Thread by @phonedude_mln

9.6 Million Links in Source Code Comments: Purpose, Evolution, and Decay
arxiv.org/abs/1901.07440

after quick read:

* ~80% of sw repos have >=1 URL in comments

* ~81% of links are still 200 (19% mix of 404, 500, 403, 405)

* expected decay nums; 1st study of links in comments?

What I didn't see:

* checking for "soft 404s" - 200s that say "can't find record" doi.org/10.1145/988672…

* content drift - still 200 but aboutness changed doi.org/10.1371/journa… (cited, but for something else)

* normalizing for time e.g, link_age = 1 day vs link_age = 1 year

Also appears that they fixed links in a method similar to "Moved but not gone: an evaluation of real-time methods for discovering replacement web pages" -- using #mementos to generate keywords to find new URL on live web (vs. linking to archived copy)

doi.org/10.1007/s00799…

Other works that could have been cited:

* 1/5 scholarly articles have link rot dx.doi.org/10.1371/journa…

* 50% of SCOTUS opinions have rot/drift dx.doi.org/10.2139/ssrn.2…

* 11% of links shared in Twitter disappear after 1 year arxiv.org/abs/1209.3026 arxiv.org/abs/1309.2648

Overall a really neat study.

Decay numbers are in line with prior pubs, but this sampled from a new & novel corpus.

Link targets (e.g., pubs, API docs) are probably more replaceable than "regular" links.

cc @hideakihata @ctreude
@hvdsomp @mart1nkle1n