I measured nspawn container startup with various IP configuration options. Total time from start to host0 carrier is consistently around 0.5s. Even with static IPv6 configuration, it takes more than 1s before container begins to respond, bringing the total start to reply time to almost 2s. DHCPv6 adds more round-trips with similarly excessive timings and takes another 2-4s. This makes IPv6 unsuitable for on-demand containers that get started to serve requests from interactive applications, and wasteful with containers that only need to run for a few seconds at a time as part of a low-frequency compute pipeline.

After spending a couple of weekends chasing down a particularly annoying bug, and ending up with half a line of a fix, I thought that the story of my investigation and the diagrams I had to draw to navigate the seedy corners of the Samizdat stack might have some value not only as an instruction for past and future contributors, but also as an illustration of limitations of the COCOMO model.