WEBCLIENT do not encode URL fully (according to RFC 3986, only query parameters

[lasteris]

Questions

Why Vert.x WebClient encodes only query part of URL ? I am not sure, that i am right, but accordingly to RFC 3986 i assumed that ALL characters, that seems not valid to US-ASCII should being encoded by percent-notation. Am i wrong ?

Version

4.5.0

Context

I've encountered this problem, while I was trying to use Cyrillic symbols in path (not in query part).

Do you have a reproducer?

GIST: https://gist.github.com/lasteris/859c75bcd8640b15a27655cb0784c8b1

Steps to reproduce

1. run
2. curl --location 'http://127.0.0.1:8881/typo/?hi=%D0%9F%D1%80%D0%B8%D0%B2%D0%B5%D1%82'
3. view logs
4. see 20:22:22.651 [vert.x-eventloop-thread-1] WARN i.v.e.w.h.impl.LoggerHandlerImpl - 127.0.0.1 - - [Sun, 19 Nov 2023 17:22:22 GMT] "GET /�����hi=%D0%A0%D1%9F%D0%A1%D0%82%D0%A0%D1%91%D0%A0%D0%86%D0%A0%C2%B5%D0%A1%E2%80%9A HTTP/1.1" 404 53 "-" "Vert.x-WebClient/4.5.0"

Extra

• Windows 10/11, Java 21

[tsegismont]

With UriTemplate you can get an absolute URI that is properly encoded:

String absoluteURI = UriTemplate.of("{scheme}://{host}:{port}/{path}")
      .expandToString(Variables.variables()
        .set("scheme", "http")
        .set("host", "127.0.0.1")
        .set("port", 8881)
        .set("path", "привет"));

But on the server side encoded path segments are not supported anyway. Would you like to contribute this?

[lasteris]

With UriTemplate you can get an absolute URI that is properly encoded:

      .expandToString(Variables.variables()
        .set("scheme", "http")
        .set("host", "127.0.0.1")
        .set("port", 8881)
        .set("path", "привет"));

But on the server side encoded path segments are not supported anyway. Would you like to contribute this?

that sounds like an interesting idea🤗

[vietj]

yeah that's what URI template are made for

[lasteris]

yeah that's what URI template are made for

can I really give it a try to implement decoding path segments at the server side ?

[vietj]

I'm not sure what you want to do since we are talking here about web client and not server

[vietj]

ah you mean Thomas suggestion

[lasteris]

ah you mean Thomas suggestion

yeah, honestly i thought about server support from the start, because if webclient does not encode, why should the server have logic for decoding (cause it all in one ecosystem) ?
and now sounds interesting, maybe i could find the "path segments" resolving logic on server side and try out to implement decoding

[lasteris]

and by webclient side i am not sure that it should be implemented, because I think that on client side that's not a problem in any case if we have UriTemplate or UrlEncoder

[vietj]

sure, I'll close this issue in the meantime then