OK, I'm a little new at this but I think this previous e-mail conversation is appropriate to this thread. If you have more than one DNS in your tcp/ip config shouldn't the OS ask each of them to resolve until you run out of DNS's regardless of whether the first one is there or not. I haven't tried this under beta 2 yet. I didn't know enough to challenge MS's answer before but I'm learning thanks to all of you.
---------- Forwarded message ----------
Date: Mon, 13 May 1996 19:37:46 -0700
From: Paul Donnelly <pauldon @
To: "'heagarty @
com'" <heagarty @
Subject: DNS Server Search Order.
I was just reviewing the following bug you reported and I'd thought I'd
clarify the resolution, since our behavior appears correct (really) at
this time. The list of DNS's listed are searched only if the primary
can't be reached. If the name is not resolved by the first one, then the
name resolution will fail. Only in the case of a DNS being unreachable
would we attempt to check another one. I just verified that we are
doing the same thing as my Unix box, and we are.
Hopefully that clarifies the DNS search list a bit.
WIndows NT Test Team
-- Original bug report to MS --
I apologize if I have not followed the proper channels while providing
this bug report.
I have noticed that NT 4.0 does not seem to use secondary DNS's setup in
its TCP/IP. For example, I have a local DNS which handles the machines
on my LAN and I have a global DNS that handles full internet name
resolution. If I have the local DNS first in 4.0's list I can see all
the local machines but can't get to anything in the outside world.
If I reverse the order of the list by putting the global DNS first and
the local one second then I can see the internet hosts just fine but not
the PCs on my LAN.
I have the same order in my Windows '95 that is installed on a different
partition and it works fine for both groups of hosts regardless of the
order of DNS's in the tcp/ip configuration.
Shouldn't the stack continue searching DNS's until it finds the host
name or runs out of places to look. I'm a little new at this so I'm not
willing to completely blame 4.0 (beta or not).
Thanks for all your hard work.
* End of bug report -
-- beginning of thread that made me think of this --
From: Jeremy Noetzelman[SMTP:jnoetzel @
Sent: Tuesday, June 11, 1996 3:44 PM
To: axel .
se; firewalls @
Subject: RE: Round-robin DNS?
At 12:05 PM 6/11/96 +0200, axel .
>Also, the Microsoft says that zone transfer is not properly implemented
>among the Unix BIND versions which caused some trouble. Due to this certain
>DNS/BIND versions on some implementations will be incompatible. I wonder if
>this is true? If so, the obvious conclusion should be that one has to avoid
>intermixing DNS/BIND from different vendors and/or platforms (indeed, a good
>principle regardless product, anyhow).
I wonder whether this is yet another case of MS insisting that they're the
standard, regardless of what's in practice. Now they want to rewrite the DNS
rules and say that everything that's been in production on a Unix platform
is a bogus implementation...
Tends to reinforce the anti-ms sentiments around, IMO.