I am begging software projects to release programs as tarballs or plain old static binaries. I don't know what a Helm chart is. I am not installing an entire Kubernetes cluster. Docker breaks LXC and firewalls and is constantly breaking for mysterious reasons. I would rather submit myself to hours of `./configure && make && make install` frustration than try to hack through the endless hellscape of containerized deployment systems.
@aphyr I am legitimately not sure where I am more annoyed by "here is our helm chart, good luck" or "here is a single static binary for you to run (assuming your system uses the same libc as ours) (assuming you trust blobs)" (looking at you here hashicorp).
@dalias Given the existence of services like nss and pam, I'm not sure statically linking libc is universally a good idea either. Discovering that one program disagrees with the rest of my system about how user IDs map to user names, or about how hostnames map to addresses, is intensely frustrating.
@owen Maybe, but the vast majority of programs have no legitimate reason to even care about the idea of users.
glibc making hostnames up to nss was just a horrible mistake.
@dalias an operational reality we may disagree with remains an operational reality nonetheless.
@srtcd424 @owen Not really. In #musl we honor nscd for user/group db because the protocol is workable, we aim not to NIH things, and it gives custom db backend support. But not the nss hostname stuff. It lacks sufficient capability (hosts backend can't do scope ids for link local addresses) and it has broken fallback logic that doesn't treat nonexistence safely, and DNS protocol already provides a more capable way to do custom backend if needed.
@dalias @owen @aphyr nodds in agreement
I want stuff to be portable and reproduceable!
The whole #PainAndFrustration with #LinuxFromScratch, bloated #GNUtils and #GlibC constantly breaking shit (among other things) is why I started @OS1337 …