spoiler

made you look

  • 0 Posts
  • 177 Comments
Joined 1 year ago
cake
Cake day: July 27th, 2024

help-circle

  • The entire number and math system is just a straightforward implementation of IEEE 754.

    Yeah, but using doubles for everything is its own downsides e.g. it’s why JSON “can’t” store 64bit integers for starters.

    They did add the BigInt class recently, which annoyingly you can’t use with JSON because it requires specialized handling (Because of the aforementioned issue with JSON).

    (So you “can” store 64bit integers in JSON, the spec just says not to, so people just ignore the spec. You just then run into silent truncation issues with clients that do follow it, like browsers.)



  • This behavior is actually in line with what I’d expect, as Unicode support in Windows predates UTF-16, so Windows generally does not handle surrogate pairs and instead operates almost exclusively on WTF-16 code units directly.

    So it’s just straight UCS-2, and the software does enforce that, pretty much the opposite of “WTF-16”.

    Edit: Pretty sure “modern” (XP+ I think) Windows actually does enforce UTF-16 validity in the system, but there’s always legacy stuff from the NT4/2K era that might turn up.







  • I’m still annoyed that “OPAQUE” never seemed to catch on. Uses a username/password combo as normal, but never actually sends the password to the server, only a proof of knowledge. Even if the server is hacked and the DB leaked the attackers can’t actually recover anything resembling a password from it, since the server simply never possesses it.

    Passkeys are superior (No password at all), if only the UX around them was better.