Let’s relearn web caching together.
Prompt Caching In Agents | EARENDIL
How prompt caching shapes the cost, latency, tools, and architecture of coding agents, and what Pi does to keep cache behavior visible.
docs/research/session-save-restore.md at main · burrito.space/peek
experiments in a post-browser web
Solutions for Link Rot on the Modern Web
Web links are fundamental to the web, enabling navigation between pages and citations in research articles. However, web links suffer from "link rot", a phenomenon in which links are likely to become inaccessible over time. This can occur if a link’s site disappears, making it impossible to resolve the hostname or establish a server connection, or the linked page has been deleted, resulting in an HTTP 404 "Not Found" error. Today, a common solution to tackle link rot is to rely on web archives, which capture snapshots of web pages for future reference. However, the modern web has evolved significantly since web archives were first introduced, leading to several limitations in their effectiveness. First, the scale of the web has grown tremendously, making it infeasible to crawl every page whenever it changes. As a result, many broken links either have no archived copies, or the archived copies have stale content. Second, modern web pages rely heavily on increasingly complex and diverse JavaScript. This shift has made it more challenging for preserving fidelity in archived copies, significantly increasing both the computational cost of operating browser-based crawlers and engineering effort required to maintain accurate replay systems. This thesis presents a set of solutions to cope with link rot on the modern web. My work aims to mitigate the various limitations of web archives. First, for broken links without any archived copy, or if the archived copy includes stale content or unavailable functionalities, I built Fable. Fable revives the dead link with the new URL to the same page whenever available. Compared to prior approaches, Fable revives 4.6K broken links—a 50% increase—with much higher accuracy. Second, for pages that require dynamic crawling, I show how web archives can achieve a better tradeoff between efficiency and fidelity. By carefully choosing 8.9% of pages to crawl dynamically and strategically reusing resources from those crawls, an archive can serve 99% of the remaining statically crawled pages without any fidelity loss. Third, for fidelity violations in archived copies that are caused by incorrect edits to crawled scripts, I built FidEx. FidEx reliably detects when an archived page differs from its original version and pinpoints the root cause. After fixing the most common errors pinpointed by FidEx, I reduced the fraction of pages for which FidEx reports a violation of fidelity from 15% to 9%.
Solutions for Link Rot on the Modern Web
Web links are fundamental to the web, enabling navigation between pages and citations in research articles. However, web links suffer from "link rot", a phenomenon in which links are likely to become inaccessible over time. This can occur if a link’s site disappears, making it impossible to resolve the hostname or establish a server connection, or the linked page has been deleted, resulting in an HTTP 404 "Not Found" error. Today, a common solution to tackle link rot is to rely on web archives, which capture snapshots of web pages for future reference. However, the modern web has evolved significantly since web archives were first introduced, leading to several limitations in their effectiveness. First, the scale of the web has grown tremendously, making it infeasible to crawl every page whenever it changes. As a result, many broken links either have no archived copies, or the archived copies have stale content. Second, modern web pages rely heavily on increasingly complex and diverse JavaScript. This shift has made it more challenging for preserving fidelity in archived copies, significantly increasing both the computational cost of operating browser-based crawlers and engineering effort required to maintain accurate replay systems. This thesis presents a set of solutions to cope with link rot on the modern web. My work aims to mitigate the various limitations of web archives. First, for broken links without any archived copy, or if the archived copy includes stale content or unavailable functionalities, I built Fable. Fable revives the dead link with the new URL to the same page whenever available. Compared to prior approaches, Fable revives 4.6K broken links—a 50% increase—with much higher accuracy. Second, for pages that require dynamic crawling, I show how web archives can achieve a better tradeoff between efficiency and fidelity. By carefully choosing 8.9% of pages to crawl dynamically and strategically reusing resources from those crawls, an archive can serve 99% of the remaining statically crawled pages without any fidelity loss. Third, for fidelity violations in archived copies that are caused by incorrect edits to crawled scripts, I built FidEx. FidEx reliably detects when an archived page differs from its original version and pinpoints the root cause. After fixing the most common errors pinpointed by FidEx, I reduced the fraction of pages for which FidEx reports a violation of fidelity from 15% to 9%.
Reorganizing Website Architecture for HTTP/2 and Beyond
This document discusses reorganizing website architecture for HTTP/2 and beyond. It summarizes some issues with HTTP/2 including errors in prioritization where some browsers fail to specify resource priority properly. It also discusses the problem of TCP head-of-line blocking where pending data in TCP buffers can delay higher priority resources. The document proposes solutions to these issues such as prioritizing resources on the server-side and writing only what can be sent immediately to avoid buffer blocking. It also examines the mixed success of HTTP/2 push and argues the server should not push already cached resources. - Download as a PDF or view online for free
Theming atproto · blog.damato.design
Taking Mise en Mode into the new web standard
Archiving URLs
Archiving the Web, because nothing lasts forever: statistics, online archive services, extracting URLs automatically from browsers, and creating a daemon to regularly back up URLs to multiple sources.
Decache - Easy Cache Exporter
Instantly check your internet cache for over 500 lost YouTube videos and Flash games.
Freeing the Web from the Browser | reinterpretcast.com
The Web is, without a doubt, the most powerful research tool currently available to man. No longer must researchers comb through endless indices and catalogues to find what they are looking for. Instead, hyperlinks allow users to associatively explore rich interconnected structures of information with ease. In just a few clicks, it is possible to navigate through a vast global space of information comprising documents on virtually any subject.
Web bloat isn’t a knowledge problem | Christian Heilmann
There is a consensus among all browser makers and lovers of the web: the current state of the web is a mess. The average web page is 2.4 megabyte big and has over 200 requests.
ArchiveBox - Open-source self-hosted web archiving
Preserve websites, media, bookmarks, feeds, source code, evidence, and research material in durable files you control.
Webrecorder: Web Archiving for All
Webrecorder provides open source solutions for everyone to archive the complex, interactive Web.
Apache Calcite | Proceedings of the 2018 International Conference on Management of Data
As part of the Digital Library's transition to Open Access, new features for researchers are available in the Premium Edition. Click here to learn more.
