Re: DNS discovery: what about well-known site-local addresses ?

[John Schnizlein <jschnizl@cisco.com>]

At 08:31 AM 7/29/2003, BELOEIL Luc FTRD/DMI/CAE wrote:

>at last IETF meeting (Vienna), I have understood that Christian Huitema was proposing not to deprecate well-know site-local addresses used for recursive DNS servers. As a consequence, such a solution could also be taken into account in our discussion.
>
>I would like to have the opinion of this working group about that point.

Why continue this insistent attempt to invent new methods when there is
a well-defined method (stateless-DHCP) that has widespread support?

First reason why this new alternative is a bad idea:
The definition of well-known addresses for each Internet service is not
the way the Internet works. A very strong analysis would be expected to
embark on such a radical reinvention.

Second reason:
Site-local addresses are contentious, to say the least. A large fraction,
if not rough consensus (see Tony Hain's rebuttal), agrees that site-local
addresses cause more trouble than they are worth.

Third reason: 
The number of special site-local addresses would limit the number of
DNS servers that could be made available to clients.

Fourth reason:
For each special site-local address, the local network administration
would have to configure appropriate routing - a non-trivial effort.

Fifth reason:
This approach would not work for the case in which a subscriber uses
a DNS server provided by his service provider. The subscriber's site
and the provider's site are likely often different.

John

#----------------------------------------------------------------------
# To unsubscribe, send a message to <dnsop-request@cafax.se>.