When should we drop the moratorium?

[MylesBorins]

If we are convinced that we will build off the minimal kernel and have an idea of what the future looks like at what point does it make sense to upstream?

The longer we allow the current implementation to continue to exist in the wild the more there will be code that relies on it.

We could also upstream to master / Current but not LTS

[devsnek]

i would say we should upstream as soon as we have functionality which matches or exceeds the current upstream

[GeoffreyBooth]

I concur with @devsnek. We’ll be well on our way (but probably not quite there yet) by the end of Phase 2.

If we think that upstreaming the new implementation is very likely to happen, and I don’t see why it wouldn’t, we could consider updating the Node docs now to add some warning that there’s a good chance that --experimental-modules might be replaced with a different implementation and public API in the near future.

[devsnek]

we could consider updating the Node docs now to add some warning that there’s a good chance that --experimental-modules might be replaced with a different implementation and public API in the near future.

fwiw, that's literally why node has experimental flags, so i don't think that's needed

[GeoffreyBooth]

I agree it shouldn’t be needed, but I think @jdalton can provide several anecdotes of having to guide people toward stopping using --experimental-modules in production or in public (non-experimental) build tools. I used it myself for a real-world application before joining this group, because I assumed it meant that the feature was in beta. Clearly the message isn’t getting across to a sizable number of users that they shouldn’t be using it. I don’t think being more explicit in the docs could hurt.

[mcollina]

I think we should send a PR to master from this repo as soon as possible, and ship it in v11. I think some people are settling in using --experimental-modules, and we should be breaking them as soon as possible.

I would recommend for the majority of the work to keep being done in this repository, and just do a PR once and then when there is consensus here.

[guybedford]

My main issue with shipping the minimal implementation as-is is that currently --loader opens up a lot of ways for users to tailor modules in Node to their own needs. Because this minimal implementation has removed --loader I feel that taking that away stops that sort of community experimentation, which is important.

[SMotaal]

I also agree with @guybedford on the --loader aspect. I wonder if we want to consider if the --loader flag can be made functional on top of the new implementation if and only if it is used with --experimental-modules flag.

I am not able to infer if the intent is to retain needing --experimental-modules when landing this unfinished minimal implementation but made the assumption, am I right?

[bmeck]

I think removing the moratorium makes sense and we can begin upstreaming things that are not expected to change. I agree that we need --loader before unflagging, but want to do a fairly large rewrite of the loader instrumentation mechanisms so I would not want to ship it as it exists currently. Given that expectation of change I would prefer to not upstreams with a --loader implementation until we have something that is not expected to change.

[MylesBorins]

Closing for now. We can revisit a bit later when we have a clearer vision