I wrote a custom BodyParser based these changes and that works properly.
Thanks @gmethvin

Isn't ISO-8859-1 the default charset, as defined at RFC 2616? On a second look, RFC 7230 apparently removed this default.

Edit: but looks like there are other parsers (TolerantText, per instance) using ISO-8859-1 as theirs default. Also, when parsing response bodies, play-ws is using the AHC default charset, which looks like is ISO-8859-1.

@marcospereira RFC 2616 defines defaults for text/* content types but doesn't mention anything about application/x-www-form-urlencoded. TolerantText is designed to be used with text content types, so that default makes sense there.

You can find a decoding algorithm for application/x-www-form-urlencoded here. We don't actually follow this algorithm completely (we don't read the special _charset_ field for example), so we might want to fix that in the future.

Thanks for the explanation, @gmethvin. Also, I think we can add a test case to check cases where the default charset is used and also when the request charset is specified. framework/src/play-integration-test/src/test/scala/play/it/http/parsing/FormBodyParserSpec.scala looks like the right place for Scala specs. Not sure if there is a specific test for the Java version.

@marcospereira tests added

@marcospereira Can you backport to 2.5.x?

@mkurz already doing it. ;-)

👍

Backport to 2.5.x: 6d63a8b