<script context="module"> -> <script module>

[Rich-Harris]

Describe the problem

We've been meaning to make this change for a while, we should make sure it happens for 5.0

Describe the proposed solution

Deprecate the old syntax, introduce the new, and update the migration script

Importance

nice to have

[dominikg]

are we going to reserve more attribute names on script or style tags? preprocessors can technically use them so defining new ones could be a breaking change.

[Rich-Harris]

Good point. Without making any commitment to use any of these in future, we could reserve client, server, test... any others?

[trueadm]

worker might be another one.

[dominikg]

another option could be to claim all attributes that are not using an xmlns namespace id to show which third party tool/package they originate from.

The only current use that comes to my head is svelte-preprocess using src to allow multi-file svelte components 🙈

[Inqnuam]

@dominikg I'm developing a meta-framework based on Svelte, where pathname attribute is used in pre-processor:

<script context="module" pathname="/home">
</script>

I hope this will still be possible in Svelte 5

[Conduitry]

Is <script $module> too ugly to consider? Could we just reserve everything with a given prefix?

[Rich-Harris]

oof, I think I'd rather keep context than do that tbh

[jackvanderbilt-visma]

The nomodule attribute is already a thing. This one and other one's you guys are suggesting to already reserve might collide with the spec someday... maybe? who knows...

[7nik]

The nomodule attribute is already a thing. This one and other one's you guys are suggesting to already reserve might collide with the spec someday... maybe? who knows...

Technically, the HTML's script and Svelte's top-level scripts are different things with unrelated sets of attributes. So it's not a problem at all.

[Rich-Harris]

Yeah, the only breakage we potentially need to worry about is people using their own attributes, like the pathname example above. But by that reasoning we wouldn't be able to introduce any new attribute (like lang="ts" or generics="T") outside a major release, and I don't know that we'd be willing to accept that. To me this falls into a bit of a semver grey area, and I think the likelihood of collisions (outside reserved attributes that we anticipate) is low enough that we can just say 'adding your own attributes to <script> is not supported and prone to breakage, but on your own head be it' and look the other way.

[Autumnlight02]

Could this not be solvable with an additional plugin to convert that before the svelte compiler get's his hands on it? only issue with that would be the language server.