Open Internet Order’s DNS Problem

The challengers to the FCC’s reclassification of Internet Service under Title II of the Communications Act filed their reply brief to the FCC’s brief on Monday. The challengers – led by US Telecom – made a number of legal arguments about the shortcomings in the way the FCC went about drafting an order without allowing comment of their proposed rules. My guess is that these arguments will prevail because the FCC Open Internet Order, 2015-style, is so different from previous net neutrality orders. The 2010 order’s rules, for example, were couched in language that was included verbatim in the Notice of Proposed Rule Making that preceded it. So commenters were able to criticize or support the actual rules before they became final.

Not only did the 2015 NPRM not do that, it asked for comment on a set of rules that were nothing like the final ones. So it was bait and switch. The FCC’s approach was rushed, clumsy, and arrogant so I don’t see it surviving the challenge. But I’m not a lawyer.

The UST brief carefully revealed the errors and inconsistencies in the Order on all of the key factual and analytical issues as well. I was especially pleased to see my own amicus brief cited twice on DNS as well as a High Tech Forum blog post on the FCC’s problems.

The first reference is about the necessity of DNS to Internet Service:

If broadband providers did not offer DNS as part of their Internet access service, mass-market consumers would find that service useless for accessing the Internet’s various applications.   See  Bennett Amicus 10-11;

Some may argue that the Internet doesn’t require DNS because we could access apps and sites by IP address, but that would impose a huge loss of functionality for the web. We access web sites by name – the HTTP “GET” protocol message sends the domain name from the browser to the web server, even though this is wrapped up in a TCP exchange. But web servers (such as the open source Apache server) allow multiple web domains to share a single IP address. This is especially useful to small blogs and other web services hosted by commercial web hosting companies. Some web servers host 100 or more web sites, all distinguished by name.

CDNs also play games with domain name resolution to direct traffic to higher-performance locations either closer or more lightly loaded than other possible servers. This isn’t routing as much as dynamic name resolution; routing is done by IP and BGP at the level of IP addresses. Domain names are a second, parallel name (or address) space to the IP address space. To fully utilize the Internet, you need both.

UST argues that today’s DNS is a richer service than the DNS we had when the FCC first classified cable modem and DSL as information services:

The FCC held in 2002 that DNS “do[es] not” fit within the management exception,  Cable Broadband  Order  ¶¶ 37-38 & n.150, and DNS has only gained functionality since then.  For instance, DNS now actively protects users from Internet attacks and scams.[4]

The footnote points to comments in my amicus on DNSSEC: “See  Bennett Amicus 11; Bennett Dec. 30, 2014 Ex Parte 8 (JA___).   Unlike SS7 signaling (FCC Br. 76), DNS does not  merely route calls to a specific destination.”

The FCC argues that DNS (and caching) are  information services when offered by third parties such as Google but are network management functions when offered by ISPs. This obviously makes no sense. Hence UST invokes the analysis of DNS and network management I wrote here on High Tech Forum:

Indeed, the fact that these services are offered by third parties demonstrates that they do not “manage[] a telecommunications network.”  Internet access providers would not (indeed, could not)  allow end users to give third parties authority or responsibility for “managing” their networks.[6]

Here’s the footnote: “See  Richard Bennett,  FCC Brief Painfully Wrong About DNS, HighTech Forum  (Sept. 17, 2015),”

Consequently, we see that the FCC’s construction of Internet Service is technically unsound and logically inconsistent. So even if the court excuses the 2015 Open Internet Order’s procedural flaws, its reasoning is in deep trouble.

  • RipeForDiscussion

    We get it, your don’t think that DNS fits within the management function found in 47 U.S.C. 153(24). That being said, it seems as if you argument concerning why DNS does not fit within the management functions is that the FCC failed to properly to describe DNS, using unlettered language such as “routing”, and that the FCC gestures towards the management function rather than making a serious attempt at explaining DNS’s management qualities.

    My overall question to you would be how is DNS not managing? When an internet user would like to reach a website (endpoint) they must go through DNS, which translates the alphabetic domain name to the numeric IP address. Once DNS determines whether the domain name entered corresponds to the IP address, the ISP provides the transmission of connectivity so that the use cna reach the endpoint. If you ask me, and nobdy has yet :), if it wasnt for DNS managing domain names into IP addresses ISPs transmission of connectivity would be much more cumbersome.

    Even if you were to argue that DNS is an information service which falls outside the management exception found in 153(24), does this remove the transmission of connectivity provided by the ISP out of the realm of Title II telecommunications service. After all, isnt that what’s at stake here … were talking about that transmission of connectivity, while your amicus brief, and subsequent articles go back and forth between couching DNS as somewhere in the shadow IP Protocol and BGP and failing to remain within the management function of 153(24).

    I respect the **** out of your work, we disagree on this one.
    I go more in depth into your article and DNS here:

    • Richard Bennett

      To fit within the legal concept of management functions, DNS would need to have some connection to the transmission of information between the ISP customer’s premise and the corresponding service endpoint, which it doesn’t. It’s extremely convenient to have a service that translates domain names into IP addresses, but in many (indeed, most) cases it’s not actually necessary for transmission. If have an IP address, wherever I got it, I can send a message to Netflix, Google, et al.

      Even more interesting, the site the user contact can reply to her without making a DNS query, since it simply uses the source IP address in the query as the destination IP address in the response. Even more interesting is the fact the ISP does most of its work on the return path rather than the initiating path due to hot potato routing. Routing on the Internet is asymmetrical, another way the Internet differs from the PSTN.

      While DNS isn’t necessary for transmission, it is necessary for security; DNSSEC protects the end user from man-in-the-middle attacks, which plain DNS doesn’t. This is an information service function, not simply a transmission function.

      So DNS is a vital part of the Internet Service for reasons that have nothing to do with transmission.

  • RipeForDiscussion

    What makes DNS not a management function? If you ask me, DNS is the starting point which manages a consumer’s choice in reaching a certain endpoint. If a consumer wants to go to a website by utilizing a certain ISP’s connectivity, DNS manages the translation of alphabetic domain name into numeric IP address.

    • Richard Bennett

      As I said on your blog: “The most glaring hole in the FCC’s argument that DNS is a management function is the fact that DNS can be provided by third parties. If a user chooses Google DNS over Comcast DNS, according to the FCC Google is then managing Comcast’s network. As US Telecom points out in its reply brief, this is an absurd claim on the FCC’s part. Google cannot manage ISP networks (other than its own), it’s simply a user of ISP services like everyone else.”

Comments are closed.