Join the discussion

Write your take first — we'll ask for email only when you're ready to publish.

  • Hacker News
  • the part that still bites is the retry loop one bad IP is recoverable one bad IP plus polling every second until it answers turns every outage into a flood
  • The talk he gave at Usenix LISA was one of the best ever!

    (LISA is now SRECon)

  • Surely these were Rogue Routers, not flawed ones? Hell bent on world domination, just like recent AI agents :-D
  • 2003, I was still young in 2003. Kinda.

    In 2003, Network Address Translation (NAT) was becoming the default LAN IP management scheme, for two reasons: (1) IPv4 space was at the beginnings of its inevitable resource restrictions, and (2) the perceived security of the NAT had effectively become what people believed a firewall performed.

    In those old days of the middle Internet and early web, it was still fairly common to visit an IP address (IPv4 - IPv6 was ratified but still being interoperability tested, barely a toy yet), you could visit an IP address directly without using the DNS.

    Peer distribution was often a derivative of the sneakernet (carrying media around on your person), emailing long lists of popular but undiscoverable media content.

    The concept of the home gateway device was young but functional. You could go to Best Buy and get a router to pair with your CPE; integration of CPEs and routers had already happened but it was still common to see them as separate devices.

    Back then, traffic was a sliver of what it is now; a systems operator might have personally seen anomalous traffic, ingress and maybe even more unlikely, egress, and had the notion "that ain't right". These days we rely on incredible solutions that have existed for a long, long time to alert us, and self manage, these types of events.

    While experimenting with codex this past week (old guy trying to participate but I gotta stretch out first, old), I tossed a pair of 8.8x3 inch monitors on a pi4 and told codex to 'go wild' making me a poor man's ATC kit, with a pile of html real-time gauges of everything I could think of - outer space, space station, planes, weather, home assistant temperatures, all of it.

    And then at the outset I asked codex, hey, how often are you hitting these endpoints? How often do you ask the web what time sunrise is? Ah, every 900 seconds.

    And so my Abe Simpson clincher, is, it's very easy for me to see how the ballast of the data on the internet is just bots, talking to bots, and how a 10 second NTP selection 2 decades ago was just the same move, before the immense scaling and false resource abundance we have today, allowed it to continue unchallenged.

  • This is likely being discussed because it's relevant to this similar modern day story involving Tesla: https://news.ycombinator.com/item?id=49686766
  • It struck me how much more useful information the graphs and diagrams here provide over their flashier counterparts you commonly see today.

    The old-fashioned aesthetic almost feels like a cue what I'm looking at won't be slop.

  •   > Are incidents such as this a likely side-effect of ubiquitous, low-cost, perhaps even disposable Internet hosts?
    
    
    The whole write-up is excellent. I was already impressed by the author, but the prescience of this comment... wow.

Explore Birbla archives