Deprecate dbus-native

[acrisci]

Thanks for your work creating dbus-native.

I have created a fork of the project called dbus-next with many bugfixes and enhancements that I believe fixes many of the issues on your issue tracker as well as the following:

• fully featured high-level and low-level interfaces
• a more intuitive type system
• comprehensive documentation for the public api
• integration tests for all library features
• used in six electron media players through the mpris-service library.

I was wondering if you would deprecate this library and send your users my way. I am also open to merging the projects.

Thanks.

[sidorares]

hey @acrisci , thanks for your efforts ( I saw your email but didn't have time to respond, I'm sorry )

Happy to see deprecate my version in favour of yours, or merge and move to https://github.com/dbusjs , whatewher you think is better

[sidorares]

At current stage, is your code backwards compatible? Do you want to re-use npm name? If yes, whe can 1) mifrate "official" repo to dbusjs org 2) I'll add you publish rights to npm and you publish latest version of your code

[acrisci]

At current stage, is your code backwards compatible?

Unfortunately no. I rewrote everything except the marshallers.

Do you want to re-use npm name?

I think this would be good because dbus-native has very high visibility.

If yes, whe can 1) mifrate "official" repo to dbusjs org 2) I'll add you publish rights to npm and you publish latest version of your code

This sounds good to me.

[sidorares]

Unfortunately no. I rewrote everything except the marshallers.

I guess with enough of communication and major version change this is still possible

[acrisci]

With the latest 0.7.1 release of dbus-next, I think I'm ready to make the api stable so I'm ready to take action on this now.

[sidorares]

What was agreed: dbus-next would take over the npm name. That never happened, and dbus-next has itself been dormant since 2022.

Where the users actually are:

package	latest	published	downloads/week
@homebridge/dbus-native (soft fork of this repo)	0.7.8	2026-07	~41k
dbus-next	0.10.2	2022-04	~19k
dbus-native (this repo)	0.11.0	2026-07	~7.4k

Roughly nine in ten users of this codebase consume it through the Homebridge fork, which is a light fork of this repo — they dropped abstract-socket, swapped optimist for minimist, and vendored long.js for an ARMv6 crash.

All three of those deltas are now resolved upstream: abstract-socket is gone (Node ≥ 20.8 does abstract sockets natively), optimist is gone (node:util.parseArgs), and the long.js workaround disappears when BigInt becomes the default (#248) — it is already available opt-in. Runtime dependencies here went from seven to three, against their six.

Since this issue was filed, this repo has also gained promises, timeouts and AbortSignal, real Error objects, TypeScript declarations, diagnostics_channel instrumentation, big-endian support, a ~2000× faster ay write path, property access control and PropertiesChanged, signal introspection, name validation, and a launchd transport for macOS.

ROADMAP §0's recommendation is therefore do not deprecate, and instead close this with a statement of direction, plus talk to the Homebridge maintainers about folding their fork back in — their fork exists because upstream went quiet, and that premise no longer holds.

Leaving it open because it is a positioning call, not a technical one.

[sidorares]

An update six years late, and an apology for the ambiguity in between — an open "should this be deprecated?" issue is itself a signal, and it was the wrong one for a long time.

The decision: work has restarted here, and this package is not being deprecated.

The npm handover agreed in 2019 never happened, and dbus-next has been dormant since April 2022. Rather than leave three half-maintained packages, I've picked this one back up.

Progress since this issue was filed — 0.4.0 (2018) to 0.14.0, most of it in the last few days:

• The type system. Variants read as their values, string-keyed dicts as plain objects, x/t as bigint. That closes Dict of string, variant -> hash map #3, open since 2013, and [proposal] Replace Long.js with native BigInt #248.
• A high-level API. bus.proxy(), bus.objects() backed by ObjectManager, defineInterface(), scoped signal subscriptions — with the low-level invoke surface unchanged underneath.
• Documentation and tests. A documented public API, a migration guide per breaking release, and an integration suite that runs against a real dbus-daemon in CI across Node 22/24/26.
• The wire layer. Promises and AbortSignal, real Error objects, a 25s default call timeout, big-endian reads, UNIX_FD, TypeScript declarations, diagnostics_channel, and a working server side.
• One runtime dependency, down from seven.

@acrisci — thank you for dbus-next, and for maintaining it as long as you did. Every criticism in your opening comment was fair when you wrote it, and closing those gaps is most of what this cycle was.

Leaving the issue open for now in case anyone wants to weigh in.