On 21-sep-2007, at 13:11, Hannes Tschofenig wrote:
I haven't received concerns regarding the suggested discovery
approach. May I assume from the lack of feedback that the suggested
approach is reasonable?
Not so fast...
I've given the draft a quick read and I find myself with some questions:
- what would be the resolution of the geography information?
- how sensitive is this information from a privacy standpoint?
The answers to these questions will determine what approaches are
workable.
As for the discovery mechanisms:
STUN -> reverse DNS -> location server seems convoluted to me. If you
look at DNS service discovery (also known as wide area bonjour in
some circles) you'll see something that's a lot easier: forward DNS -
> server. The advantage here is that DHCP supplies a domain name
that can be used for the forward lookup, and home gateways tend to
push the domain name they learn from an upstream server out to local
hosts so this would work fairly well in an IPv4/NAT setup. And it can
piggy back on the DNS service discovery that already happens if
applicable.
But there are two issues: IPv6, which is typically deployed without
DHCP and also often without a working reverse DNS. Apart from that,
there is no built-in limitation on who gets to discover the server
address. This can be fixed by letting the location server handle
privacy limitations.
Although anycast was widely shut down for DNS resolver discovery, I
think it fits really well here, because it works both for IPv4 and
IPv6 and doesn't require additional infrastructure such as the DNS or
DHCP. Also, the anycast mechanism automatically limits who gets to
talk to the server.
Last but not least, the draft seems to assume a centralized model
where there is one authoritative server and clients need to talk to
that server. I'm not entirely comfortable with that. In general, it's
always preferable to be in charge of your own destiny as a user. A
more practical problem could be that service providers don't set up
servers. We've seen that there are groups who will spend time and
money to discover information like this independently. It would be
great if that information could be used if there is no other choice
and/or the user prefers this.
It would also be great if home gateways could (re)transmit location
information, which is either learned from an upstream server or
entered by a local administrator. But in these cases it's important
to consider the possibility of having wrong or conflicting
information. (I.e., user moves and forgets to change location setting
in a wifi base station.)
_______________________________________________
Int-area mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/int-area
