Small ISPs such as Glass’s (and yes, they do exist, even today) have none of the incentives that network neutrality advocates attribute to major carriers to discriminate against voice, video, or other content. In fact according to Glass, new applications such as VoIP are great because they provide new business. “This week we hooked up a VoIP company which was dissatisfied with the quality of service it was getting from the incumbent in our area. We deployed a low-latency, high-bandwidth radio link just for them, at a cost (parts and labor) of about $1,000. We can justify the cost because we will be paid for the service. It’s cost-shifting without compensation that’s the big issue for all ISPs–large and small.”
Glass has a stake at least as precarious in the current Internet economy as the media companies using peer-to-peer transmission as part of their business plans. Laws or regulations that fail to take economics into account, in his view, could put small ISPs out of business. He defends his position fiercely, and gets plenty of flak in return. I consider Glass a friend and have even planned to tap him as an author on some projects. So I want his view heard as a balance to the “just throw more bandwidth at it” proposals.
That said, I wonder whether the problem is really peer-to-peer protocols, which Glass focuses on, or high-volume media such as video. What architecture could handle the video experiences Internet users want. Compression can achieve impressive quality at reasonable bandwidth, but the sheer volume of everybody sharing the network stresses current transmission systems.
Theres’ no denying Glass’s argument that companies using BitTorrent-style protocols shift the bandwidth burden from the central servers to the edges. That was long touted as a benefit of P2P. And one researcher and system administrator told me recently that BitTorrent is nowhere near as much of a problem as Skype, another peer-to-peer system.
But users who videoconference will use gobs of bandwidth anyway. When you and I videoconference, we’re responsible only for our own traffic, not videos requested by for dozens of other users in our area. But we’re still transferring lots of streaming data. In his network neutrality principles, Glass is essentially asking for new rules governing one particular mode of setting up connections (what we know as peer-to-peer). I’m not convinced this fix aimed at one current mode will prevent users from finding other problematic modes until we find an economic solution to high bandwidth on packet-based networks.
Glass summarized the economics as follows: “Applications with high data volumes require us to ensure that we have sufficient capacity to handle them, but so long as someone pays the ISP, it has an incentive to build out infrastructure as needed. But when the cost is shifted to the ISP without compensation, there’s no incentive (and in fact, no available capital) to grow to accommodate it.”
Does this extend to the leap required to support video streams (and ultimately, multiple video streams?) The economics of bandwidth are incredibly complicated, and attempts to meter it simply increase complexity rather than solve the problems. I reported on the debate over Quality of Service back in 2002, and laid out another report on the economics of bandwidth in 2006, which is still relevant. But Glass’s observations are a contribution to the debate that deserve more consideration.