The following text is copyright 2009 by Network World, permission is hearby given for reproduction, as long as attribution is given and this notice is included.

 

Cox Communication attracting attention

 

By Scott Bradner

 

On the surface it seems like almost exactly the wrong time for the announcement that Cox Communications made on Jan 27th.  Cox announced that they were about to start unequal mucking with the Internet traffic of their residential customers.  Maybe the Cox folks know something I donŐt but it sure seems to me that Cox has just figured out how to be the main target for a new administration and a new FCC chair that have made it clear that they might be looking for just such a target.

 

At least this time Cox did let its customers about their plans, unlike what Cox and other cable companies have done in the past.  Cox published a FAQ about their plans (http://www.cox.com/policy/congestionmanagement/) that does provide some information, but not nearly enough to be able to tell just how it works or what will happen to customer's traffic.  The FAQ says that "applications that are tolerant of delay ... may be momentarily slowed" during times of congestion.  The Cox FAQ provides a list of the types of traffic that will not be targeted for slowdown, including, it appears, traffic it can not otherwise categorize.  In spite of the FAQ here are few things I'd like to know from Cox:

 

o           just how is the slowing done?  (prioritizing packets?, dropping packets, controlling cable timeslots?, or?)

o           just how does Cox figure out what application the traffic is from?  (port numbers?, deep packet inspection?, or?)

o           where is the congestion experienced?  (local loop?, head end?, uplink?, backbone?, or?)

o           why only residential (not business) customers?  (business links do not get congested?, or?)

o           how often is Cox's network congested ?  (1% of the time, 50% of the time?, every afternoon at 3pm?, or?)

o           how overloaded is Cox's network when it gets congested?  (5x oversubscribed?, 50x oversubscribed?, or?)

 

One thing I'd like to know from the developers of p2p protocols: how long (in hours) do you think it will take you to make your traffic look like it is not one of the applications Cox is targeting for slowing?

 

I fully expect that the FCC will demand the answers to at least some of these questions before too long and I will be interested in seeing what Cox has to say.

 

But, it just might be that Cox's customers will not have to wait for Cox to fess up, at least for some of these questions.  On the same day that Cox announced its traffic mucking plans Google and a few partners announced Measurement Lab (M-Lab) (http://www.measurementlab.net/).  M-Lab provides tools that Internet users can use to see if their ISP is mucking with their traffic.  A fortuitous coincidence -- a set of tools and a demonstration why the tools are needed, both on the same day.

 

Google is working with the New America Foundation's (http://www.newamerica.net/)  Open Technology Institute (OTI), the PlanetLab Consortium (http://www.planet-lab.org/), and a number of academic researchers to make performance testing software available and to deploy testing servers (36 in 12 locations early in 2009) across the Net.

 

It is vital that ISPs be able to employ reasonable, but fair, network management technologies and processes to protect their networks.  Maybe Cox is doing just that, and maybe not.  We will all know in time. Cox has ensured that the question will be addressed earlier, rather than later, in the new regime.

 

disclaimer: Asking questions is Harvard's reason d'etre, but I've not heard Harvard ask the above ones so they are mine.