some of this I've said elsewhere, but not here. sorry if you've already
seen it.
IMHO this is fundamentally a very dubious option because DNS is the
authoritative source of name-to-address mappings, and the way to find
out what DNS name is assigned to a particular network address is to
query the DNS for PTR records at the appropriate in-addr.arpa or
ip6.arpa locations.
It's bad design to have two authoritative sources of the same piece of
information. If there's a conflict between DNS and DHCP, which one
wins? And given that DNS server information can be obtained from DHCP,
why not just query the DNS to find out the FQDN that corresponds to the
network address assigned to the host? Or is DHCP just acting as a kind
of DNS cache in order to make life simpler for relatively
unsophisticated networked appliances? If so then the rules for caching
need to be observed.
Note also that in the IPv6 world it is considered perfectly normal for a
host to have multiple IPv6 addresses and IPv6-aware applications are
expected to deal with this somehow (currently, by a combination of luck
and trial-and-error). There's no reason to assume, and good reason to
not assume, that all of the addresses associated with a host are on the
same network. For this reason it makes even less sense in IPv6 than it
did in IPv4 to assume that "the" network to which a host is attached
configures the host. So any information obtained from DHCPv6 should be
considered as only meaningful in the context of that particular network
attachment - not as something that applies to the entire host.
Even if this option can be found to have value (with suitably narrow
applicability), IMHO it's nuts to define the "domain suffix" (components
2..n of the FQDN) separately than the "host name" (the leftmost
component of the FQDN). Any attempt by a host to infer anything about
the relationships between hosts based on their domain name suffixes is
extremely dubious anyway. Why not just define a FQDN option and be done
with it? Seems like it would be less likely to be misused.
Also, one should be careful to avoid making the assumption that a host
has a single FQDN, or even a distinguished FQDN, or that an FQDN maps to
at most one host. None of these is true in practice.
-------- Original Message --------
FWIW, "domain suffix" is used in RFC 3263, 3588, 4183 and 4620. In none
of these documents does it seem that the author has seen a requirement
for a definition; "a domain name that is intended to be used as a suffix
of a complete domain name" seems to be the implied definition.
_______________________________________________
Ietf@xxxxxxxx
https://www1.ietf.org/mailman/listinfo/ietf