Quick questions about using Hyper with HTTP/2

[OldMansKid]

Hello community,
I notice that http1::SendRequest doesn't implement Clone while http2::SendRequest does. On the other hand, hyper::service::Service isn't required to implement Clone. So, I assume that in HTTP/2, a single Service's call() function will be called with requests from all HTTP/2 streams, and cloning an http2::SendRequest will open a new stream, am I right?

I also notice that Hyper doesn't expose any APIs to operate on HTTP/2 pseudo headers nor on frames like sending RST_STREAM, while the h2 crate gives more control over these things. Does it mean that I need to use h2 instead of SendRequest/Builder APIs if I have such needs? Thanks in advance.

[seanmonstar]

I assume that in HTTP/2, a single Service's call() function will be called with requests from all HTTP/2 streams, and cloning an http2::SendRequest will open a new stream, am I right?

Not quite. In HTTP/2, sending any request will always result in a new stream. The reason the http1::SendRequest isn't Clone is because an HTTP/1 connection cannot send new requests until the previous one has finished. Whereas http2::SendRequest can be cloned to send multiple requests at the same time. (The Clone call doesn't create a new stream. It's the send_request() call.)

HTTP/2 pseudo headers

Most of them are translated to http types, either pieces of the Uri, or the StatusCode.

frames like sending RST_STREAM

If you need to deal with those frames directly, yes you'll need to look at the h2 frame. Though, hyper will use those protocol-specific frames by translating from the level that hyper exposes.