zlacker

[return to "Tell HN: t.co is adding a five-second delay to some domains"]
1. kens+71[view] [source] 2023-08-15 04:20:25
>>xslowz+(OP)
I can confirm. NYT shows a five-second redirect delay: "wget https://t.co/4fs609qwWt". It redirects to gov.uk immediately: "wget https://t.co/iigzas6QBx"
◧◩
2. mzs+r52[view] [source] 2023-08-15 19:18:22
>>kens+71
I don't see it:

  % curl -gsSIw'foo %{time_total}\n' -- https://t.co/4fs609qwWt https://t.co/iigzas6QBx | grep '^\(HTTP/\)\|\(location: \)\|\(foo \)'
  HTTP/2 301 
  location: https://nyti.ms/453cLzc
  foo 0.119295
  HTTP/2 301 
  location: https://www.gov.uk/government/news/uk-acknowledges-acts-of-genocide- committed-by-daesh-against-yazidis
  foo 0.037376
◧◩◪
3. lapcat+yd2[view] [source] 2023-08-15 19:57:49
>>mzs+r52
I think Twitter, err, X, just turned off the delay now that it's getting big media attention. I could reproduce it over and over again a little earlier, but now I can't anymore: >>37138161

[Edit:] I'm still seeing it with threads.net:

  curl -v -A 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/16.6 Safari/605.1.15' https://t.co/DzIiCFp7Ti
◧◩◪◨
4. mzs+nj2[view] [source] 2023-08-15 20:29:34
>>lapcat+yd2
I don't see it with your URL either:

  % curl -gsSIw'foo %{time_total}\n' https://t.co/DzIiCFp7Ti | grep '^\(HTTP/\)\|\(location: \)\|\(foo \)'
  HTTP/2 301 
  location: https://www.threads.net/@chaco_mmm_room
  foo 0.123137
Doesn't matter if I do a HTTP/2 HEAD or GET:

  % curl -gsSw'%{time_total}\n' https://t.co/DzIiCFp7Ti 
  0.121503
HTTP/1.1 also shows no delay:

  % curl -gsSw'%{time_total}\n' --http1.1 https://t.co/DzIiCFp7Ti
  0.120044
I chalk this up to rot at X/twitter that is being fixed now that it was noticed.
◧◩◪◨⬒
5. lapcat+3k2[view] [source] 2023-08-15 20:32:52
>>mzs+nj2
> I don't see it with your URL either

That's because you're not spoofing the User-Agent to be a browser rather than curl.

◧◩◪◨⬒⬓
6. mzs+3r2[view] [source] 2023-08-15 21:21:09
>>lapcat+3k2
Oh that's it, thanks! In fact it it returns a 200 not 301 then:

  % curl -gsSw'%{time_total}\n' -A 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/16.6 Safari/605.1.15' https://t.co/DzIiCFp7Ti
  <head><noscript><META http-equiv="refresh" content="0;URL=https://www.threads.net/@chaco_mmm_room"></noscript><title>https://www.threads.net/@chaco_mmm_room</title></head><script>window.opener = null; location.replace("https:\/\/www.threads.net\/@chaco_mmm_room")</script>4.690000
  % curl -gsSIw'%{time_total}\n' -A 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/16.6 Safari/605.1.15' https://t.co/DzIiCFp7Ti
  HTTP/2 200 
  ...
  content-length: 272
  ...
  x-response-time: 4524
  ...
  
  4.660211
The delay is not there for nyti.ms (anymore) but once you use the Safari UA it's handled as 200 response:

  % curl -gsSIw'foo %{time_total}\n' -A 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/16.6 Safari/605.1.15' https://t.co/4fs609qwWt https://t.co/iigzas6QBx | grep '^\(HTTP/\)\|\(location: \)\|\(foo \)'
  HTTP/2 200 
  foo 0.126043
  HTTP/2 200 
  foo 0.037255
It really does seem that twitter is adding a 4.5s delay to some sites from web browsers. Could be malicious, could be rot...
◧◩◪◨⬒⬓⬔
7. lal+nn5[view] [source] 2023-08-16 17:42:40
>>mzs+3r2
The specific logic with user agents is that it happened (I think they've ended it now?) whenever the word "curl" was not in your user agent string. If the substring "curl" was contained anywhere in your user agent string, it did not have a delay. I cannot imagine how it could rot in that specific way non-maliciously.
[go to top]