8

About a year ago, several migration scripts were run to change old HTTP network-internal links across the network to HTTPS links. However, it appears that the script didn't modify links that lack the ending /. For example:

But...

So, excluding cases where the posts were made after the script ran, were edited back after the script modified them, or were somehow overlooked by the script (probably a separate bug), all of the HTTP links in posts which were added before the script ran have been converted into HTTPS links, if they contained the ending /. Links that didn't contain the ending / were not converted.

  • Won't they redirect to https anyway? – Journeyman Geek Mar 8 '18 at 4:46
  • @JourneymanGeek I actually asked this question in chat, and Shog responded that "they don't necessarily work just fine" – Sonic the Inclusive Hedgehog Mar 8 '18 at 4:51
  • 2
    In context... He was also grumpy at you. And was a reason for fixing most of them. I'm sure folks will fix broken links they come across in the normal course of things. – Journeyman Geek Mar 8 '18 at 4:54
  • Shameless plug: SOUP fixes all those HTTP links (and others still scattered in various dusty corners of the SE user interface) automatically. And yes, it handles old meta links correctly, too. – Ilmari Karonen Mar 8 '18 at 22:06
  • @IlmariKaronen So that's why I had trouble telling what Shog was referring to in his link... – Sonic the Inclusive Hedgehog Mar 8 '18 at 22:06
7

There are a few potential issues with the old links floating around:

  1. Badly written code might not understand them and do something sub-optimal.
  2. A redirect through HTTP potentially loses referral data, making it harder to tell how much traffic (say) Meta Stack Exchange is sending to Software Engineering.
  3. A redirect through HTTP potentially leaks sensitive information to intermediate networks.
  4. Crappy extensions that try to convert all links to HTTPS will break upon encountering one of the old meta domains (example).
  5. Possibly a bit slower.

...Most of these are not particularly concerning. The goal of the re-write scripts was to catch the bulk of the old links and thus reduce the amount of redirecting and associated nonsense... But re-writing every link is infeasible; as you've observed, some get edited back in!

That said, if you come across one on a post you're editing anyway, don't hesitate to update it.

  • Can you please explain the one link to Apple that did contain the / but was not edited by the script? – Sonic the Inclusive Hedgehog Mar 8 '18 at 16:37
  • No idea. If it doesn't show up in a precise URL search, that would explain it. – Shog9 Mar 8 '18 at 16:41
  • So you're saying this is a problem, but choosing not to fix it. If that's so, why did you mark this question as "by design"? It doesn't seem from this answer that not converting old links without the ending / was an intentional decision, or that they're supposed to be that way. Can you please remove the "by design" tag, and replace it with "declined"? This implies the correct thing, that you've acknowledged this as an issue but have chosen not to fix it. – Sonic the Inclusive Hedgehog Mar 23 '18 at 18:47
3

I'm not actually sure those things would typically be broken - we'd simply just see a redirect to HTTPS.

If its actually broken - its a semi trivial edit. If its not broken, its a massive task, or many edits, that doesn't fix anything.

It's not a good use of staff time IMO.

If you happen to come across a http link in your normal course of things, try it. Then do the edit and any other that's needed. If its actually broken, you'd be fixing it. I don't really see a major point at this time to go after them other than "It's not quite consistent"

We have other bananas to fry.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .