One of the design guidelines of Jailer is don’t break FreeBSD. As in if someone installed and used Jailer, and then deleted the Jailer binary and libraries, their Jails would still run without any issues. We do this with minimal intervention, for example,
jailer init patches FreeBSD’s
/etc/rc.d/jail, but in a way that you wouldn’t feel the difference much. We don’t create new
rc.conf variables, we just change couple of loops. In a way, you can keep these changes even if you delete Jailer so your system would be much improved. Obviously, we do sent these patches to FreeBSD src.
But I’m in front of an issue right now. On one side, I want to keep these guidelines, on the other, pushing the limit will allow me to improve Jailer way more than I expected.
These are the things that I think about before sleep, or during the shower. I gave a promise, that I will not break the Jail ecosystem. But what if, just what if, the ecosystem was broken in the first place?
Some of you might know, that we’ve been working on integrating libucl with Jail. The experiments have been going well, in such that I feel I want to integrate these experiments with Jailer already, even before they get into FreeBSD (and they might even not get in at all).
My dream of Jailer and its ecosystem is complex. I feel that these integration would do good on the long-term, but I want to keep the short term alive as well.
One idea is to fork Jailer, keep two versions of it. One version that’s FreeBSD compliant, and another one that is pushing the limits.
This is going to be an interesting week…
That’s all folks…