This adds HTTP2 keep-alive support to client and server connections
based losely on GRPC keep-alive. When enabled, after no data has been
received for some configured interval, an HTTP2 PING frame is sent. If
the PING is not acknowledged with a configured timeout, the connection
is closed.
Clients have an additional option to enable keep-alive while the
connection is otherwise idle. When disabled, keep-alive PINGs are only
used while there are open request/response streams. If enabled, PINGs
are sent even when there are no active streams.
For now, since these features use `tokio::time::Delay`, the `runtime`
cargo feature is required to use them.
- Renamed `keep_alive_timeout` to `pool_idle_timeout`.
- Renamed `max_idle_per_host` to `pool_max_idle_per_host`.
- Deprecated `keep_alive(bool)` due to confusing name. To disable the
connection pool, call `pool_max_idle_per_host(0)`.
This adds support for calculating the Bandwidth-delay product when using
HTTP2. When a DATA frame is received, a PING is sent to the remote.
While the PING acknoledgement is outstanding, the amount of bytes of all
received DATA frames is accumulated. Once we receive the PING
acknowledgement, we calculate the BDP based on the number of received
bytes and the round-trip-time of the PING. If we are near the current
maximum window size, the size is doubled.
It's disabled by default until tested more extensively.
When the body type of a `Request` or `Response` implements `HttpBody`,
the `Request` or `Response` itself now implements `HttpBody`.
This allows writing things like `hyper::body::aggregate(req)` instead of
`hyper::body::aggregate(req.into_body())`.
Closes#2067
This is *just* a "trait alias". It is automatically implemented for all
`Service<Uri>`s that have the required bounds. It's purpose being public
is to ease setting trait bounds outside of hyper. Therefore, it doesn't
have any exposed associated types, to prevent otherwise relying on any
super-traits that hyper requires.
Instead of returning a tuple `(impl AsyncRead + AsyncWrite, Connected)`,
this adds a new trait, `hyper::client::connect::Connection`, which
allows querying the connection type for a `Connected`.
BREAKING CHANGE: Connectors no longer return a tuple of
`(T, Connected)`, but a single `T: Connection`.
It is not possible to connect to an IPv4 address from an IPv6 address or
vice-versa so don't waste time trying. If no remote addresses match then a
"missing connect error" will now occur.
Closes#1903
BREAKING CHANGE: The `Resolve` trait is gone. All custom resolves should
implement `tower::Service` instead.
The error type of `HttpConnector` has been changed away from
`std::io::Error`.
The `Connect` trait is now essentially an alias for
`Service<Destination>`, with a blanket implementation as such, and is
sealed.
Closes#1902
BREAKING CHANGE: Any manual implementations of `Connect` must instead
implement `tower::Service<Destination>`.
The only important trait for a user is the `tower::Service` trait, which
is now available also at `hyper::service::Service`. The other "trait
aliases" are no longer publicly exported, as people thought they had to
implement them.
Also removes dependency on `tower-make`, which is trivial but otherwise
shouldn't affect anyone.
Closes#1959
This takes the same strategy as golang, where the timeout value is
divided equally between the candidate socket addresses.
If happy eyeballs is enabled, the division takes place "below" the
IPv4/IPv6 partitioning.
The `HttpConnector` and `AddrListener` types which make use of
`tokio::tcp` have been made their own optional feature. This allows
using them without requiring the *full* tokio runtime.
