About the address allocation convention between ISPs

Hello everyone,

I have a question about the convention of address allocation between ISPs.
If a smaller ISP tries to establish connection with its provider, does
this small ISP configure one of the interface on its boarder router using
an IP address obtained from the provider, or it is the other way around,
that is, the provider uses one of the IP address belongs to the customer
to configure the provider's boarder router?

I have this question because I am trying to identify the link between two
organizations from traceroute measurements. How the addresses are
allocated will affect the identification of the inter-domain link by
exactly one hop.

I am not sure if there is such a convention at all, or the address
assignment is randomly decided according to the agreement between the
customer and the provider?

Since I know there are many seasoned network professionals on this mailing
list, I think it might be a proper question to ask here. Would anyone
kindly be willing to share your experience? Thank you very much!



Many larger networks (with multiple interconnections
will split the chore, where the numbering reflects
exactly who is responsible for the physical circuit.

So alas, there is no one "right" answer to your
question, unless you're going to try to put together a
table based on the naming conventions...

for instance, probably hop "6" is the actual interface
between 7018 and 209 in NY according to this view.

  5 gar4-p300.n54ny.ip.att.net ( [AS 7018]
4 msec 4 msec 4 msec
  6 att-gw.ny.qwest.net ( [AS 7018] 4
msec 4 msec 4 msec
  7 jfk-core-03.inet.qwest.net ( [AS
209] 4 msec 4 msec 4 msec

-David Barak

Dear Mr. Barak,

Right, in your example, AS7018 and AS209 are both backbone ISPs and I
guess they are peering at NY. So if the addresses at both ends of
the "real" inter-domain link were assigned with qwest's address, hop 6
would be the inter-domain link. But if the addresses were assigned with
AT&T's address, hop 5 would be the inter-domain link. DNS name
seems would not help in this case.

And here is another question that confuses me. Using your example. If
the link was as follows,

--+ +-----------+
  >IP1(domainnameA) <------> IP2(domainnameB)| border rt |IP3(domainnameC)
--+ +-----------+

If the router on the left(only half was drawn) physically belongs to att,
and the router on the right physically belongs to qwest. the Link between
IP1 and IP2 is the inter-domain link. IP1 and IP2 should be a pair of /30
addresses. My question is how are the domain names of IP1 and IP2
assigned? If say IP1 and IP2 are both addresses from AT&T's address
block, the for IP2, is it usually foo.att.com or foo.qwest.com?

Thank you very much!




You can also use whois to determin who owns the /30 between the hop in
question. If ISPx comes up as the holder of the netblock in the
registration then that traceroute result is comming from ISPx. Hopefully
this helps!

- Robert

Dear Mr. Fei,

I surmise that hop 6 is the actual interface by seeing
gw-att.qwest.net <-- the fact that both qwest and att
appear in the same DNS lookup to me implies that qwest
intends that link to be the border. hop 5 is the
ingress link to ATT's border router, and hop 7 is the
next qwest router upstream.

Generally the ISP who numbers the block provides the
DNS. This breaks down, however, because there is no
single naming structure, and many providers don't do
reverse DNS for their infrastructure.