Hmm. that was interesting. When I tried the nmcli command with $NAME of System\ eth0 (the name of the connection), I found that I was getting two ipv6.dhcp-hostname values. I’ve now got a new configuration file: /etc/sysconfig/network-scripts/ifcfg-System_eth0 (!) I’m not very familiar with how NM is supposed to translate to/from underlying networking implementations (e.g. the redhat ifcf-* files + networking-scripts). I’d guess that NM abstracts the different models to its own object model and reads/writes to the underlying configuration files + some files of its own. I can see it being a challenge keeping both models consistent as the data must be round-tripped (e.g. read from fileX, modify, write back into the correct part of fileX.), and sometimes data elements may be ignored (so their fate may be ambiguous). Should I report a bug for the above behaviour (assuming that I can get a repeatable approach)? Tim |