Add closed property to OutgoingMessage.

[ronag]

Similar to HTTP2ServerResponse.closed. This is an alternative to 8589c70 which provides an API similar to h2 compat and won't break existing OSS projects.

Fixes #15523

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• tests and/or benchmarks are included
• documentation is changed or added
• commit message follows commit guidelines

Affected core subsystem(s)

[jasnell]

ping @nodejs/http @mcollina

[mcollina]

I'm -1 on this one. I still hope we can get writable = false in HTTP1, without breaking the ecosystem. It's worth spending time doing that, we should not add new properties lightly.

[ronag]

@mcollina: In that case, what do you think about removing closed from h2 compat? It should be essentially the same as writable? #15527

[mcollina]

I'm 👍 in removing closed from HTTP2 compat layer if it's not on on HTTP1

[apapirovski]

To be clear, we should still track closed internally but not expose it — correct? (It's useful for internal stuff.)

[mscdex]

If it's only used internally then it shouldn't be exposed to the public, use a Symbol or something instead.

[ronag]

before removing closed... do we have any way for the user to track it, e.g. writable with correct behavior (unlike h1)? #15527

[mcollina]

@ronag I don't know. At this point we aim for compatibility, so I guess we would have to reproduce the "flaws" as well.

[BridgeAR]

I think it is clear that this is not the way we want to move on and instead should just remove closed from http2 compat. Is everyone OK with closing this therefore?

[apapirovski]

I have a PR open for http2 that removes it so 👍 from me.