I think it is fine keeping it here – will measure time to first logical block, no matter if any additional logic like `_redirects` was applied or not.

nit: We try to avoid adding new dependencies if we can avoid it. This is mostly for security reasons. This one is small enough that we could remove this dependency and add necessary `urlpath.Match` logic to a subpackage in https://github.com/ipfs-shipyard/go-ipfs-redirects, so no third-party libraries are needed.

nit: switch to https://github.com/ipfs-shipyard/go-ipfs-redirects

nit: unsure about duplicating all this – the only difference between NonUnixfs and Unixfs versions is special handling in "default" blocks in the Unixfs one. Could we have a single `handlePathResolution` that takes additional arg `responseFormat` and do the conditional `isUnixfsResponseFormat(responseFormat)` in the single, shared `default` block instead?

We need to be extra careful here and protect `_redirects` functionality against regressions caused by future refactors (we are [planning to extract gateway code into standalone lib](https://github.com/ipfs/kubo/issues/8524#issuecomment-1195788188), so we know it is imminent). To guard against shipping unsupported behaviors, this PR needs tests for negative behaviors: - [ ] expect HTTP 500 when `_redirects` is triggered but contains a status code that is not supported (see my note on the https://github.com/ipfs/specs/pull/290 that we should be explicit about which codes are supported) - [ ] expect HTTP 500 when `_redirects` is invalid (just put `hello` in a text file and write test that confirms it produces human-readable error) - [ ] expect HTTP 500 when `_redirects` is invalid (includes `!` "forced" directive → should produce a human-readable error informing that forced redirects (aka shadowing) are not supported)

nit: include `Location: ` header name in the match

nit: include `Location: ` header name in the match

nit: confirm `301` redirects return valid destination in `Location` header (add one more `test_should_contain`)

nit: I think you need to re-do the CAR to fix `:splat` line in `_redirects` fixture – details in my comment at https://github.com/ipfs/specs/pull/290#discussion_r942868437

nit: add similar checks for `410` and `451`, (to return correct error code, call `webError` func with `http.StatusGone` and `http.StatusUnavailableForLegalReasons`)

This loop should error when unsupported code is used in `_redirects` file.

nit: check if `rule.Status >= 301` and `rule.Status <= 308` and return redirect only if in this range of codes

@justincjohnson mind making me an Admin (Owner?) in that repo? Seems only that level can move it between orgs. (sadly under the water and I'll be reviewing things after Iceland, but we could move the repo before, so we have one thing less on our plate)

Addressing new linting errors in CI

Mind creating a CAR that has the test dir, and using it instead of constructing it on the fly? We have some prior art in `test/sharness/t0118-gateway-car/carv1-basic.car` That way we could use its root CID in the ipfs/specs RFC under "Test Fixtures" section, ensuring other implementation of IPFS test against things we want.