discuss: remove domain module

[bnoordhuis]

• require('domain') was deprecated 7.5 years ago

• modern code uses AbortController

• bug tracker activity suggests it's not in active use (or it works so well people never hit bugs⁰ - unlikely!)

• the supporting infrastructure for domain is a kind of technical debt that also has performance implications¹; it's not as bad as it used to be but it's still there

Proposal:

• remove it for good in the next major release

• leave a stub lib/domain.js that prints a warning, a la the sys module.

---

⁰ the prime reason for deprecation was because it was so buggy due to part poor design, part poor implementation

¹ at one time it was so bad that if even one module imported domain, whole-program performance tanked

[MoLow]

a good idea would be to test how this will effect citgm

[marco-ippolito]

this is the previous discussion opened more than 5 years ago: #10843
Other discussion about removal: #17143

Domains were deprecated in node v4, after all. As of v13, still no answer as to what to use instead.

It seems that it is something that has been around for quite some time now, I think is time to remove domain.

[bnoordhuis]

They've been deprecated even longer, the deprecation notice was already there in the first io.js release! (ref. 6a29356)

I'll open a trial PR and see how it fares.

[bnoordhuis]

https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/3060/ - honestly, there are so many random test failures (also in other, unrelated citgm runs) that it's hard to say whether it's indicative of anything.

FWIW, I didn't find any mentions of the domain module while randomly clicking through some of the failing test suites.

[targos]

There are a few (all I saw come either from mocha or tap):
https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/3060/nodes=fedora-last-latest-x64/testReport/junit/(root)/citgm/bluebird_v3_7_2/
https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/3060/nodes=fedora-last-latest-x64/testReport/junit/(root)/citgm/browserify_v17_0_0/
https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/3060/nodes=fedora-last-latest-x64/testReport/junit/(root)/citgm/gulp_v4_0_2/
https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/3060/nodes=fedora-last-latest-x64/testReport/junit/(root)/citgm/node_gyp_v9_3_0/
https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/3060/nodes=fedora-last-latest-x64/testReport/junit/(root)/citgm/rimraf_v3_0_2/
... and more with tap

[bnoordhuis]

Ah, thanks. Okay, so that means a runtime deprecation first before the hammer comes down.

[ORESoftware]

I for one, vote to keep Domains - I use them for multiple use cases - is there a clear replacement for the functionality? Until there is a clear replacement, I vote to keep in core. Make something better, before just removing it. Test libraries like tap and others use domains, and I use domains in production.

In other words: The domain API is simple to use, intuitive, and works, however it is a bit slow in performance. I would like AsyncLocalStorage (or whatever replaces Domain) to be beyond "experimental" before removing Domains when there is no replacement for the Domain functionality (which is critical for some people who use in production).

For example middleware similar to this can improve production servers:

app.use((req,res,next) => {
    
    const d = Domain.create();
    
    d.once('error', e => {
       res.status(500).send(e.message);
    });
    
    d.run(next);

});

if there is a new api that can what the above does, and improve performance, I am all for it. 👍
However AsyncLocalStorage is still experimental?

For example:
https://chat.openai.com/share/2736160d-2f1c-4663-b7e7-758d6a922a2c

[lroal]

Keep Domain. The AsyncLocalStorage and async_hooks just aren't reliable.
However, I see there is a TC39 proposal in stage 2 for AsyncLocalStorage

[Flarna]

The AsyncLocalStorage and async_hooks just aren't reliable.

FWIW domain uses async hooks internally so I doubt it's more reliable.