Thanks, Bob, for writing a draft doc on DNS Discovery requirements.  We 
need something concrete to center this discussion.

So, here are some discussion points around the draft doc...


Is a server-less solution the best solution for DNS Discovery?
==============================================================
The requirements statement starts off with the same premise as the DNS 
Discovery team: "server-less" (as the term is defined in the doc) operation 
is a design requirement.  The DNS Discovery team report stated "It is a 
further requirement that the above information be obtained without using a 
DHCP server."

The issue with the DNS Design Team analysis and the recommended DNS 
discovery mechanism is not "is this the best server-less solution"; rather, 
the issue is "is this best server-less solution the best solution" or "do 
we need this best server-less solution".  Bob's requirements will give us 
another answer to "is this the best server-less solution" and won't answer 
the question of whether the server-less solution is needed.

So, we should either:

* throw out the requirement for a server-less solution and
   find the best solution; or,
* do an apriori analysis to confirm that a server-less
   solution is fundamentally required; or,
* have and end-game analysis of "best server-less" solution
   vs. server-ful solution and decide if we need both

Do we need a server-less solution for every IPv6
configuration problem?
================================================
I read an unstated assumption in the first sentence, "IPv6 provides two 
approaches to basic IPv6 configuration." that server-less and server-ful 
configuration are two mutually exclusive, stovepipe approaches, where both 
alternatives must be provided for each component of basic configuration.  I 
believe we've shown that server-less configuration for IPv6 addresses is 
useful and a solved problem.  It may be the only required solution - I can 
suppose situations in which address assignment might require constraint and 
control, but we don't have any deployment experience that proves such a 
requirement.

But a solution to server-less configuration of IPv6 addresses does not 
imply that a solution to server-less configuration of routing information, 
DNS resolution and other basic configuration is a good idea or even 
possible.  I think we should be careful to consider each type of 
configuration separately to make sure we're generating useable solutions to 
real problems.

Server-less DNS Discovery or server-less DNS resolution?
========================================================
Just to be careful - the basic requirement in Bob's draft requirements is 
"to provide a server-less mechanism to enable a host to learn the address 
of an DNS server."  This is a more stringent requirement than to provide a 
server-less mechanism to perform DNS resolution.  I would say that the most 
recent "Stateless DNS Discovery" mechanism 
<draft-ietf-ipv6-dns-discovery-04.txt> does *not* meet the requirement laid 
out in Bob's requirements statement.  A host using the Level 1 compliance 
mechanism defined in <draft-ietf-ipv6-dns-discovery-04.txt> does not find 
the address of an arbitrary DNS server.  Rather, the host can do DNS 
resolution in a site that has specifically configured its network routing 
and DNS infrastructure to support DNS resolution through the well-known DNS 
server addresses.

Scope of required information for DNS Discovery
================================================
Others have pointed out the a device using DNSSEC may need NTP.  The 
requirements should be expanded to read something like "learn the address 
of a DNS server and any other servers required for use of DNS".

At 04:23 PM 4/17/2002 -0700, you wrote:
>I took a cut at a requirements statement for IPv6 DNS Discovery.  Comments 
>are appreciated.
>
>Thanks,
>Bob
>
>----------------------
>
>IPv6 DNS Discovery Requirements Statement
>
>IPv6 provides two approaches to basic IPv6 configuration.  One is 
>server-less and is defined in IPv6 Neighbor Discover [RFC2461] and the 
>IPv6 Stateless Address Autoconfiguration [RFC2462].  The other is server 
>oriented and is defined in DHCPv6 [dhcpv6 id].  IPv6 Neighbor Discovery 
>includes flags to direct an IPv6 host to use either approach.
>
>In order for an IPv6 host to communicate on the Internet it needs an IPv6 
>address, IPv6 prefix for the link it is attached, default router address, 
>and a DNS server address.  This information allows the host to use basic 
>internet services like the web, email, file transfer, etc.
>
>The server-less approach currently provides mechanisms for the IPv6 host 
>to learn an address, IPv6 prefix for the link, and default router address, 
>but does not provide any mechanism for the IPv6 host to learn any DNS 
>information.  Without any DNS information the IPv6 host can not use basic 
>internet services with out some other kind of configuration (except by the 
>user typing in literal IPv6 addresses).  Requiring users to enter literal 
>IPv6 addresses in difficult in IPv6 given the size of the addresses.
>
>The basic requirement for IPv6 DNS Discovery is to provide a server-less 
>mechanism to enable a host to learn the address of an DNS server.  This 
>will complete the IPv6 server-less configuration mechanisms.
>
>For this requirement, Server-less is defined to mean the DNS information 
>is learned with out any dependence on resources other than are needed 
>(i.e., links, interfaces, routers, and DNS server) to communicate with the 
>DNS server.
>
>IPv6 DNS Discovery should work inside of a single site where the DNS 
>servers are in the site and between sites where the DNS servers and hosts 
>are located in different sites (e.g., small home office where DNS servers 
>are in the ISP's network).
>
>It is also desirable to learn other related DNS information such as 
>default DNS for the IPv6 host, search path, LLMNR enabled flag, etc. in a 
>server-less manner.
>
>Desirable aspects of a solution the meets these requirements include:
>
>   - Minimal changes to existing implementations
>   - A solution that could be standardized in a short amount of time.
>   - Minimal use of bandwidth
>
>
>References
>
>[DHCPv6]  J. Bound, et. al., Dynamic Host Configuration Protocol for IPv6
>           (DHCPv6), Internet Draft, <draft-ietf-dhc-dhcpv6-23.txt>,
>           February 2002.
>
>[RFC2461] T. Narten, et. al., Neighbor Discovery for IP Version 6 (IPv6),
>           RFC2461, December 1998.
>
>[RFC2462] S. Thomson, et. al., IPv6 Stateless Address Autoconfiguration,
>           RFC2462, December 1998.
>
>
>
>
>
>--------------------------------------------------------------------
>IETF IPng Working Group Mailing List
>IPng Home Page:                      http://playground.sun.com/ipng
>FTP archive:                      ftp://playground.sun.com/pub/ipng
>Direct all administrative requests to [EMAIL PROTECTED]
>--------------------------------------------------------------------

--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page:                      http://playground.sun.com/ipng
FTP archive:                      ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------

Reply via email to