May 3, 2006

Day 1 - SIP, Network Neutrality, ICANN, and the Internet

Well, here I am again, for the 19th year, working behind the scenes at the Interop trade show.  Yes, that's me.

This year I am working with SIP based VOIP, particularly in conjunction with firewalls, NATs, secure VLANs, QoS enabled wireless, and induced packet loss, delay, duplication, jitter, and reordering (and many other impairments produced by my Maxwell system.)

We have seen some of what we expected: that NATs and VOIP live more in a state of mutual hostility rather than halcyon peace.  And it is also fairly clear that the quality that people perceive from VOIP often evaporates fairly quickly when network conditions begin to erode away from some laboratory standard of perfection.

Simply throwing 50 milliseconds of bidirectional jitter on top of 50 milliseconds of bidirectional delay, coupled with a couple of points of burst packet loss, will cause a SIP+RTP/RTCP based call on top of a TCP based VLAN (SSL) to become nearly unusable.  That's not a particularly unusual degree of network degredation.  And the resulting behavior of the phones on the plane of an automobile that loses its tires if it hits a small pothole while traveling at 50mph.

What this means to me is that there are reasons why certain classes of network traffic do need special handling - and thus why the so-called network neutrality debate is more nuanced than might first appear.  As I've said elsewhere, it's my feeling that non-equal treatment of traffic is appropriate, and that slightly increased prices for enhanced service might be appropriate, but that the choice about how traffic is to be handled must be in the hands of the user, not the provider.

Which brings us by a commodius vicus of recirculation back to ICANN and environs...

The news and blogs have recently agreed with my prior notes that ICANN's "add grace" system has resulted in a huge number of unpaid, 5-day name registrations.  The ratio of these unpaid registrations to paid registrations is on the order of 200:1.

And that, in turn, means that the entire foundation upon which ICANN has approved the $7 fee registry fee in the proposed ICANN-Verisign contract has no basis in reality.  There are strong indications that rather than $7 a more appropriate registry fee would be about $0.02 if the freebee 5-day registrations had to pay for the registry services they consume.

ICANN's board should realize that they have been hoodwinked about the Verisign registry fee.  ICANN's board should quickly rescind its approval of the contract until it receives a believable and deep audit of the actual registry costs and isolates the effects of ICANN's "add grace" period.

 But does ICANN's board have the courage to stand up to its staff?  I doubt it.

Posted by karl at May 3, 2006 1:46 PM