NNSquad - Network Neutrality Squad

NNSquad Home Page

NNSquad Mailing List Information


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[ NNSquad ] Re: Comments on NNSquad Purpose

Phil Karn wrote:

[ This message is passed to the list since it contains discussion
of technical parameters useful for measurement toolset R&D.
Any discussion that strays from the technical (and you know
what I mean) in response to this message will not be approved through to the list. -- Lauren Weinstein
NNSquad Moderator ]

Jay Sulzberger wrote:

our packet streams.  But the fact is that if we give two numbers,
we have specified how good a Net connection is.



For if you do, I will produce evidence gathered by the
hardworking folk of NNSquad, who use this small suite of Net
tests, which give me back two numbers: bandwidth and lag.

I can think of three independent figures of merit for network neutrality:

1. Latency. I want to know how network latency changes with independent variables such as packet size, packet rate, source, destination, transport protocol, application protocol, user data, etc. Several other metrics can be expressed entirely in terms of latency: "Packet loss rate" is just the fraction of packets with infinite latency. "Bandwidth limit" is the packet size * rate product above which latency rises to keep the delivered size * rate product constant.

2. Transparency. Does the carrier deliver my packets to their specified destinations as I sent them, or are they intercepted and modified in some way? Transparent web proxying would be an example of non-transparent behavior. NAT would be another, although most NATs are under the customer's control and the ISP is to blame only for not making enough routeable IP addresses available.

3. Packet spoofing. Does the network inject packets that appear to be, but are not from the party with whom I am speaking? (apologies to Lily Tomlin) Comcast's TCP reset injection would be an example here.

What do people think of this list? Should anything be added? Changed? Redefined?


I think this is a very good start. Perhaps we'd want to alter the
wording on #2 so that it is not uni-directional. I.e., the delivery of
packets is not just "as you sent them" because the packets could be
coming towards you. Port 25 blocking (for better or for worse) is a good
example. Does the ISP block direct outbound connections from your
machine to port 25 on machines on external networks, thereby allowing
you to only use their email server? Similarly, does the ISP block
inbound port 25 connections to your machine, preventing you from running
an SMTP email server?

Again, this is not a discussion of port 25 blocking; I can see reasons
for doing it (and I myself hate it, as I run an email server at home). I
think this is just the sort of metric that NNsquad should factor into an
overall ISP "report card rating."