Merge pull request #654 from capnproto/harris/allow-underscore-in-hostname

Allow underscores in URL hostnames

c++/src/kj/compat/url-test.c++

@@ -265,6 +265,10 @@ KJ_TEST("parse / stringify URL") {
 
   // Scheme and host are forced to lower-case.
   parseAndCheck("hTtP://capNprotO.org/fOo/bAr", "http://capnproto.org/fOo/bAr");
+
+  // URLs with underscores in their hostnames are allowed, but you probably shouldn't use them. They
+  // are not valid domain names.
+  parseAndCheck("https://bad_domain.capnproto.org/");
 }
 
 KJ_TEST("URL percent encoding") {

c++/src/kj/compat/url.c++

@@ -38,7 +38,11 @@ constexpr auto END_QUERY_PART = parse::anyOfChars("&#");
 constexpr auto SCHEME_CHARS = ALPHAS.orGroup(DIGITS).orAny("+-.");
 constexpr auto NOT_SCHEME_CHARS = SCHEME_CHARS.invert();
 
-constexpr auto HOST_CHARS = ALPHAS.orGroup(DIGITS).orAny(".-:[]");  // [] is for ipv6 literals
+constexpr auto HOST_CHARS = ALPHAS.orGroup(DIGITS).orAny(".-:[]_");
+// [] is for ipv6 literals.
+// _ is not allowed in domain names, but the WHATWG URL spec allows it in hostnames, so we do, too.
+// TODO(soon): The URL spec actually allows a lot more than just '_', and requires nameprepping to
+//   Punycode. We'll have to decide how we want to deal with all that.
 
 void toLower(String& text) {
   for (char& c: text) {