Polka interprets percent-encoded slashes the same as regular slashes

I have an app with a query endpoint like /query/:input and sometimes the input can be something like 1/7.

I would have a url like /query/1%20%2F%207 (percent encoded 1 / 7) and end up getting a 404 page. I added a console.log call and it turned out this was being decoded into /query/1 / 7 in request.url. Express doesn't have this issue, for comparison.

I was using the next version from NPM because that was the default in the Sapper template.

Repro

const polka = require("polka");

polka()
  .use((req, res, next) => {
    console.log("url", req.url);
    next();
  })
  .get("/query/:input", (req, res) => {
    res.end(
      JSON.stringify({
        url: req.url,
        input: req.params.input,
      })
    );
  })
  .listen(3000, (err) => {
    if (err) throw err;
    console.log("> Running on localhost:3000");
  });

$ curl http://localhost:3000/query/1%20%2F%207

Expected behavior / Express behavior

200 OK with body:

{
  "url": "/query/1%20%2F%207",
  "input": "1 / 7"
}

Actual behavior (5.2.0)

200 OK with body:

{
  "url": "/query/1%20%2F%207",
  "input": "1%20%2F%207"
}

(input is urlencoded when it should be decoded)

Actual behavior (next)

Server prints url /query/1 / 7.

Curl sees 404 Not Found.

Start with the provided Polka reproduction and run the curl request against the /query/:input route. Trace how the request URL is decoded before routing, then verify that the encoded slash stays in req.url, input becomes "1 / 7", and the request returns 200 rather than 404.