Re: [mif] [DNSOP] [dnsext] 2nd Last Call for MIF DNS server selection document

[Keith Moore <[e-mail]>]

On Oct 21, 2011, at 11:11 AM, Ted Lemon wrote:

> On Oct 21, 2011, at 10:04 AM, Keith Moore wrote:
>> And honestly I don't see why handling of non-DNS names like "foo" is in scope for MIF.    
> 
> Because such names are typically resolved using DNS search lists, and at lease one mechanism for setting up search lists is interface-specific.

I don't think it's MIF's job to try to make all existing hacks work, while limiting its scope to specifying how hosts and apps implement things.  

I think it's potentially reasonable for MIF to say things like "here's how you should configure your networks if you want them to be usable from hosts with multiple interfaces."

Also, the subject of multiple active interfaces per host exposes a number of cracks in the Internet architecture, and also exposes cracks in some of the hacks that people have used to work around cracks in the Internet architecture.   IMO, MIF should not be trying to add more hacks.  MIF should primarily do what's best for the Internet architecture in the long term, realizing that IPv4 and therefore RFC 1918 are at EOL.  Hacks to accommodate the existing world should be considered of secondary importance, and should only be considered if they don't pollute the architecture in the long run.

Also, it's arguable that v6 link-local addresses should not be used by applications, even on ad hoc networks, because randomly-generated ULIA prefixes are much better.  And the way you figure out which interface(s) to use in order to reach ULIAs is via routing protocols, not via assuming that a DNS query is specific to a particular network interface.