“Welcome to systemd conference. Lunch is served.”: Lennart Poettering's keynote

* 1'17" "Breakfast and dinner is on you."

* 2'06" "If you're not there, then the boat will leave without you."

* 11'42" "I'm surprised, myself, that GitHub isn't perfect. Because it's so popular, right?"

* 13'55" "We won everything that we could have won. There's nothing else we can win. Except for Gentoo maybe."

* 14'40" "Fedora's and Ubuntu's cloud editions nowadays default to systemd networkd. Which I personally find pretty amazing given that networkd doesn't even have a [Desktop] Bus interface."

* 19'01" "I use resolved nowadays on my laptop all the time. But we have not pushed it into any of the distributions. Simply because ... well I think it's better in many ways than the glibc resolver. It actually doesn't deliver anything really big for anybody yet."

* 19'40" "Most of the distributions, like Fedora, currently do that with running a local DNS server and things like that. That's something that we really don't want."

* 20'16" "The way I see resolved, what it should be, is that it does name resolution. And whether that name resolution is traditional DNS, or it is LLMNR-style Microsoft-style link-local DNS, or whether it is Apple-style link-local DNS called mDNS, should just be an implementation detail."

* 22'37" "[sd-event] doesn't try to be portable. It's actually the opposite of being portable. It's basically epoll(), a wrapper around epoll()."

* 27'34" "I personally think that user-space IPC is a really crazy idea. I really hope that we can deliver something with kernel IPC."

* 32'28" "I know that distributions have these callout scripts so that when a DNS server changes they rewrite BIND's configuration files and things like that. That's the wrong way around. Make BIND listen to the events. BIND should just take the data. It's what we do with resolved, for example."

* 34'11" "Something we will probably be not working on [in systemd networkd] is Wi-Fi support, though, any time soon. That basically means that it's not going to be a consumer network daemon. It's not going to end up on your desktop because it's useless, if you don't have Wi-Fi."

* 35'27" "I personally think that the way that [user namespacing] is implemented in the kernel is borderline crazy. I also don't think that it's usable yet."

* 37'29" "In a systemd-nspawn container you get 16-bit user IDs, how it used to be in Unix. And the upper 16 bits of the host user IDs are basically just container IDs. And the lower bits are the normal ones. What we can say is that we implemented that in systemd, and we kind of try to push people to a certain kind of [user ID] allocation when they do that."

* 39'04" "And then there's machinectl shell. I got some very confused reporting in the press: that we'd reinvented sudo, or something."

* 42'57" "Eventually I'd really like to get rid of systemd ask-password and move it to some D-Bus interface."

* 54'54" "We are not the guys to do abstracted back-end support supporting ten different back-ends. We really are the guys who want to push people towards using one and not the other."