name resolution

[Xen]

Liam Proven schreef op 26-11-2017 22:24:

> I heard that in Brno, too. But actually, they are there, there were
> multiple small shops selling them, others repairing them.
> 
> People take "I don't notice anyone using Apple kit" with "nobody uses
> Apple kit."

I didn't try to make that statement.

If Apple is popular in other countries it changes the face of IT that is 
global, but I was just saying that where I lived, there was practically 
none of it.

> The point is, for 3y it was shipping. Nobody could use it, because MS
> didn't even offer drivers for it.
> 
> MS added drivers. _Still_ nobody used it. It remained rare and obscure.
> 
> Then Apple made it the only way to get stuff in and out of an iMac
> except over the network, and suddenly, USB was everywhere.

You mean because hardware vendors started shipping it in order to sell 
to the mac.

> They were only introduced in 1984. Before that, I suspect you may be
> thinking of the Apple II.

Yes.

>> I am talking about the lifetime of USB.
> 
> Yup -- from ~1996 to date. Pre Mac OS X. Apple was a very significant
> force, from that time, and shaped the PC and Windows and the whole
> industry, not just its own little niche.

Then apparently it didn't go along with USB 2 because it wanted to 
promote Firewire for that ;-).

Well I guess if it prompted or inspired pheripheral manufacturers to 
produce USB devices I guess you are right.

> And it still does. For example your Ubuntu machines print via CUPS, an
> Apple product. And Ubuntu machines find local printers via a standard
> that Apple implemented and made popular, even though Apple didn't
> write it or create it originally.

>> From 1995 till 2003 when I was *VERY* busy with computers as hobby, 
>> with
>> friends and at university, NO one every even mentioned the world 
>> AppleTalk.
> 
> No, because as you keep saying, you didn't use Macs and didn't know
> anyone who did. It's an Apple protocol for Macs to talk to other Macs
> and Mac printers and nothing else.

Friends started switching to the Mac during later OS X era.

One of my university friends became a Mac user -- he was more of a 
scientist and not a system administrator.

The other one is a system administrator and programmer and also became 
Mac.

> So did everything in the early, pre-Internet, pre-router days.

You mean before TCP/IP?

>> You consider that fair?
> 
> Yeah.
> 
> A previously-meaningless name, not a TLD, never available to real
> routed networks talking over the Internet, only available  inside
> internal networks, suddenly meant something.

You know the DNS stats that Tom linked proves otherwise.

.home and .local produce high traffic, I mean together they dwarf 
requests for something like .org.

That means they are not meaningless. And they are used a lot.

The .home case cannot result from mDNS.

So some people (a lot of people) have devices with .home suffixes but no 
router that blocks it.

The problem is here the attitude that all of those "people" are 
incompetent and misusing something.

But this "misuse" generally means "functionality".

Linux people are more than willing to sacrifice functionality for 
perfection.

So even though those people are having fun, watching movies, and having 
sex,

the Linux people then say with sour faces that they are polluting the 
internet.

> My home LAN was called KEEPIER.LOCAL. I had to rename it. But that was
> about 15y ago. Long time.

You are not a programmer.

So for you that is the logical choice.

You can't change those systems.

At least you say you can't.


> Why would it? Windows had its own protocol. Netware had its own too.
> WfWg and 9x supported both, out of the box.

People claimed that today, because "Apple", it now HAS to permeate the 
Linux world, and it's almost the same thing (ex the total network stack 
protocol suite idea) today as back then.

People in Linux don't really understand inspiration or what prompts 
what.

They think inspiration is something you do by force.

>> Unless you say that SMB/CIFS was inspired by it.
> 
> Totally different thing.
> 
> AppleTalk was a network transport protocol, like NetBEUI and IPX/SPX.
> 
> SMB is a file sharing protocol, like AppleShare and Netware Core
> Protocol and NFS.

SMB ran on top of NetBIOS Frames which Microsoft called NetBEUI.

NetBEUI originally or in its intended meaning was not a network 
transport protocol, but an application programmer's interface.

Microsoft abused that name, but anyway.

It's this NBF that I have referred to as "NetBIOS" because for all 
practical purposes, the API and the protocol came side by side and 
that's why Microsoft did that.

So when I mention "SMB" what I really mean is the NetBIOS 
auto-discovery.

I don't know why you keep making distinctions when in the end it comes 
down on the same thing.

That's like becoming mad because someone calls a lazy chair a couch even 
though it is not relevant to the use case (sitting down).


>> I am saying that we did fine without their mDNS-like features.
> 
> Pre-mDNS, if you had no name server, you had no name resolution.
> 
> Windows had a solution. Unix didn't.

They could use Samba too.

Which doesn't work for SSH access and the like.

But the target audience for Linux isn't exactly...

people who cannot configure dnsmasq, although there are plenty (and I 
was one too)

But I was one too because the community doesn't provide enough sensible 
defaults.

If more people do this thing, and tell each other, you discover how to 
do this sooner.

Does the recent auto-discovery feature in Windows rely on mDNS? Or 
LLMNR?

You know the "double click a device and open its web configuration web 
page".

That's the only thing ordinary people use it for.

And there are some places where you get shown a list of other computers 
in the network, which I wasn't too happy about (privacy).

> Post-mDNS, Unix had a solution too. A solid, reliable, standards-based 
> one.

Not used for Samba
Not used for NFS
Maybe used for SSH

My older devices don't support it so I would get a mix IF I chose to use 
it.

But I cannot address the older devices using the same domain.

So the entire system is inflexible and inextensible.

It is like Microsoft.

It is the stuff Linux has traditionally fought.

What do you expect people to do?

Use .local for new mDNS Linux boxes and for new hardware devices (how to 
guess?)

But use .home for older stuff? At the same time?

You don't see how this solution actually doesn't work?

It's the same as that GB -> GiB move. It only confuses the hell out of 
me.

I NEVER KNOW whether something is going to be REAL GIGABYTES or the new 
"fake" GB.

Particularly also because many apps by necessity use shorthand "G" form.

One app will have G == 1000. The other will have G == 1024.

It has created a complete mess that I could have told you in the 
beginning.

It has been totally forced.

It is not a natural solution.

You don't allow people to do what they want, you get trouble.

Coercion is the new Linux strategy.




The older system worked with everything.
Now I have to use two, incompatible systems side by side, OR choose to 
NOT use the new mDNS until I can use it *completely* which still means I 
have no choice in the matter at all.

I cannot decide anything about it. It does its thing and I have no say.

I can also not really move it to a new domain if I wanted that.

We are also "locked" to .local now.

How does that fit in with DNS' design?

You don't see how crappy much of a solution this is?

Everyone who uses local DNS is now in trouble. EVERYONE.

They can say they can live with it.

But if you have some other domain,


and you have hardware mDNS devices.

Your "search" will now point to your own domain.

You won't have unqualified resolution for your new hardware mDNS.

You now have to remember which devices are mDNS.

Or you have to constantly address them as .local,

but in that case you might as well dump your own domain.

You really didn't think this through, did you.


> It just means you can't use _one_ make-believe invalid TLD internally.
> This is a very small price to pay.

I am concerned with shaping the future.

This is not a small price.

1) Two now incompatible systems, one of which is non-extensible, 
non-negiotiable, non-transferrable.

2) The dynamics of the new system basically cause extinguishment of the 
older system

3) Not inherent in the technology, but only in the choice in how to do 
/etc/nsswitch.conf

4) Using the two systems side by side using different domains is an 
incoherent mess

5) The only way around it is to use 2 different search domains and never 
address anything qualified.

6) You are now eternally locked to using two different domains if you 
want to use your own DNS or not use mDNS AT ALL (except through GUIs) 
which means that soon you will suggest...

7) ...drop the personal domain, don't run a DNS server.

8) At that point it becomes clear.

9) That it's an extinguishing technology, and it will make it impossible 
for new people to start playing with DNS and learning the technology.

10) Zero configuration will not just be a possibility. It will be the 
only reality.

11) You won't be able to manually configure anything and have now 
created a generation of people that don't know how to do it.

12) http://www.coding2learn.org/blog/2013/07/29/kids-cant-use-computers/



It is perfectly clear that this new technology, or the "only one way" 
implementation thereof seeks to extinguish local DNS.

A number of people here have voiced opinions to the extent of "no more 
local domains".

You call paranoia.

I see what's happening.

There are several proponents of mDNS here who try to kill off local 
private domains.

This is also a facet of IPv6.

No more internal sheltering.

Everything 100% internet connected, only a transparent invisible 
firewall.

Internet-addressable private addresses.




This is the evolution.

Your home network is going to be killed off.

Where will you go then?


> Yes, it was fine, on Windows, on Netware, on AppleTalk.
> 
> But it didn't scale. As Ken said, once you got to hundreds of nodes,
> it got slow. With thousands, it got very very slow or failed.

But mDNS is also not meant for large scale is it.

> Yes, they are more work. In that sense, they are inferior. But the
> Internet could not have been built on anything else.

I'm not calling it inferior.

I am saying the manual work is a design that allows for automatic stuff 
_when you want it_.

If you start out with the automatic stuff, you have no choice to go back 
to static stuff.

>> You are not free to do what you like.
> 
> Oh, get over it!

Go to bed America, your government takes care of you.


> You can't call yourself the President of the USA or King
> Willem-Alexander of the Netherlands. You can't call your LAN
> xen.local.

I can as long as I do it in my own home.

What do you not understand about the difference between local or 
private, and global or public?

My own home should be MY domain.

You are eroding basic rights, privileges, and attainments.

You are eroding dominion over your own life.

Calling myself (in public) King Willem-Alexander of the Netherlands 
affects everyone. Calling my PRIVATE LAN .local effects only ME.

THIS IS FREEDOM because you don't have to compromise with others.

This is a gliding scale Liam.

If you can't even defend personal choice in the privacy of your own 
home, where will you defend it?

> This is not destroying your freedoms!

The Party would be proud that you have become such a good citizen.

>> I said "don't", plural. I meant that virtually no one runs a 
>> serverless
>> system.
> 
> That is utter nonsense.
> 
> Every home LAN is a serverless network. I have no home server. (Well,
> OK, I do, but [a] I'm a geek and [b] it's not even plugged in and
> hasn't been since I moved to Prague in August. And it doesn't have a
> DNS server installed, because I don't want to maintain one.)
> 
>> Unless you just want to connect phones, everyone has a router in their
>> homes.
> 
> That's not a server.

Yes it is, it runs DHCP.

> Some routers have services, but not all. It's not a defined part of
> the standard.
> 
> I know people who turn off their router when they want to work
> undisturbed. Your proposal means that then they could no longer print.
> _This is not an improvement_.

So you now have Linux people who have internet routers... no, network 
printers,

who can't afford or won't use a separate router for their LAN, apart 
from the modem of their ISP that can hack their home network if the ISP 
wants it,

also will not connect the printer by USB or give it a static IP + local 
name,

and yet still demand functionality under adverse conditions.

And who can't solve a printing problem by using an USB cable.

Or if they predict such happenings to configure the printer statically,
or just buy a cheap router so that DHCP works statically.

You know it's all a high amount of laziness and wanting everything.

You are dependent on bad technology and when it's gone you suffer.

Yeah, obvious.

Isn't that absurd in itself?

To base your life on that?


> NOTHING you have suggested is an improvement. You are suggesting
> over-complex kludges, which are totally unnecessary and will break
> under common circumstances.

But your solution you can only get working by (a) killing off another 
solution,

(b) using the Mac as the model for how to do business, while the Mac is 
the complete opposite of Linux,

(c) Make it impossible to actually use the .local domain on purpose 
(it's only meant for autoconfiguration and discovery) if you were to 
also use unicast DNS,

So you wanted a solution by all means possible, and ended up with this 
poor crap heap of a design that does what you want, but calling it 
beautiful goes pretty far?

It should be completely obvious that merging the two systems is the 
ultimate goal.

Not the exclusion of one over the other.

But you're not imaginative enough to find a solution for that.

This split solution is untenable.

It's a band-aid solution that will work "for now".

Because if you consciously use the .local domain for it, and you also 
have another domain, it will create a mess.

"It's not great, but it will have to do" <-- that's the solution we now 
have.



> What you're proposing is _worse_ than what we have.

All for checking a SOA record at system boot, or at network restart,

and only because it sends .local queries to the upstream DNS server that 
really ought to blackhole it?


> All because you've got your knickers in a twist because an Internet
> standard stops you calling your home network whatever you want.

The standard doesn't do that.

The standard allows for it.

Stop blaming or vincidating your solution on a standard that doesn't 
mandate it.

> It's _ridiculous_. You are being absurd.

It's absurd that you keep saying the standard forces you to do this, 
when it doesn't.

-- 
Highly Evolved Beings do not consider it “profitable” if they benefit at 
the expense of another.