Carlos E. R. wrote:
> On 2013-06-12 21:07, Jim Henderson wrote:
>
>>> Well, their choice to have software that doesn’t work the way it’s
>>> supposed to, unless they can cite something that says that anything
>>> other than 127.0.0.1 is invalid. Clearly the IETF thinks differently,
>>> and they’re the standards body involved.
>> Here’s an idea: Point the openSUSE developers (via a bug) at the postfix/
>> whomever developers’ bugs where they say that 127.0.0.2 shouldn’t be
>> valid and suggest that the two groups of developers work together for a
>> solution that works for everyone.
>
> I already did. Kind of. 
>
> I have not seen an email archive of the postfix mail list, so I can not
> post a link to our exchange.
Google quickly showed several archives listed at
http://www.postfix.org/lists.html, for example:
http://news.gmane.org/gmane.mail.postfix.user/cutoff=237564
It seems to me that this discussion isn’t getting very far very fast. I
think that may be because people haven’t really understood what the
issue is. I certainly don’t feel that I understand it properly yet.
I would suggest starting again, with a focus on understanding what the
problem is, rather than whether or not we should vote for a particular
solution. A difficulty with understanding the problem is that it appears
it is an old one with various bits of evidence and discussion scattered
around and a lot of context to pick up.
So I think it would be a good idea to focus on two particular points.
Perhaps Carlos could explain and confirm them as well as he can and
others can fill in with more information.
(1) One claim is that some applications try to resolve the hostname and
fail (or take a long time to initialize) when there is no network
present. This sounds like a bug in these applications. Is this still
true? Which applications currently have this bug?
(2) It seems that in order to work around #1, SUSE adds a line to
/etc/hosts for 127.0.0.2 associating it with a hostname. In the example
Carlos gave in the postfix archive, the hostname was identical with the
canonical name associated with a real LAN address also listed in
/etc/hosts. Note that the hostnames were obviously fake, not real, so it
does not prove a problem. Most of my machines have an association for
127.0.0.2 listed, but in no case does it match one for any other
address. However, I manage my machines in weird ways, so my
contraindication doesn’t mean much So what exactly does openSUSE
currently put in /etc/hosts under what circumstances, and does it ever
list two IP addresses for the same hostname? If so, is that valid thing
to do under whatever rules govern the content of /etc/hosts.
(3) It seems that for some people, the workaround #2 causes knockon
problems with another set of applications, of which postfix is one
example. Apparently, the postfix conf supplied by SUSE provokes a
problem, whilst the default conf supplied by postfix upstream does not.
So why does SUSE supply a different config and can the default be used
instead? What other applications currently have what knockon problems
from this #2 cause
Note that there is no point in wasting much effort on #3 (except for
finding workarounds on affected systems) until it is established that #1
is still an issue nowadays and so workaround #2 is still necessary and
also that workaround #2 is actually valid and is installed by the system
rather than by manual intervention on particular systems, for example.
