Recursion

[Barry Margolin]

In article <94sg8o$9e at pub3.rc.vix.com>,
Johnny Fribert Lauridsen  <jlaurids at cisco.com> wrote:
>So, if I set my Client DNS Server to non-recursive, and make sure that I
>have TTL=0 (to avoid caching)
>on the actual primary/Auth DNS server (delegated from client DNS server
>with NS and A),  that the client DNS server 
>'will do a 'referral', as explained above?  I guess this will work OK,
>according to Douglas Comer and others, BUT,
>AND, most important - Will the Win2K, NT and Win9X DNS resolver libraries
>on the end-client understand and
>use the referral and do the lookup at the referred DNS Server?

No.  Almost no resolver libraries do iterative queries, they mostly send
recursive queries to the nameservers listed in their configuration, and
depend on the nameserver to iterate for them.  Resolver libraries like this
are called "stub resolvers", and it's the usual implementation.

One reason for this is efficiency.  With an iterative resolver, each
process would probably have its own cache of recent responses, and would
have to work its way down from the root servers to the servers for the
domains it's accessing.  While it would be possible to design a resolver
library that uses a system-wide shared cache, it's difficult (Unix didn't
have shared memory when much of this stuff was first implemented), whereas
moving the functionality to a server makes it easy.

-- 
Barry Margolin, barmar at genuity.net
Genuity, Burlington, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.