North American Network Operators Group

Date Prev | Date Next | Date Index | Thread Index | Author Index | Historical

Re: Two Tiered Internet

  • From: Steve Gibbard
  • Date: Wed Dec 14 14:49:48 2005

On Wed, 14 Dec 2005, Marshall Eubanks wrote:

To me, this seems likely to lead to massive consumer dissatisfaction, and a disaster of the
magnitude of the recent Sony CD root exploit fiasco.

Typical Pareto distribution models for usage mean that no matter
how popular "tier 1" sites are, a substantial part of the user time will be spent on degraded "tier 2" sites.

If these don't work, people will complain. Just imagine for a second that cable providers started
a service that meant that every channel not owned by, say, Disney, had a bad picture and sound. Would this
be good for the cable companies ? Would their customers be happy ?

Of course, based on some recent experience this probably means that this will be adopted enthusiastically.
I'm seeing a lot of comments here that appear to be looking at this as a very binary issue -- either it's ok, or it will cause the customers to defect en masse to the competition. This seems to ignore questions of how it would be implemented, and what the competition's offering would be.

If I've got a choice between two providers, both of which are offering a 3 Mb/s pipe, but one of them restricts services from other networks to half of that pipe, that's going to effectively be a situation where one provider is only offering half the Internet bandwidth the other offers.

On the other hand, there could be a scenario in which one network offered a 3 Mb/s unrestricted pipe, while the other offered a 6 Mb/s pipe, with prioritized traffic potentially eating 2 Mb/s of it. That would still be 4 Mb/s of unrestricted traffic vs. the other provider's 3 Mb/s.

In other words, a provider with sufficiently better last mile technology than the competition should be able to do lots of stuff like this and still come out ahead. Providers in markets that are technologically more even might have more trouble.

That assumes rate limiting in the last mile. If what's instead being talked about is QoS tagging of last mile packets, that should be completely irrelevant to those who don't use the services that are prioritized.

Of course, if they're restricting capacity in the backbone and using QoS there, that may be a different story, but that seems unlikely to be what's being talked about. Backbone congestion doesn't tend to happen much in major American cities these days, but individual DSL lines saturate pretty easily.