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: Internet User's Bill of Rights [was: Google Hijacked -- Major ISP to Intercept and Modify Web Pages

Sean was (as far as I understand) simply talking about IP packet reordering. A (modern) network will deliver all packets in a flow in the order that they were sent, but it is perfectly permissible for packets to arrive out of order. This can happen for a bunch of reasons -- from per-packet load-balancing (which is universally recognized as a bad idea these days) to path changes (if packet X is sent over a path that has 20ms latency and packet X+1 is sent over a path that has 10ms latency (eg: an MPLS LSP changes or a circuit goes down, etc), X +1 will arrive before X). Yes, TCP will deal with this, but the fact remains that IP is a best-effort service and FIFO queuing is not guaranteed, or in many networks even preferred.

Your ISP CAN and WILL drop packets during times of congestion -- the only reason that your ISP can over service anywhere near the price that they are is because of statistical multiplexing and oversubscription.

Also, most of the "HTTP Hijacking" is performed with DNS tricks or transparent cache redirection. It is much simpler and cheaper to push things through a proxy to insert stuff into the HTTP than to perform some sort of TCP session MITM (which requires specialized things in the data path, etc..).

W


On Dec 12, 2007, at 3:04 PM, Dan Doyle wrote:

Hmmm, I think you are playing coy here. If we are talking TCP, it
already handles ordering. So, my interpretation of no reordering means
do not mess up the sequence that is embedded in the TCP header such
that my machine will not be able to piece the packets back together.
This also means to me that ISPs do not modify the sequence number so
that ISPs can insert their own packets with the correct sequencing
such that the client machine can't identify that the original host
stream has been compromised. I can't believe that ISPs have the
audacity to not see what they are doing as compromising the integrity
of the original stream in a world where  authenticity is critical.

By the way, I can tell you that my company has a problem with network
neutrality. Where did you get your information that you can make that
broad generalization?

Dan

On Dec 12, 2007 11:27 AM, Sean Donelan <sean@donelan.com> wrote:
On Mon, 10 Dec 2007, John Bartas wrote: 
"To me it's even simpler: What am I paying my ISP to do?"

     I'd like to propose an answer, in two parts:

1) I'm paying my ISP to move a stream of packets between my site and other
internet sites. They are not to edit, drop, reorder, examine, modify or add
to the packets other than the minimal RFC-compliant header changes required
for forwarding the packets.
2) They must handle all packets in FIFO order except by mutual pre- argeement.
For example, If I consent to delaying my bit-torrent traffic to improve
timing of my RTP streams, then they may do so; but not without my prior
informed consent.

You may have confused a packet switching network and a TDM private line.

TDM private lines have strict FIFO bit handling, they are also much more
expensive. Packet switching networks are cheaper, but also don't make
the same promises about packet ordering, variable bandwidth, packet loss,
or even data corruption, etc.

If you want to pay a lot more, would all these issues disappear? On the
other hand, is there a reason why they are lower cost? There doesn't
seem to be much concern about "network neutrality" for expensive network
connections usually bought by commercial companies.

You might be interested in the paper "Characterizing Residential Broadband
Networks":
http://www.imconf.net/imc-2007/papers/imc137.pdf