“As long as we can choose our networks…”
I suppose that should be refined to “as long as we can choose what routes our traffic
takes…” — that is, which peers, what types of lines and routers, perhaps even what
last-mile providers. It should be possible to say “if
there’s no way to send the following content along routes I trust,
don’t send it.”
You don’t have to be paranoid to want this. You might distrust a
route because you expect it to attempt to reconstruct, alter, and
resend content; because you suspect it of not accepting content from
certain areas or sites, because you worry that it keeps track of what
you send when, without your permission… You might not want to
send content through any router that doesn’t respect the “return
receipt” flag which sends back information on how your packets
travelled on their way to a destination. Or you might just not want to
support in any way certain traffic providers, explicitly asking to
patronize other providers whenever possible.
house to the Arlington Center hub, ‘Hub of the World’ or ‘OpenBelNet’
from there to the Harvard U. hub, ‘HU Internet2’ to the Cambridge
Internet2 backbone hub, and any lines/routers run by WorldCom or
UUNet, or on the dynamic ‘Debian-Class1’ network list.”
- Latency : as long as total latency is under 3 seconds, stick to the above networks rather than leave them.
-
I use Return-receipt packet delivery, which sends me back a packet for
every packet I send out, announcing the route it’s taken, or that it’s
been dropped. This more than doubles my bandwidth bill, but has
its advantages… feed the resulting datastream into a route-analyzer;
and one can set up all sorts of useful triggers. When my
preferred routes are all down, I can opt to use secondary networks –
either outside my normal prefs, or at a higher rate. I can also
tell my Net-enabled applications not to complete sensitive transactions
if there have been any fishy routings in the past minute. (Besides,
whenever my provider makes too many routing mistakes, I get a free
month of service.)”
Implementing this would seem to take significantly more intelligent routers and middleware than currently exists.
For a great coverage of some of the topics brought up at the Web of Ideas discussion tonight, see Geoff Huston‘s killer essay on the finance of networks, with its diversity of options laid out in gory detail.