Implement new config file that is honored regardless of MINIMAL configuration

[agoode]

In issue #46 and #64, it is clear that we want to have at least these three options configurable independently in all cases:

• Restrict to 169.254.0.0/16 for reverse lookup (currently true in MINIMAL, false otherwise)
• Enable two-label limit heuristic (currently true in MINIMAL, false if non-MINIMAL and /etc/mdns.allow exists)
• Enable unicast SOA heuristic (currently true in MINIMAL, false if non-MINIMAL and /etc/mdns.allow exists)

Right now, all configuration is done through /etc/nsswitch.conf and /etc/mdns.allow. If a MINIMAL variant is used, a set of hard coded settings are honored. For non-MINIMAL, a combination of different hard coded settings and configuration from /etc/mdns.allow is used.

I propose a new set of config files that can be used to specify all these parameters for any nss-mdns configuration:

• /etc/nss-mdns.conf
• /etc/nss-mdns_minimal.conf
• /etc/nss-mdns4.conf
• /etc/nss-mdns4_minimal.conf
• /etc/nss-mdns6.conf
• /etc/nss-mdns6_minimal.conf

We might not need all these files, in which case we'll just use /etc/nss-mdns.conf.

[lathiat]

I think the main angle of #64 is firstly that it broke after upgrade. While I understand that concern, I don't think that should be fixed and the default should match the expected standard as implemented in other operating systems, i.e. the current shipping version.

I don't mind the idea of allowing an option to allow >2-label hostname lookups in minimal-mode as a config option for users such as @Jack12816 who are relying on this behavior. But I don't think you should optimize for that case.

I find it relatively unlikely people are going to want to configure this behavior differently for IPv4 & IPv6. So I think if you want to go through the config file route, a single nss-mdns.conf would suffice.

But the second angle was that the only fix was to potentially have reverse dns lookups timeout slowly by using non-minimal.

Based on this logic

Enable two-label limit heuristic (currently true in MINIMAL, false if non-MINIMAL and /etc/mdns.allow exists)
Enable unicast SOA heuristic (currently true in MINIMAL, false if non-MINIMAL and /etc/mdns.allow exists)

Couldn't in theory you use mdns_minimal with an mdns.allow of:
.local.
.local
254.169.in-addr.arpa
8.e.f.ip6.arpa
9.e.f.ip6.arpa
a.e.f.ip6.arpa
b.e.f.ip6.arpa

To match the behavior they desire?

It is perhaps slightly obtuse, but I really think this is going to be a relatively uncommon use case. The only time this happens basically is if you setup entries in /etc/avahi/hosts - which I am sure some people have done, but I doubt it's many and this wouldn't happen with other operating systems.

Wouldn't that save adding extra configuration file logic over the existing logic, or have I misunderstood the logic and it doesn't actually work like that?

[lathiat]

Sorry I meant use mdns (not minimal). This basically lets you use mdns, resolve all .local hostnames including reverse hostnames, but only for the link local subnets (the normal _minimal behavior)

[imrehg]

The braking changes as #64 mentioned is an issue that got me here too. ArchLinux ships this by default, and your current readme says:

If the configuration file is absent or unreadable nss-mdns behaves mostly as if a configuration file with the following contents is read:

# /etc/mdns.allow
.local.
.local

which is clearly not the case. Thus it's both the breakage, and incorrect documentation.

[agoode]

Sorry about the documentation. The wording was made clearer in fd0cab5.