Vincent's Blog

Fapws4 is my small and fast Python web server: a single C file built on libuv and picohttpparser, embedding Python, with one process, one thread and one event loop. Over the last few days it received its biggest batch of changes in a long time: two releases, 0.7 on 9 October and 0.8 on 10 October 2026.

This post summarises what changed, why, and what you have to check if you run an application on top of it.

In my previous post I built a minimal thick jail from pkgbase packages and ended with a baseline ZFS snapshot. That snapshot is more than a safety net: it is a template. With a single command you can stamp out as many jails as you need, each one ready in a second and using almost no extra disk space.

But clones come with a property that surprises people: a ZFS clone is not a plain copy. It stays linked to the snapshot it came from, and that link decides what you can and cannot destroy later. In this post I clone one jail into four, list the few things that must be changed in each copy, explain how upgrades behave, and show three ways to deal with the dependency between a clone and its origin.

For years, building a FreeBSD jail meant downloading base.txz, extracting it by hand, and later juggling freebsd-update(8) to keep it patched. With FreeBSD 15, the base system is also available as regular packages, a method known as pkgbase, and jails are where it shines: you install exactly the pieces you need, update them with the same tool you use for everything else, and keep the whole thing on its own ZFS dataset that can be snapshotted, rolled back and cloned in seconds.

In this post I build a minimal but functional thick jail (a full, self-contained copy of the userland) from scratch, from dataset creation to a running system with SSH. I also share the one mistake that cost me an afternoon, a userland newer than the host kernel that made the jail hang on shutdown, along with the commands to avoid or fix it. Everything here comes from a real FreeBSD 15.1 setup. Adapt the pool, dataset and network names to yours.

For years, keeping a FreeBSD system up to date meant running freebsd-update(8), which patches the base system as a single, opaque block. pkgbase changes that model: the base system is split into ordinary packages that you can install, query and upgrade with pkg(8), just like any other software. The pkgbasify script can convert an existing installation, but it rewrites the whole base system, which makes it a poor first experiment on a machine you care about. A jail is a much better laboratory. It is isolated from the host, cheap to create and, with ZFS, trivial to roll back. In this article I walk through converting a classic jail to pkgbase from start to finish, including the one jail parameter that makes the conversion fail if you forget it. I then show how to upgrade the result to a newer minor release, all without touching the host.

I've changed my dhcpd daemon from a Actif/passif system managed via ifstated, to an Actif/Actif system without ifstated. I decided to change mainly because this is a builtin feature. A bit complex to setup, but once I've found the correct setup, it's in act easy. My biggest problem was coming from the -Y and -y parameters of dhcpd.

On OpenBSD, /etc/daily and /etc/weekly quietly watch over the machine and mail you a report. On FreeBSD, periodic(8) does the same. I love that model, but when I wanted ZFS on my laptop, Void Linux turned out to have better drivers than FreeBSD for my hardware. The catch is that Void ships no periodic maintenance at all. So I wrote two small POSIX shell scripts, one daily and one weekly, that mimic the BSD reports and add checks specific to ZFS and to a laptop. They log everything, compare today with yesterday, and interrupt me only for real problems such as a degraded pool, a failing disk, or a kernel installed without its ZFS module. This post walks through each check, the command behind it, and why it is there.

I can put both at the top of void-zfs-maintenance-blog.md if you want. I can also write a shorter, punchier title, for example "Daily and weekly maintenance on Void Linux with ZFS, the BSD way", which matches the current heading of the post.

vdcron is my small cron-like scheduler for machines that are not always on, such as laptops. It is a plain shell script that runs the tasks that are due each time the machine resumes, boots or someone logs in. At the end of September I reviewed the whole script, and the result is version 0.9. It fixes several bugs that could lose tasks from the queue, corrects the date calculations, and adds an option to edit the configuration safely. Above all, progress is now saved after each task, so a shutdown in the middle of a run no longer starts everything again.

If you run FreeBSD on ZFS, you probably can't imagine upgrading without bectl anymore. Take a boot environment, run freebsd-update or pkg upgrade, and if something goes wrong, reboot into the previous state. No backup to restore, no rescue USB stick.

Three days in Brussels with the people who build BSD. A training with Kirk McKusick, talks on FreeBSD's future, a museum visit by tram, and long conversations with names I'd only known from mailing lists. I went in wondering how many people actually make a living from BSD. I came back with an answer, a few new friends, and a lot of respect for a community that mostly does this for love, not money.

This whole thing started because of Stefano and his blog. His guide to installing Void Linux on an encrypted ZFS root with hibernation support was the thing that got me curious enough to actually try it — and, as these things go, curiosity turned into an afternoon spent repartitioning a laptop I use every day. My setup below is simpler than his (no LUKS-encrypted swap, no hibernation), because I just wanted ZFS-on-root dual-booted alongside my existing FreeBSD/GhostBSD/OpenBSD rEFInd menu, but the bones of the approach — and the credit for pointing me at zfsbootmenu in the first place — are his.

This month, this blog turns 10 years old.

I want to use this anniversary post to look back at everything I've written over the last decade: what has stayed the same, what has evolved, and what I've replaced along the way.

Increasingly, new ISP boxes bundling fiber, TV, and all the bells and whistles no longer allow inbound connections from the outside. Unfortunately, that's exactly what I need — I want to reach my home server from anywhere. Thanks to WireGuard as set up on OpenBSD, there's a very simple solution to this. This post explains how to do it.

In the context of the DMA developped in this post, this blog post improve it so to have accents and all nice icons in emails delivered by DMA

This blog provides the precise configuration required to build a localized, self-learning mail server on OpenBSD. We assume your mail is hosted on a remote IMAP service and you want to filter it locally before your best email client (K-9 Mail in my case) ever rings.

This is an intermediary setup where I keep my provider as MX records, but I'm allowed to perform my own spam filtering.

Then by using imap server, I can publish cleaned inbox.

I've made lot of modificaitons inside Fapws4 those last days/weeks. It remains a strong and very fast WSGI server offering all flexibility of Python. This is mainly than,ks to the libuv library and the picohttp parser. This blog will show how Fapws4 resist to massive requests and to compare results with other SWGI solutions. I will even compare it with apache24.