Link to full paperAbstract
In 2016, the FCC offered broadband providers a safe harbor for certain disclosures required under the Open Internet transparency rule, if the broadband provider disclosed on FCC-designated broadband labels accurate information about broadband commercial terms, performance, and network management. The Restoring Internet Freedom Order eliminated this safe harbor. However, the 2021 Infrastructure Investment and Jobs Act will require ISPs to display these broadband labels, and directs the FCC to update them. The FCC issued a Notice of Proposed Rulemaking, asking in part if it should add and/or remove information from the 2016 broadband labels.
Comments filed in response to the NPRM reveal several disagreements between stakeholders about what the performance and network management sections of the broadband labels should disclose. Very few of these arguments are grounded in any data or analysis. In this paper, I analyze these disagreements.
Several ISPs assert that packet loss is an esoteric metric. I demonstrate that packet loss has been one of the three fundamental network performance metrics ever since the Internet began. I also demonstrate that the quality of interactive applications such as video conferencing is dependent on speed, delay, and packet loss, and that this has been understood for many decades. Some ISPs assert that packet loss cannot be decreased without decreasing throughput or increasing latency. I show that there is no such inherent tradeoff, that this has been repeatedly proven mathematically, and that this has been widely understood for many decades.
Several ISPs argue that network performance should not be measured during peak usage periods, because they assert that peak usage period network performance has little relevance to customers’ typical experiences. I analyze very large sets of measurement data, and show that there is a large variation in mobile network performance between the peak usage period and off-peak usage periods.
One trade association argues that mobile broadband providers should be allowed to disclose an “expected” speed range that is not based on actual measured speeds. I examine such past disclosures of expected speeds, and I show that these expected speeds were far off from the actual measured speeds.
One stakeholder suggests that disclosures about blocking, throttling, or paid prioritization practices are irrelevant and/or confusing, and that the broadband label should not include them. I give examples of application-specific throttling, and I suggest simple disclosures of these application-specific practices that are relevant and useful.
One trade association argues that disclosure about subscriber-triggered network practices do not convey meaningful information to consumers, and that the broadband label should not include them. I give examples of subscriber-triggered throttling, and I suggest simple disclosures of these subscriber-triggered practices that are relevant and useful.
In addition, I propose several improvements to the 2016 labels. First, I use large sets of measurement data to demonstrate that there is a large variation in mobile network performance metrics, and I propose specific percentiles that convey this variation. Second, I discuss the routes over which network performance may be measured, and propose requirements about these routes.
Open Zoom
