North American Network Operators Group

Date Prev | Date Next | Date Index | Thread Index | Author Index | Historical

RE: ISPs as content-police or method-police

  • From: Hank Nussbacher
  • Date: Tue Nov 21 00:50:49 2000

At 09:42 20/11/00 -0800, Roeland Meyer wrote:

No, we cannot all agree to that. I cannot condone anyone else applying their
ideology, by force or default, on someone else, unless specifically
requested. As a downstream customer, the backbone is nearly unreachable from
a services perspective. If CERFNET started to do port-filtering then the
only means I (I assume that everyone here has sufficient clue to find my
upstream) have to change that is to either sue CERFNET or start looking for
an ISP that has a different backbone provider. CERFNET tech support will not
even talk to second or third level customers. In essence, it eliminates the
secondary re-seller market from contention. You risk getting sued, not only
from the downstream customer, but your own downstream as well.

The ONLY one that should be even dreaming about doing something like this is
the direct upstream to the leaf nodes, and then ONLY with permission.
Otherwise, no ports should ever be filtered by any transit provider.
Do you block traffic to broadcast addresses (inbound or outbound) - .255? Do you rate limit (CAR) ICMP or SYNs at your border routers? If so, you too are doing content filtering. Part of your AUP should be a full disclosure of what you block or rate limit and then let the customer ask for it to be opened.

Your definition of content filtering is just one of many. Any blocking of *any* bit of IP data flow is some form of content filtering.

In the end it comes down to a subjective analysis of what each ISP considers harmful traffic that can be used to disrupt their network. This differs from true content filtering (sex, hate, terrorism, etc.) and should therefore be renamed 'harmful network data' filtering.


By God, we PAY for open pipes and there are standard remedies when we don't
get what we pay for.