[dhcwg] Re: Last Call: DHCP Domain Search Option to Proposed Standard

Keith Moore <moore@cs.utk.edu> writes:

(in a message sent to the iesg)

> The problem I have with this option is that it invites implementors
> and network administrators to presume that a host's domain should be 
> tied to a host's location in the network - when these are intentionally, 
> and by design, entirely separate.  For instance, a mobile computer user 
> who wishes DNS lookups to be performed relative to a particular domain 
> should not have this choice overridden by each POP to which he connects.

> (The fact that protocol designers often try to use DNS lookups as a 
> means to discover topologically close services does not change this.
> There is an inherent conflict between using DNS as a way to discover
> services relative to a domain and using it as a way to discover
> services relative to a network location, but the intent of DNS 
> clearly favors the former.)

> Host DHCP implementations therefore should not blindly accept Domain 
> Search information from a DHCP server.  Local configuration of any 
> DNS information (including not just search domains, but also DNS servers) 
> should override anything supplied by DHCP.

> Keith

I am sympathetic with your general point. I'll also note that this
arguably applies to a number of other existing DHCP options. I don't
believe right off that this point is explicitely addressed in the
other cases either.

There is a general issue of when should (and shouldn't) a DHCP-learned
option override a manually configured one. My guess is that
DHCP-learned options generally do not override manually configured
ones. But I suspect that this is all considered implementation
specific. Can other DHC WG members comment? What do existing RFCs say
about this point?

Finally, what do you propose? I.e., "send text" :-) Would it suffice
to add some language stating that in the case of this option, it
should not override a manually configured search list?

Thomas

_______________________________________________
dhcwg mailing list
dhcwg@ietf.org
http://www1.ietf.org/mailman/listinfo/dhcwg