for as long as I remember, it has a bug making it show just a white screen instead of its UI. Many of the apps for this platform are maintained by a single person. UTPass is no exception.
To my surprise, after logging into KDE, pass(1) didn’t really seem to work here either. Just doing a pass -c something popped up a window asking for my password (fair enough) but then pass(1) would just sit there without giving me my password or my prompt back, as if something was getting in the way.
Unsure what was going on, I took a long hard look at my phone laying next to the keyboard and decided not to spend the time troubleshooting why gnome-keyring was still active and probably screwing up ssh-agent, but to take the opportunity to again install KeePassXC, the last password manager I used before switching to pass(1).
Let’s go
There was one small problem though. While there are plenty of scripts to import passwords to pass(1), exporting from pass(1) to KeePassXC appeared not already facilitated. So I spend the afternoon creating a python script that traversed my ~/.password-store, decrypted all the GPG files and put everything in a CSV file that KeePassXC would accept. You can find it where I publish my code.
After some debugging, the script worked, and I could import my passwords into KeePassXC. I added a folder to the directory that I share with all my devices using Syncthing and put the password file there. Then I opened up the Open Store on my phone to install KeePassRX, finding that an update for UTPass had become available, fixing the white screen bug.
I wonder whether I will keep using KDE or that new or old annoyances will (re)surface, making me escape back to the windows with too crowded title bars and with corners that are so round that, a couple of years from now, they will probably be circular.
The other day I was listening to the latest PostmarketOS podcast in which they spoke about their participation DEFCON 33, WHY 2025 and FrOSCon 2025, and it sounded like they had fun, sold merch and got stickers. They also briefly mentioned that a meaningful difference between Ubuntu Touch and their own PostmarketOS is that the former still uses a layer of Android for hardware abstraction (Halium) while they themselves attempt to go without that.
Ubuntu Touch uses Libhybris to talk to Android HAL
They didn’t mention it, but this could be what makes porting devices to PostmarketOS very much harder than for Ubuntu Touch, as indicated by the number of devices offered on both sides with full and complete hardware support. Ubuntu Touch has many, PostmarketOS just one, the Purism Librem 5, a device that wasn't build with Android in the first place. GPS fixes with Ubuntu Touch can take some time, but the warning notice about that is just to prevent disappointments. I have a OnePlus Nord N10 that gets a GPS fix within half a minute. Although much slower than with phones tied to Google or Apple, this really is fine.
Sharing pictures
Another thing they mentioned was that people who daily drive Linux phones always find workarounds for what they’re missing, such as actual snapshot cameras. I do that too, but not only won't let it me share immediately the pictures I took, I also cannot show them on my phone when I'm with someone. But while hardware problems mire PostmarketOS, it’s the state of the software that is Ubuntu Touch’s problem. Native apps are few and far between, and more than a few need fixes and updates.
The default calendar app sometimes ignores timezones, the running tracker doesn’t reliably track runs, the contacts app doesn’t sync contacts and sharing images to the matrix client doesn’t work, to name just a few issues. It particularly doesn’t help that PostmarketOS apps don’t run on Ubuntu Touch (and vice versa). You can install snaps and even NixOS packages, but those are hit or miss. With so few apps, there will typically only be one app for each of your use cases, if at all, and if that one doesn’t suit your needs, you’re out of luck.
Great, but not too great
Just yesterday, at an orchestra rehearsal (I always have a lot of those at this time of the year, which is precisely why I’m blogging less at the moment) a friend told me he ditched his smartphone altogether. Too many stimuli, he said. I think that is great, if not outright brave. Sometimes I’m in a shop wondering why my payment isn’t accepted and realising I didn’t put enough money in my account. Not having a smartphone means not being able to correct that then and there.
But you don’t just use smartphones for yourself. My friend is in fact the ensemble’s project leader. Today I spoke with someone else from his team. She told me that she and others now have to send him SMS text messages, while the rest is reading everything in their WhatsApp group. So now everyone has to type everything twice, which often turns out to mean that he misses things.
Perfectly
I like to tinker with technology and am generally okay with things not working perfectly for myself, but often I don't have the time and things do need to work. I also need people to be able to reach me without it becoming too much work for them. So even though PostmarketOS gives me more freedom, Ubuntu Touch is in fact more suited for me. And adb(1)’ing into a phone and then adb(1)’ing from it’s own terminal into a virtual LineageOS device running on it (you’ve got to have a banking app) is just funny. And now that I can use Matrix for WhatsApp and Signal. I could probably daily drive my N10.
In July 2024, I bought a monstrous laptop, a huge Acer Nitro 5. I installed Pop!OS on it, called it Flexo because its red and black colour scheme can make people think it's evil, and left it at that. Since then it has been a gaming machine for my kids and served me well during cold astrophotography sessions.
# echo flexo > /etc/hostname
Actually, later I installed Arch on it, simply because it is more up-to-date, and I’m used to it. But that wouldn’t be the last operating system it’d have to endure. Just last week, I needed to flash an older phone with an even older version of its own operating system and in order to do that, software, provided by its chip manufacturer, was needed, and this software was not available for Linux, Arch or otherwise.
Because astrophotography is a rather data intensive hobby (or potentially: the climate of my country doesn’t allow for lots of clear and moonless night skies), inside Flexo there are an extra M.2 and SSD, both of 1TB capacity. Since I expected the Windows installer to name them differently, I copied the files of both of them to the other while downloading Windows 10 and putting it on a bootable USB stick.
That last command took a very long time, and maybe it wasn’t even needed, because the Windows installer still complained that there was no suitable partition. I was able to delete the primary partition of one of the 1TB disks and then chose to install Windows on it. I had no way of knowing then if that was the M.2 or the SSD1. After installing, the installer rebooted Flexo directly to Windows. And then again. "Sit back and relax while we work our magic", it said 😂. And then it did a third reboot.
After all this I had a working Windows 10 machine. Flexo started without even offering me to boot Linux. Earlier, I had booted it with an Arch USB stick to fix the /etc/fstab file (it needed removal of a line referring to a disk) and reinstall GRUB to create a boot menu with both operating systems.
But that boot loader didn’t load. Instead, Windows was booted automatically and F2, the key that would previously get me into the laptops firmware, just locked the machine. Booting with F12 did allow me to select either the Windows or the Linux bootloader I installed earlier. Selecting that let me boot either system as well, so that was the one I wanted to automatically pop up when starting the machine. But since I couldn’t get into the firmware, there was no way for me to set it as the default.
The problem appeared a bug in the firmware making it choke on the Linux bootloader entry. The Arch forums, as always, had a solution. I had to boot into Windows and open Command Prompt as Administrator, then Mount the EFI partition:
This indeed landed me in Arch’s UEFI partition, where I could rename the bootloader file to stop it from confusing Acer’s firmware:
S:\>ren grubx64.efi grubx64.efi.bak
I rebooted, after which F2 presented me again with the firmware interface. The only problem now was that in its boot order form, I didn’t see any of my UEFI entries, just the number “2”. When pressing F5/F6 to change the values, only unreadable junk appeared. What could I do but to select one at random and hoping for the best.
After selecting the wrong one first, the one starting with "2sy" appeared the right one. So now Flexo is a dual boot machine, something I haven’t owned for two decades. I had flashed that phone days before finding the solution above, but perhaps I will have to do a similar thing in the near future, so I figured, as long as its not in the way, why not keep it there?
Last week I wrote how I basically used my new Atari STE to recreate my father’s professional music setup to let it once again play back the music I created between 1992 and 2002. But while C-Lab’s Notator software ran so that I could load and edit the files, one thing was still missing: A Roland SC-55mkII Multi-timbral Sound Canvas:
Neatly tucked away sits my Roland SC-55mkII in the previously unused space below the monitor
Or at least, something that would actually play the music so that I could hear it. The SC-55, being one of the MIDI devices that was recommended to me by the Atari community as one that plays nice with games that support MIDI, was not cheap, but does create a very high quality MIDI sound. On eBay, I found a Japanese shop simply called Books and Music, and they carried one. In the package they send me, they included a cute anime styled leaflet with a 10% off coupon code valid until one minute before the end of this year.
Perhaps, if they have a 240 to 100 volt adapter, I might buy it from them, because at the moment the SC-55 is occupying the only one I have, which means my also Japanese Sony MSX2+ is currently without power.
According to soundprogramming.net, the SC-55mkII is the successor to the SC-55, mostly adding some samples and increasing sound resolution from 16 to 18 bits. It can do 28 sounds at the same time, has an orange backlit LCD screen, pulls 9V and 900mA, is 218 mm x 233 mm x 44 mm small, and weighs 1.4 kg.
The Roland SC-55 MK II, front
When it turns on, the SC-55 presents its type number through a fancy little animation. Below the power button is a second MIDI in port. Incoming signals from this port are mixed with those received by the first MIDI in port in the back. This way, a computer can play a song through the SC-55 via port one, while a pianist can add impromptu accompaniment over port two.
To the right of these are a volume dial and a 3.5mm headphone jack, but inserting headphones won’t cut the output from the output jacks in the back. To the right of the screen are the “all” and “mute” buttons. To mute all channels, press both, to mute a single part, press “mute” and select the part with the parts buttons.
The other buttons come in pairs that change values, such as reverb, pan, or instrument, up and down. This can be done for all channels at once or, using the parts buttons, for one or more parts. The values can be seen in the left part of the LCD screen. The rest of the screen is filled with a bar graph with one bar for each of the 16 parts, where each part is a MIDI channel.
Originally, this device came with a small remote control, but mine didn’t. It doesn’t matter. I need it as much as the buttons on the front panel, not at all, that is. My sequencer, C-Lab Notator, has a much more convenient interface for everything these buttons can do.
The Roland SC-55 MK II, back
The rear panel has stereo audio input and output ports at its far left. What comes in through the input ports is mixed with the sound created by the SC-55, so this is where I connect my Atari STE’s stereo output ports to. The result is a perfectly balanced combined audio signal that continues to my little mixer panel that also has incoming cabling from my MSX and its many audio expansions.
Centred in this panel are three sad looking MIDI ports, the one at the far right labelled “IN 1”, which is what I connect the MIDI out port of my Atari STE with. Unlike the Atari, the SC-55 has a proper Thru port, which sends a duplicate of whatever the “IN 1” port receives further downstream.
There is a switch that lets you use other input mechanisms than MIDI and at the right is the AC adapter jack with a cable hook to prevent accidental disconnects. I don’t use it, but I can imagine it being a lifesaver on stage.
It’s funny to see how little is needed for a computer that is barely fast enough to play an MP3 file to drive a device like this. C-Lab’s Notator allows you to do edits while playing your song, all the while updating its progress numbers without pause. Considering there were even MIDI sequencers for 8 bit 3.5Mhz MSX computers, this is hardly surprising, but impressive nonetheless.
