In the beginning was the ARPANET,
When SRI and BBN handled the NIC and NOC for the Internet, there was a
field in the WHOIS database called "Connected Status." People could
request IP networks (Class A/B/C) from SRI, but BBN would only route
networks with the Connected Status on the ARPANET. You could do Whois
queries to check the status of any network address. Companies could
request IP addresses for internal use, without connected status.
Around the time the NSFNET took over, the database sources started
diverging. Connected Status was deleted from the WHOIS database. There
was only one WHOIS database. NSFNET kept a seperate database of which
networks were allowed to use the NSFNET.
I know about all the routing specificiation language efforts, but is it
possible to go back and do something simple?
For networks which are announced on the Internet, add a Connected field
to the regional address registries listing the AS Number(s) which could
announce the network. Private, internal only networks would have an
You know, that would be a great idea except for one thing.
It's just too simple
Well, this is exactly what route objects are for. Can't get too much
simpler than that.
The problem is that you have to protect the authoritative database(s)
in order to avoid bogus origin declarations. RIPE does that quite
nicely with their maintainer concept. A maintainer protecting and
IP address range (inetnum) does also protect the creation of route
objects for this IP range (but authority for routing information can
be delegated to some other entity by using mnt-routes attribute).
Without this kind of protection scheme, route origin documentation
becomes worthless. See RADB with the hundreds and thousands of
automatically generated "proxy route objects" which some people
seem to believe being a great idea.
Then again, I've pretty much given up hope that ISPs filter their
customers properly. Filtering customers means losing money to the