we live in a timeline where ! isn’t urlencoded but : is.

[domi@donotsta.re]

we live in a timeline where ! isn’t urlencoded but : is.

what happened?? this feels wrong

[theresnotime@fox.nexus]

@domi urlencodes u

[lanodan@queer.hacktivis.me]

@domi Kind of wonder if : being urlencoded is due to some pre-http legacy.
Or urgh, Windows.

[domi@donotsta.re]

@lanodan well, it can be a part of the user/password? so that might have been a bad example

but same can be said for ; which genuinely has no right needing to be urlencoded imho

[easrng@pleroma.envs.net]

@domi @lanodan some servers used ; instead of & as a separator iirc

[domi@donotsta.re]

@easrng @lanodan well, then they’re violating the URL spec. while the urlencoding is defined very broadly (mayhaps a bit too broadly), the parameters are defined quite strictly: ? to begin, = to separate rhs and lhs, & to separate further indices

[lanodan@queer.hacktivis.me]

@domi @easrng Wouldn't be surprised of it being super-legacy from pre-RFC1738 (December 1994).

[erincandescent@akko.erincandescent.net]

@lanodan @domi @easrng prospero:// URLs from 1738 use ;parameters

also RFC2396 requires escaping it in path segments b/c its used for path params:

      path          = [ abs_path | opaque_part ]
      path_segments = segment *( "/" segment )
      segment       = *pchar *( ";" param )
      param         = *pchar
      pchar         = unreserved | escaped |
                      ":" | "@" | "&" | "=" | "+" | "$" | ","

An example of a URI scheme using params is IMAP: imap://[email protected]/INBOX/;uid=20/;section=1.2

[erincandescent@akko.erincandescent.net]

@lanodan @domi @easrng anyway the URL generic syntax specs a bunch of extra reserved characters so the person *defining* a given URL or URL scheme can use them as delimiters, use their encoded form to escape values that contain said delimiters, and not have generic URL parsers mangle things