13:35 <dustymabe> walters: let me know when you pull https://github.com/openshift/machine-config-operator/pull/3720 out of draft and I'll look over it again

16:37 <dustymabe> release notes for last week's releases (needs two reviewers): https://github.com/coreos/fedora-coreos-streams/pull/723

17:24 <baude> bgilbert: got a second to quick glance something ?

17:25 <bgilbert> baude: sure, what's up?

17:25 <baude> https://paste.centos.org/view/dee21512

17:25 <baude> seem reasonablish ? just want to ask before i start testing

17:30 <bgilbert> yeah, looks reasonable generally. one small thing and one medium thing:

17:31 <bgilbert> the small thing is that we should use unique identifiers. "1024" is a round number and "/" is a pretty generic HTTP path. let's pick an arbitrary Ignition port number and maybe use "/ignition" as the HTTP path?

17:32 <bgilbert> the medium thing is: do we actually want HTTP? or should we just have the server send the Ignition config down the socket and close it?

17:33 <baude> well, one nice thing about http in how i used it is that the protocol is known and no buffer size control needed

17:33 <baude> for podman, i did intend to kill after the file is sent

17:35 <bgilbert> buffer size control?

17:35 <bgilbert> presumably it's a blocking socket? you just send, and the kernel will block you if the client hasn't read enough

18:09 <dustymabe> spresti[m]: marmijo[m]: mind a review on the releaset notes PR (linked above)?

23:33 <bgilbert> baude: also note that if the config causes Ignition to change any kernel arguments, Ignition will fetch the config twice