I doubt this is actually the case. I wouldn't trust the help in smit as documentation. <div>&nbsp;</div> If the hostname is unqualified, the system uses the 'domain' parameter or first domain listed in the 'search' parameter of resolv.conf. <div>&nbsp;</div> Many people have operated this way for many years. I've not had IBM support once question having the hostname set to an unqualified name. <div>&nbsp;</div> http://pic.dhe.ibm.com/infocenter/aix/v6r1/index.jsp?topic=%2Fcom.ibm.aix.files%2Fdoc%2Faixfiles%2Fresolv.conf.htm <div>&nbsp;</div> Maybe you could get an official answer or explanation for us from IBM engineering?

No, really, it is the actual case. You may get away with short names but it is not correct and if you use software that connects to machines across sub-domains and/or have machines with the same name in different sub-domains you can get into a right mess. If you want to run your computer room using luck and with no thought to future software then you carry on. As for an "official" statements and engineering - very funny. If you don't trust me then you would not trust them either.

So help me understand here: If IBM requires fully qualified domain names for hosts, and thus puts this in the help information in smit, then why doesn't smit give an error if you try to put only a short hostname in there, while at the same time a domain name is provided - clearly DNS is then used, so smit should not allow a short hostname, right? <div>&nbsp;</div> And, if an underscore is not allowed in the hostname, why does smit allow it, and why isn't that mentioned in the help info?

Meilinkm - Smitty was written 25 years ago and when networks were optional! I think the idea is that smitty assumes the user is right and knows what they are doing - just like most UNIX commands do what you ask and what is right nor what your really meant to type! If you think this is wrong in this case (baring in mind not every one uses DNS) the you can raise a Design Change Request (DSR) and send to IBM and if deemed important and a high priority then AIX/Smitty may get changed. On the underscore this is part of the DNS RFC specification that only allows 26 letters (upper and lower case), numbers 0 to 9 and the hypen. See http://www.ietf.org/rfc/rfc1035.txt Every one knows this :-) as it is standard bed time reading for techies. You could raise further DCR's for this but smitty can't force you not to add duff characters in to the relevant files or commands directly so the philosophy is the user knows best. Personally, I like smitty to be fast rather than enforcing loads of rules.

Actually, I believe that smit(ty) is correct. Rember to read what is there, not what you are thinking. <br /> Literally, WHEN DNS is used for hostname resolution FQDN must be used. This is to prevent errors when only the short hostname is sent to DNS plus, "POTENTIALLY" a list of domains to search. THE DNS might not give the correct IP address back. <div>&nbsp;</div> To be able to use shortnames AND to not have startup issues in case DNS is unreachable I have adopted the standard of adding the line to /etc/netsvc.conf: <br /> hosts=local4,bind4 <div>&nbsp;</div> (note the 4 behind local and bind mean only IPv4 requests are done) <div>&nbsp;</div> so that for THIS host (actually all requests, but the only entries are localhost and my hostname). <div>&nbsp;</div> Yes, MOST of the time a shortname will work fine with DNS, but it might fail. Using FQDN should prevent that. If not, you have a configuration elsewhere (i.e. in your DNS database(s)).

bother - formatting here (all the <br /> in my previous comment) is a real real bore. And no chance to correct spelling errors either! :(