Hi Bill,

I think the carriers are just trying to make themselves useful and avoid this 'dumb pipe' syndrome. I'm not against the idea of carriers making known to consumers about the work these DPI equipment but from my understanding it seems DPI rae helping carriers to be able to provide appropiate services to targeted consumers. Google has been using the same model to monetize customer information and no one sems to be bothered about that.

Privacy, piracy and deep packet inspection

HI Bill,

We have looked at the Max Planck Institute's Glasnost software. From the README:

"Glasnost is a tool to detect manipulation of BitTorrent traffic by ISPs,
in particular blocking of BitTorrent traffic using forged TCP RST packets.
This type of blocking was found to be used by the traffic management
boxes of Sandvine Inc., deployed, e.g., by Comcast in the USA. "

As you can read in our recent P2P filter solution test (see http://www.internetevolution.com/document.asp?doc_id=148803), sending TCP Reset (RST) packets is only one method to filter P2P traffic - actually, as we have seen in the Sandvine/Comcast experiment, it is not the best alternative :-)

Unless there is some hidden magic in the Glasnost source code not mentioned in the documentation, it is unlikely that the MPI tool recognizes the much more common - and more advanced - ways of limiting P2P traffic by slowing down TCP connections. Advanced filter systems like the ones we tested (Ellacoya, Ipoque) do not simply block sessions but slow them down. If you block a session, a BitTorrent client would simply try to open additional ones, leading to up to 2,000 simultaneous session attempts per user. This would blast P2P filter device tables and eventually service provider edge router tables. In fact, large-scale experiments in the U.S. (not calling any ISP names) have likely experienced this - we could have told them before :-)

In addition, the simple RST filter trick can be blocked easily under Linux by using the netfilter Linux kernel support (see http://en.wikipedia.org/wiki/Iptables), in something like:

"iptables -A INPUT -i $IF -p tcp --dport $BTPORT -m tcp --tcp-flags RST RST -j DROP"

(where IF stands for the WAN interface and BTPORT stands for a user-selected P2P application destination port)

In short, if the Glasnost tool says your network does limit P2P traffic, it is probably right.  If it claims the network does not limit P2P traffic, you can assume it just did not detect a more advanced P2P filter.

We will set up a Glasnost test in our P2P emulation test bed soon and will keep this forum posted about the results.

Best regards,

Carsten Rossenhoevel
EANTC (European Advanced Networking Test Center)

Bill,

This article (and some of its comments except cjon316) unfortunately mixes up three very different things:

1. Privacy
2. Net neutrality
   
3. ISP traffic management

Privacy is not at all an issue here.  The protocol-based P2P filter technology (like the infamous Comcast/Sandvine experiment) does not collect user-related data. Its goal is to manage large volumes of traffic, not to collect data about user behavior.  There are certainly some systems - U.S. government agencies have driven their development - which collect and analyze user-specific data (a.k.a. "spying") but that is a completely different type of application.  ISPs are not interested in this activity and could face serious charges in most democratic countries if they did.

Net neutrality is a specific U.S. issue that may be affected by P2P protocol filters.  Let's picture this using the analogy of a narrow road.  Like in broadband access, 99% of all car owners would use the road, say, twice a day. 1% of the car owners however use the road 24x7, driving back and forth all the time.  This way they reach around 200 times the average use, in total 400 rides a day (these are realistic figures).  Now the road gets blocked and nobody wants to pay for an upgrade (flat rate prices cannot be increased due to competition).

If one takes net neutrality literally, each n-th car passing the road would need to be rejected randomly. Now the few car owners who run their cars continuously along this street would still get through 200 times (50% of 400), but the occasional users would only get a single ride (50% of 2 times).  Given that all users pay the same (the 24x7 car owners refuse to pay more because of net neutrality), the occasional users would turn away in the end because they are obviously treated unfair.

Now what alternative do we want? I can see four:

1. Create toll roads a.k.a. volume-based Internet plans. Nobody wants to go there as it would discourage Internet usage. We have seen the disadvantages of volume-based charges in the past.
   
   
2. Increase the flat rate prices for everybody to co-finance the 1% power users requirements. This is unfortunately not possible in a competitive market, unless the government regulates all ISPs at the same time. Net neutrality could have such an effect in the U.S. but this would not be beneficial for the government goal to make the Internet available to all citizens.  I don't know if the government and legislation have got to this conclusion already...
   
   
3. Keep prices stable and the term "flat rates" in place but limit the flat rate to a maximum volume per time per user. This is what ISPs do in many places, either openly (like in the U.K.) or secretly.  I agree that limiting a flat rate secretly is not a fair practice and will likely be turned down by courts some time in the future.  Also, unfortunately network congestion management is not easy to implement using such a per-user penalty scheme (because there might still be bursts of traffic at specific times when the penalty counters have been reset).
   
   
4. Modify the "flat rate" term, creating a standard "pseudo flat rate" with limits for specific types of "bulk" traffic and a premium "real flat rate" which works 24x7 and would be billed like a business service (some $200-500?/month). Prioritization has always been implemented on the IP layer (business customer packets have often higher priority on an ISP backbone than residential customers) and nobody has complained. Now it is technically possible to differentiate bulk and priority residential applications - limiting only the bulk applications. Suddenly people complain quoting net neutrality, however what ISPs do is mostly traffic management to keep their networks operating.

For further technical information about P2P filtering, please see our recent test published at Internet Evolution: http://www.internetevolution.com/document.asp?doc_id=148803

Best regards,

Carsten Rossenhoevel

EANTC (European Advanced Networking Test Center) 

I can see an ISP wanting to facilitate efficient communications by managing traffic. That to me follows the same logic as timing traffic lights to accomodate vehicle flow through a city. But when they start taking pictures of my license plate and passengers to reconstruct the digital billboards based upon my vehicle registration, that gets a bit more dicey.

If an ISP wants to do this, they should simply tell their customers that they are doing so, or wanting to. This is not as objectionable if the customer knows it will happen than if it is done by surreptitious means.

Privacy must be guarded. I don't know if laws need to be put in place, but perhaps that is a necessary method to guard ones privacy. Are the providers violating any current laws by doing this?

Where are there examples of true internet transparency? 

Privacy is an important right that must not be allowed to be abused for any reason. So it is important that consumers are made aware of what is being done with the information they access over and ISP's network. That way the consumer will have a choice (hopefully) on which service provider to choose.

If DPI is important to managing traffic so as to facilitate efficient communications. laws must be put in place so that peoples privacy are not compromised.