查看更多>>摘要：This January 2025 issue contains one technical paper and two editorial notes. The technical paper, An Analysis of QUIC Connection Migration in the Wild, by Aurelien Buchet and Cristel Pelsser, provides a comprehensive examination of the support of the QUIC connection migration mechanism over the Internet. The authors perform Internet-wide scans revealing that despite a rapid evolution in the deployment of QUIC on web servers, some of the most popular destinations do not support connection migration yet. Then, we have our two editorial notes. The first, When Something Looks Too Good To Be True, It Usually Is! AI Is Causing A Credibility Crisis In Networking, by Walter Willinger and colleagues, raise awareness about a deeply concerning and yet muchoverlooked development in the use of Artificial Intelligence (AI) and Machine Learning (ML) for solving problems in science in general and in networking in particular. The second editorial note, Learning Algorithms for Dynamic Call Routing: Lessons from Yesteryears, by Deep Medhi, presents an overview of various dynamic call routing schemes, and in particular, learning algorithms for dynamic call routing from yesteryears, discusses control mechanisms that were deployed for network stability, and most important, presents lessons learned from this work, which could hopefully be useful in applying artificial intelligence or machine learning (AI/ML) to networking in today's world.
查看更多>>摘要：The paper presents a well-executed measurement study on QUIC migration when HTTP3 (H3) is used, offering an important early snapshot of QUIC migration deployment. The study is thorough, and the paper is well-written, with a clear outline of the study's limitations. It effectively separates the measurement results from discussions and assumptions that go beyond what the data can support. Strengths include a clear and structured study process, an insightful presentation of results in Table 1, and the open-sourcing of both tools and data. The paper makes significant progress by addressing earlier concerns, such as clarifying its focus on HTTP3+QUIC, reducing the emphasis on connections without SNI, and ensuring that speculation is clearly separated from concrete results. While there are still some challenges, such as the study of certain QUIC parameters and the simplified handshake diagram in Figure 1, the paper provides a strong foundation for future research. It offers valuable insights into the current state of QUIC migration deployment and sets the stage for further work in this area.
查看更多>>摘要：As QUIC gains attention, more applications that leverage its capabilities are emerging. These include defenses against onpath IP tracking and traffic analysis. However, the deployment of the underlying required support for connection migration remains largely unexplored. This paper provides a comprehensive examination of the support of the QUIC connection migration mechanism over the Internet. We perform Internet-wide scans revealing that despite a rapid evolution in the deployment of QUIC on web servers, some of the most popular destinations do not support connection migration yet.
Walter WillingerRonaldo A. FerreiraArpit GuptaRoman (Sylee) Beltiukov...
10-15页
查看更多>>摘要：The purpose of this editorial note is to raise awareness about a deeply concerning and yet much-overlooked development in the use of Artificial Intelligence (AI) and Machine Learning (ML) for solving problems in science in general and in networking in particular. To put it simply, in today's age of AI/ML, the much-publicized and well-documented “reproducibility crisis” in science is further compounded by an inconspicuous and rarely mentioned “credibility crisis.” More to the point, by focusing on the area of networking research, we provide evidence that among the already small number of reproducible scientific publications that describe AI/ML-based solutions, even fewer, and often none, describe trained AI/ML models that are “credible;” that is, can be trusted to not only perform well in their original training domain but also in new and untested environments. We elaborate on the root cause of this credibility crisis, discuss why the credibility of AI/ML models is of paramount importance for their successful use in practice, and put forward an aggressive but imminently practical proposal for addressing this crisis head-on so as to pave the way for a future where networking research can reap the full benefits of AI/ML.
查看更多>>摘要：With the advent of softwarization of digital telephone switches, many dynamic call routing schemes were explored in the 1980s to provide better network performance. In particular, we highlight two learning algorithms for dynamic call routing from that era. We note that while the learning algorithms have the adaptive capability to benefit dynamic call routing performance, they alone cannot address network instabilities in certain network load conditions. Additional controls are necessary. This note is to present an overview of various dynamic call routing schemes, and in particular, learning algorithms for dynamic call routing from yesteryears. We also discuss control mechanisms that were deployed for network stability. Finally, we present lessons learned from this work, which could hopefully be useful in applying artificial intelligence or machine learning (AI/ML) to networking in today's world.
NETL
NSTL