Tony Hain <alh-ietf(_at_)tndh(_dot_)net> said:
Mobility is not the only reason to use DDNS. Consider the case of Dan's
residential gateway. If it provided a consumer-friendly automated DDNS
server for a sub-domain delegated to the residence, what are the hard
issues? First would be security, but that is reasonably addressed by
making the dynamic registrations only possible by devices on the lan
side, and by a simple web-based cert mechanism between that device and
the ISP DNS infrastructure. This aligns the DDNS trust boundary with the
basic service boundary. Second would be getting past the brain-dead
perspective that consumer connections to the Internet should not be
hosting services. The entire set of peer-to-peer applications is based
on the fundamental assumption that a service endpoint can exist anywhere
and be found through simple resolution of a name. What are the reasons
to do it? First the consumer would have simple consistent access to name
resolution for all devices on the home network. Second, they would be
able to expose services (peer-to-peer games, appliance diagnostics) that
fit directly into the naming framework they are already accustomed to
for other Internet services. Third, it scales much more realistically as
the infrastructure side only has to support updates based on the
attachment frequency of the consumer network, not every device as they
power up, or move between subnets. This would also allow for very short
TTLs where they make sense without requiring them to be everywhere.
Dan says:
Well, this makes me feel better and there is certainly a lot of good
thinking in the above. I wonder, though since I know almost nothing about
IPNG whether maybe its handled there better.
It seems to me for troubleshooting, its awefully handy to think of the DNS
as more or less static. If the connection that used to be somebody's WWW
pointing to there childrens playground is instead the sex-with-goats hotline
for 20 minutes, its harder to troubleshoot if everything is dynamic.
I'm arguing both sides clearly becuase it a subtle tradeoff. The scalability
thing is a good point.
In my implementation, every house it going to have a WWW server, some with
fixed Ip's some just pointed to by a corperate resource, some an
intentionally obscure port and (maybe dynamic) DHCP assigned IP, etc.
I think TOny is perceiving the DNS process as just another service, not a
framework per se.
But with the name resolution Internet board, etc, it has a quasi-legal
status already.
I guess among other things I don't quite get is why if an ISP buys an IP for
$0.35 they rerent if for ten times that, per month.
I'm rambling. Its a fun topic though.
Regs to all
Dan