[breaking change] Implement HTTP/1/2-only support #130
Conversation
Me either. |
| /// Returns a mutable reference to the Http builder. | ||
| pub fn http_builder(&mut self) -> &mut Builder<TokioExecutor> { | ||
| &mut self.builder |
There was a problem hiding this comment.
I'd like to see this API remain available, as I use it to configure many settings for both HTTP/1 and HTTP/2 👀
There was a problem hiding this comment.
The idea here is to expose anything Builder exposes through axum-servers interface.
What method exactly are you unable to access with the new API?
There was a problem hiding this comment.
Hmm, I don't see how I would access these:
serve
.http_builder()
.http1()
.timer(hyper_util::rt::TokioTimer::new())
.keep_alive(true)
.header_read_timeout(Duration::from_secs(5))
.max_buf_size(32768)
.http2()
.timer(hyper_util::rt::TokioTimer::new())
.keep_alive_interval(Duration::from_secs(16))
.keep_alive_timeout(Duration::from_secs(5))
.max_header_list_size(512 * 1024)
.max_send_buf_size(64 * 1024)
.max_concurrent_streams(16);There was a problem hiding this comment.
Ah yes, I assume if we expose Builder::http1() and Builder::http2() this would cover all the methods you have listed?
The PR should probably be in draft mode.
There was a problem hiding this comment.
I have an intermediate solution that does not breaks compatibility that can be used, see #156. Let me know what you think.
There was a problem hiding this comment.
Glad it doesn't break compatibility. Still don't personally understand the API or the use-case but you have to convince the maintainer, not me ;)
This PR changes dependencies to be more careful about the default features enabled.
Now HTTP/1 and HTTP/2 are enabled by default but can be disabled if so desired.
There is also now a way to set the server to only enable one of the two protocols.
This would re-open #80 and I have no clue how to solve that without a major redesign of the public API.
The only easier way that comes to mind is cloning the
ServerConfigand then modifying the ALPN if it doesn't match while handling theAcceptor.I will leave this open until the next version upgrade.