[HN Gopher] More tales about outages and numeric limits
___________________________________________________________________
More tales about outages and numeric limits
Author : todsacerdoti
Score : 42 points
Date : 2025-11-21 13:44 UTC (9 hours ago)
(HTM) web link (rachelbythebay.com)
(TXT) w3m dump (rachelbythebay.com)
| taeric wrote:
| I agree that hammering random things isn't where you should
| start. I do sadly have to admit that "just do something" is
| unreasonably effective. Which means if you have other things
| lined up, having teams that are somewhat randomly attacking
| things is probably going to see more success than you'd expect.
| nonameiguess wrote:
| This didn't cause an outage, but I was endlessly amused about
| regression test failures caused by synthetic data when the Navy's
| maritime domain awareness switched over to a tip file format that
| embedded pngs of each ship detection into an xml file and had
| some limit I'll never remember on the maximum number of such tips
| it could fit into a single file.
|
| It was actually a testament to the thoroughness with which the
| NRO regression tests its shit, but the complaint was that the
| simulated scan the program was running detections on crapped out
| after something like 30,000 detections, which is already more
| ships than the US Pacific Fleet and probably more ships than any
| foreign Navy has in every ocean on the planet, certainly more
| than will ever be contained in a single image.
|
| But in case I'm being 1981 Bill Gates level naive here, they are
| in fact prepared for the eventuality of gigantic swarms of
| microships taking over the oceans because of that regression
| test.
| bee_rider wrote:
| That seems like a good test, DDOS-ing the detector is
| considered "part of the game" in war, right? (most things are,
| I guess).
___________________________________________________________________
(page generated 2025-11-21 23:01 UTC)