URL scheme characters admitted by DECODE-URL more restrictive than those admitted by TRANSCODE

[Siskin-Bot]

Submitted by: meijeru

DECODE-URL parses the scheme part of an URL (before :) with the following charset:
A - Z a - z 0 - 9 + - . (this is in accordance with RFC 1738).
Then it does TO-LIT-WORD, which eliminates the case of an initial digit or + -, which seems to be allowed by RFC 1738.
TRANSCODE (i.e. the lexical scan) admits the following characters before the characteristic : of a URL! literal:

in initial position A - Z a-z ! & = ? * . ^ _ ` | ~ (note the absence of digits and + -).

In subsequent positions: anything from ! to ~ except [ ] { } ( ) " / :
Thus TRANSCODE is much more permissive than either RFC 1738 or DECODE-URL.

The restrictions mentioned would merit documenting, I think.

---

^{Imported from: CureCode [ Version: alpha 94 Type: Issue Platform: All Category: Datatype Reproduce: Always Fixed-in:none ]
Imported from: metaeducation#1327}

Comments:

---

Rebolbot mentioned this issue on Jan 12, 2016:
DECODE-URL and url! syntax don't obey the url encoding rules
TO-URL returns incorrect value for % symbol
LOAD does not handle correctly URLs containing encoded delimiters
MOLD does not handle correctly URLs containing encoded characters
wrong error on load "1abcde"

---

IngoHohmann mentioned this issue on Feb 5, 2020:
Semicolons should be legal in URL

---

[Oldes]

@meijeru and what do you recommend to do?

1. Limit the transcode? I think it would be a shame.
2. Make the decode-url more permissive?

[meijeru]

What about making transcode and decode-url both conform to RFC-1738?