A place to cache linked articles (think custom and personal wayback machine)
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

index.md 2.2KB

5 years ago
123456789101112131415161718192021222324252627282930313233
  1. title: Web Decay Graph
  2. url: https://www.tbray.org/ongoing/When/201x/2015/05/25/URI-decay
  3. hash_url: b8dd4f6d72c7810d42bc733cbab7509e
  4. <p>I&#x2019;ve been writ&#xad;ing this blog since 2003 and in that time have laid down, along with way
  5. over a mil&#xad;lion word&#xad;s, 12,373 hy&#xad;per&#xad;links. I&#x2019;ve no&#xad;ticed that when
  6. some&#xad;thing leads me back to an old piece, the links are bro&#xad;ken dis&#xad;ap&#xad;point&#xad;ing&#xad;ly
  7. of&#xad;ten. So I made a lit&#xad;tle graph of their de&#xad;cay over the last 144 month&#xad;s.</p>
  8. <div><a href="https://www.tbray.org/ongoing/When/201x/2015/05/25/-big/decay.jpg.html"><img alt="URI decay at ongoing by Tim Bray" title="URI decay at ongoing by Tim Bray" src="https://www.tbray.org/ongoing/When/201x/2015/05/25/decay.png" /></a></div>
  9. <div><p>The &#x201c;% Decay&#x201d; val&#xad;ue for each val&#xad;ue of &#x201c;Months Ago&#x201d; is
  10. the per&#xad;cent&#xad;age of links <em>made in that month</em> that have de&#xad;cayed. For
  11. ex&#xad;am&#xad;ple, just over 5% of the links I made in the month 60 months be&#xad;fore May
  12. 2015, i.e. May 2010, have de&#xad;cayed.</p>
  13. </div>
  14. <p>Longer ti&#xad;tle &#xb7;
  15. &#x201c;A broad-brush ap&#xad;prox&#xad;i&#xad;ma&#xad;tion of URI de&#xad;cay fo&#xad;cused on links se&#xad;lect&#xad;ed for
  16. blog&#xad;ging by a Web
  17. geek with a cam&#xad;er&#xad;a, com&#xad;put&#xad;ed us&#xad;ing a Ru&#xad;by script cooked up in 45
  18. minutes.&#x201d;
  19. Mind you, the script took the best part of 24 hours to run, be&#xad;cause I was too
  20. lazy to make it run a hun&#xad;dred or so threads in par&#xad;al&#xad;lel.</p>
  21. <p>I sup&#xad;pose I could regress the hell out of the da&#xad;ta and get a pret&#xad;ti&#xad;er line
  22. but the sto&#xad;ry these num&#xad;bers are telling is clear enough.</p>
  23. <p>Another way to get a smoother curve would be for some&#xad;one at Google to
  24. throw a Map/Re&#xad;duce at a his&#xad;tor&#xad;i&#xad;cal dataset with hun&#xad;dreds of bil&#xad;lions of links.</p>
  25. <p>This is a very sad graph &#xb7;
  26. But to be hon&#xad;est I was ex&#xad;pect&#xad;ing worse.
  27. I won&#xad;der if, a hun&#xad;dred years af&#xad;ter I&#x2019;m
  28. dead, the on&#xad;ly ones that re&#xad;main alive will be&#xad;gin with &#x201c;en.wikipedia.org&#x201d;?</p>