I found the following survey of keep-alive behaviors to be useful:
https://blog.cloudflare.com/when-tcp-sockets-refuse-to-die/

That post also covers the TCP_USER_TIMEOUT socket option, which configures another aspect of how dead TCP connections are detected. I'm not certain if anything other than Linux supports that option, so I've left it out of this proposal.

Presumably SetKeepAliveConfig would return a non-nil error if the configured keepalive behavior can't be provided on the current platform?

Presumably SetKeepAliveConfig would return a non-nil error if the configured keepalive behavior can't be provided on the current platform?

That was my thought. If a caller doesn't mind a platform not supporting an option, they can just not check the returned error.

Presumably SetKeepAliveConfig would return a non-nil error if the configured keepalive behavior can't be provided on the current platform?

Maybe we should return an error that wraps errors. ErrUnsupported from #41198

This proposal has been added to the active column of the proposals project
and will now be reviewed at the weekly proposal review meetings.
— rsc for the proposal review group

This seems reasonable, well-motivated, and well-justified. Thank you.
I think you can drop the "and KeepAliveConfig.Enable is false" in the
KeepAlive docs, since it already said the field is ignored when Enable==true.

It would help to say what the current defaults are in the config struct docs.

And in Count maybe "should be sent" should be "can go unanswered".

Updated proposal with the suggested changes.

I needed to pick default values for KeepAliveConfig fields. I went with the current behavior for Idle and Interval (15 seconds), and the Linux kernel default for Count (9). We could also say that the default for Count matches the current behavior of picking the OS default, but it seems less surprising for us to set all three.

package net

// KeepAliveConfig contains TCP keep-alive options.
//
// If the Idle, Interval, or Count fields are zero, a default value is chosen.
// If a field is negative, the operating system default is used.
type KeepAliveConfig struct {
        // If Enable is true, keep-alive probes are enabled.
        Enable bool

        // Idle is the time that the connection must be idle before
        // the first keep-alive probe is sent.
        // If zero, a default value of 15 seconds is used.
        Idle time.Duration

        // Interval is the time between keep-alive probes.
        // If zero, a default value of 15 seconds is used.
        Interval time.Duration

        // Count is the maximum number of keep-alive probes that
        // can go unanswered before dropping a connection.
        // If zero, a default value of 9 is used.
        Count int
}

type Dialer { // contains existing unchanged fields
        // KeepAlive specifies the interval between keep-alive
        // probes for an active network connection.
        //
        // KeepAlive is ignored if KeepAliveConfig.Enable is true.
        //
        // If zero, keep-alive probes are sent with a default value
        // (currently 15 seconds), if supported by the protocol and operating
        // system. Network protocols or operating systems that do
        // not support keep-alives ignore this field.
        // If negative, keep-alive probes are disabled.
        KeepAlive time.Duration

        // KeepAliveConfig specifies the keep-alive probe configuration
        // for an active network connection, when supported by the
        // protocol and operating system.
        //
        // If KeepAliveConfig.Enable is true, keep-alives probes are enabled.
        // If KeepAliveConfig.Enable is false and KeepAlive is negative,
        // keep-alive probes are disabled.
        KeepAliveConfig KeepAliveConfig
}

type ListenConfig {
        // same fields as Dialer above
}

// SetKeepAliveConfig configures keep-alive messages sent by the operating system.
func (c *TCPConn) SetKeepAliveConfig(config KeepAliveConfig) error {}

Based on the discussion above, this proposal seems like a likely accept.
— rsc for the proposal review group

Proposal details are in #62254 (comment).

No change in consensus, so accepted. 🎉
This issue now tracks the work of implementing the proposal.
— rsc for the proposal review group

Proposal details are in #62254 (comment).

I've done some investigations about TCP keep-alive on various platforms and had a CL for this proposal drafted, I'd volunteer to implement this if you don't mind.

Change https://go.dev/cl/542275 mentions this issue: net: add KeepAliveConfig and implement SetKeepAliveConfig

Thanks everyone for the discourse! Looks like it is too late in the cycle to get in, forward rolling to Go1.23.