sendmail init error: Can't assign requested address

[Chuck Swiger]

On May 15, 2007, at 9:06 AM, Ernest Sales wrote:
[ ... ]
> Honestly, I don't understand what each of this four daemons is  
> supposed
> to do. I just want the minimal working sendmail config in a NATed  
> host,
> the /etc/defaults/rc.conf reads as your sample, and init says
> sendmail_outbound_enable is set to NO, which seems odd but dunno the
> consequences.

There are only two daemons, actually: the MTA, and the client mqueue  
runner.

The separation was made because sendmail used to run as a single,  
setuid-root executable, and has had a rather infamous security  
history as a consequence.  If you want sendmail to be running and  
listening on port 25 as a MTA, you need to set the sendmail_enable/ 
sendmail_outbound_enable to YES.

[ ... ]
>>> Is there any standard, anything like the CIDR blocks reserved for
>>> private networks?
>>
>> The zeroconf/rendezvous stuff likes to use ".local" as the domain
>> unless other info is available.
>
> Cool. Tried .local and works too. Looks like sendmail is happy with
> finding 'dot anything' after the hostname. So far, my problem is  
> fixed.
> But the init behavior for unqualified hostnames is less than optimal:
> having to wait one minute until sendmail agrees --and it finally
> agrees-- is annoying; and this happens for every sendmail daemon  
> launch.
> As more end-users using PCs without FQDN jump to FreeBSD this could be
> more heard of. Wonder if filing a PR; comments welcome.

The standard period for a DNS timeout is anywhere up to about two  
minutes, depending on how many resolvers are configured in /etc/ 
resolv.conf.  It's possible to tell sendmail not to use DNS, and  
avoid this timeout, but normally people run mailservers only on  
machines with working DNS and a sensible hostname.  This isn't a bug,  
it's just an assumption that sendmail makes which is typically  
appropriate, but not for the case of a random client machine without  
working DNS....

-- 
-Chuck