name resolution

[Xen]

Tom H schreef op 26-11-2017 16:02:

> Because they're come to the conclusion that it's in their users' 
> interest.

Well that's a bit like saying I wear shoes because I want to.

Doesn't mean it's not cold outside or that the terrain isn't rough.

> Coexistence is simple and reasonable: if you want to use the defaults
> that Ubuntu and other distributions ship, don't use ".local" as a
> private domain name.

That's not really coexistence.

That's bailing out.

That's like living on Mars in order to coexist with Earth on which you 
have been excommunicated.

> In fact, with the new TLDs, private domains might have to disappear.

I... think that's nonsense.

> There was a fedora-devel@ thread two or three years ago about dnssec
> where Lennart pointed out that Fritzbox is the most widely-used home
> routers and that the admin page is reached by going to "fritz.box".

That's nice.

> ".box" must be registered by now (I'd guess by "box.com" or
> "dropbox.com" but I don't care enough to check) so Fritzbox'll have to
> change something in its setup.

No the two contenders are (I just checked for you/me/us) Amazon and a 
company that wants to use it for personalized cloud, decentralized 
cloud, in which you have your own cloud domain name.

It hasn't been decided with.

Yet.

But Fritzbox can register fritz.box if the other company wins out and 
even if they didn't, the only thing that would happen would be that the 
actual fritz.box domain would be unreachable.

I think it is folly to assume that someone else would get or should get 
precedence on that name.

At worst what you will get is that certain private domain names will get 
reserved for private use if it gets out of hand.

I think.

I mean why assume the worst?


>> Like I said, the solution outlined....
>> 
>> Is not terribly hard.
>> 
>> 1) accept a max 30cs delay in local uncached requests and accept 
>> leakage
>> onto the internet towards domain servers
> 
> Why should avahi and the distributions implement such nonsense for the
> handful of people who are too inflexible to avoid one specific domain
> name on their lans?!

And you don't consider yourself inflexible???



>> 2) have no delay but accept leakage onto the internet towards domain 
>> servers
> 
> There's already too much leakage. IANA should probably add ".local" to
> the blackhole/prisoner dns servers.

That's only used for reverse lookups.

>> 3)
>> 
>> check for the existence of a local SOA record in one of the configured
>> nameservers and if it exists, let dns queries for .local precede mDNS
>> queries for .local
> 
> Same answer as (1). What a complex mess for so little benefit.

Why.

You just don't care about me right, or people like me.

Of which there are plenty.

Your "complex mess" is not more complex than the mdns_minimal plugin 
itself, and in fact much less complex.

What is so complex about a rule you can describe in 3 lines of text, 
takes an hour to implement if you are competent, and wouldn't take much 
time to turn into something official if you had the expertise?

I mean if you don't care about my use case at all you should just say 
so, instead of pretending you care but that you think I won't be harmed 
that much.

But I know for a fact now that I will actively fight this nonsense.

Because you declare war on people like me.

So I declare war on your nonsense.

The RFC indicates that a mixture of solutions is perfectly fine.

But none of you even comments on that.

You cite the RFC and the ICANN and the IETF when it is to your benefit,

but the moment the official document actually proves you wrong, you shut 
up.

Or you start screaming like Liam.

Like LIAM.

;-).

Sorry.



You were all raving about STANDARDS but now that the standard proves you 
wrong I hear nothing?


>>> But
>>> it's not full coercion. If you want to use ".local" as an internal
>>> domain name and you only use Linux and BSD on your lan, you can.
>> 
>> I know. I just didn't know how until Liam told me.
>> 
>> I mean for zeroconf that is pretty hard.
> 
> Because changing the default config isn't ZEROconf!

That default config shouldn't need to be changed.

It should work for all.

The only reason it doesn't is a certain lack of caring about people.

Do you realize there would be millions of people who would run into this 
if they actually did use Linux?

How utterly unintuitive it is that if they set up some DNS server using 
the most obvious domain, that it wouldn't work?

You don't see this system as flawed at all?

A completely unexpected thing?

Good software doesn't do unexpected stuff.

Even more you make it impossible for routers, if they did support this, 
to actually deploy this.

You have made impossible a mass-market deployment of DNS-equipped 
routers because you want to have it YOUR way.

What are those routers supposed to pick?

Everyone has been using .local for decades.

It is the most obvious choice, which is why the RFC picked it.

And the RFC also explicitly allowed or suggested a living together, but 
you ignore that.

There is no good reason why it shouldn't happen.

Oh of course there is: if you don't want people to have their way.



> I have no idea whether Windows uses some form of avahi by default;
> there are probably many Windows systems with iTunes installed and
> iTunes brings in (or used to bring in) Bonjour unless you unticked its
> installation. MS didn't like Bonjour so it created its own, less
> powerful LLMNR (on a different multicast port).

Microsoft long suggested and promoted .local as the local domain.

It sounds a lot better.

Thank you for the link.

But of course it also has its own I am so important statement: ;-):

"Stub resolvers supporting both DNS and LLMNR
    SHOULD avoid sending DNS queries for single-label names, in order to
    reduce unnecessary DNS queries."

Which is also a bit bullshit.

But at least it doesn't say MUST you know.

So now we have competing systems that all want you to behave in a 
certain way.

mDNS actually doesn't mandate it, it is just Apple's and Linux' 
implementation thereof, that isn't required.

This one wants to kill the "search" directive, lol.