• Imagine running a measurement study to understand how an application behaves under different network conditions, or generating data for supervised learning problems that require ground-truth labels collected under controlled settings. • You log into Instagram, scroll through several reels, generate video traffic, and observe how performance shifts as latency or bandwidth changes. • Everything works today. • But the next day, the workflow stalls on a slightly altered login screen. • A few days later, a new cookie banner appears, a button moves, or a dialogue loads differently - and the automation you relied on collapses. • Whether the goal is to conduct a rigorous measurement study or to collect labelled data for a supervised learning problem, the underlying issue is the same - we need application workflows that behave consistently, even as the applications themselves evolve continuously.
Article Summaries:
- NetGent introduces a new framework for automating application workflows in networking research. Current browser‑automation tools fail when UI changes-new login screens, cookie banners, or layout shifts-break scripted interactions, limiting data collection for measurement studies and machine‑learning training. NetGent decouples workflow intent from execution, treating data generation as a structured, portable system component. By separating what a task should accomplish from how it is performed, workflows can be regenerated automatically as apps evolve, enabling consistent, repeatable data collection across diverse network conditions. The result is a scalable, agent‑based substrate that improves reproducibility and supports large‑scale experimentation.
Sources: