[Chapter-delegates] IETF position on Paid Prioritization

[Fred Baker]

On Sep 3, 2010, at 7:41 AM, Lucy Lynch wrote:

> As long time IETF participants, you and I both know that trying to explain how the IETF sausage is made may be more confusing and not less. I had to go look up the RFC you sited and then decode the type and the abstract in order to post my last explanation and I know how this works!

Yes. Definitely.

I can tell you, as one who was involved in the development of the technology and had been using predecessors to it for well over a decade before, that we expected that networks that provided services based on the diffserv architecture would probably charge money for them or would in some other way build them into their service level agreements. For example, RFC 2597 was based directly on the Frame Relay services that many providers of the time used. They would have edge networks come through the Frame Relay cloud to them, and the Frame Relay cloud provided bi-level (Cascade offered tri-level) drop precedence. It was built at the request of ISPs that wanted to offer a similar service over the Internet, enabling them to offer predictable service to all of their customers, enable customers to burst over their nominal rates, and preferentially drop traffic from such bursts.

RFC 3246, and its predecessor RFC 2598, was similarly designed to enable a paying customer to get a specified rate through the network on a "virtual wire". It was originally designed to support large volume TCP file transfers such as between research institutions, but its greatest use is for voice and video.

Ditto RFC 3662 and 5865.

We of course didn't tell the ISPs how to structure their service offerings. But we provided them with the tools they asked for.

This was discussed at INET '98, btw. http://www.isoc.org/inet98/proceedings/3e/3e_2.htm#s10