On Mon, Oct 07, 2013 at 05:22:24PM -0500, Derek Martin wrote: > On Mon, Oct 07, 2013 at 10:56:13AM +0300, > Alexander Gattin wrote: > > In my opinion `hostname` in the kernel is just > > a hint. > > Well, I mean seriously, if you're going to tell > me that a required part of configuring > networking is just a "hint" then I thank you for > giving me what I need to excuse myself from this > conversation.
You're just not bold enough to consider other possibilities. Yes, /proc/sys/kernel/hostname is just a kludge/simplification. > It is non-negotiable that all machines > configured for TCP/IP networking require a > hostname, and that that hostname must resolve to > a valid IP on that machine. Nope, hostname is just host's notion of its name, nothing else. Message-ID is based upon host's and user's notion of how his/her machine should be named [in a mail/thread/conversation]. If you think that Message-ID is kind of biometric passport and should represent some notion approved by authorities or at least [technically approved] by networking infrastructure (name resolution and/or routing), than your patch is still wrong. > Your position is essentially that it's better > for Mutt to get it wrong sometimes when the > config is 100% correct, in order to be able to > sometimes deal with a configuration that's > absolutely broken Nope. The function you've hacked is called > int getdnsdomainname (char *s, size_t l) and IMHO it should return "primary" domain. Your patch does this in much worse way than what was originally implemented. P.S. Sorry for everyone for this long discussion, it just concerns complex matters. -- With best regards, xrgtn
signature.asc
Description: Digital signature
