Jesse Wilson

I’m not sold on this. I think the top-level goal is reasonable, but this implementation isn’t consistent with our other `with` methods on interceptors – this collects all events; not...

Do we get around this by dropping support for jsLegacy?

I love this analysis. Thanks!

Is the best solution to make OkHttp lazy internally? Or to encourage our users to instantiate their OkHttp clients lazily? I think making OkHttp lazy potentially introduces other problems by...

Looks like `:` is permitted when the boundary is in a quoted-string. We need to fix OkHttp’s `MultipartBody`, to quote `boundary` when necessary. See [RFC 2045 section 5.1](https://datatracker.ietf.org/doc/html/rfc2045#section-5.1). Cloudfront needs...

@yschimke that’s a good idea. Or this? ``` class RequestBody { ... /** * Returns a non-null key if this request can be cached. If non-null, * this key is...

From [HttpUrlTest](https://github.com/square/okhttp/blob/master/okhttp/src/commonTest/kotlin/okhttp3/HttpUrlTest.kt#L536), // HttpUrl is quite lenient with what characters it accepts. In particular, characters like '{' // and '"' are permitted but unlikely to occur in real-world URLs. Unfortunately...

We’ve got [another relevant test](https://github.com/square/okhttp/blob/master/okhttp/src/jvmTest/java/okhttp3/HttpUrlJvmTest.kt#L30-L36): ```kotlin /** This one's ugly: the HttpUrl's host is non-empty, but the URI's host is null. */ @Test fun hostContainsOnlyStrippedCharacters() { val url = "http://>/".toHttpUrl()...

I think this makes more sense at Coil’s layer vs. OkHttp’s? One problem I have with guards like this is their data is necessarily incomplete. The only way to know...

Excellent!