Intent to Remove: Multiple Shadow Roots for Shadow DOM V0

[Hayato Ito]

Primary eng (and PM) emails

hay...@chromium.org, ko...@chromium.org

Link to “Intent to Deprecate” thread

https://groups.google.com/a/chromium.org/forum/#!msg/blink-dev/9qDsRePDALE/7c8kOkJfLgkJ

Summary

Remove the support of multiple shadow roots of Shadow DOM v0.

Multiple shadow roots feature enables users to put more than one shadow roots on a shadow host.

See this for an explainer of this feature.

This is a step to deprecate/remove all Shadow DOM v0 features, in favor of Shadow DOM v1.

Given that Blink now supports Shadow DOM v1 as well as Shadow DOM v0, multiple

shadow roots feature has been one of the biggest causes of code complexity. It

would be a huge win if we can remove this feature from Blink, from the DOM code

health's perspective.

Specifically:

- ShadowRoot#olderShadowRoot: Will remove this from IDL.

- Element#createShadowRoot() on a shadow host: Will throw an error.

Compatibility And Interoperability Risk

Edge/IE: Not supported

Firefox: Not supported

Safari: Not supported

Alternative implementation suggestion for web developers

No alternative.

Usage information from UseCounter

0.018% for ElementCreateShadowRootMultiple

https://www.chromestatus.com/metrics/feature/timeline/popularity/779

It looks the usage decreased significantly at Apr 2017.

OWP launch tracking bug

https://crbug.com/489947

Entry on the feature dashboard

https://www.chromestatus.com/features/4668884095336448

[Rune Lillesveen]

In general non-owner lgtm with some questions in-line.

On Mon, Aug 21, 2017 at 2:04 PM, Hayato Ito <hay...@chromium.org> wrote:

Primary eng (and PM) emails

hay...@chromium.org, ko...@chromium.org

Link to “Intent to Deprecate” thread

https://groups.google.com/a/chromium.org/forum/#!msg/blink-dev/9qDsRePDALE/7c8kOkJfLgkJ

Summary

Remove the support of multiple shadow roots of Shadow DOM v0.

Multiple shadow roots feature enables users to put more than one shadow roots on a shadow host.

See this for an explainer of this feature.

This is a step to deprecate/remove all Shadow DOM v0 features, in favor of Shadow DOM v1.

Given that Blink now supports Shadow DOM v1 as well as Shadow DOM v0, multiple

shadow roots feature has been one of the biggest causes of code complexity. It

would be a huge win if we can remove this feature from Blink, from the DOM code

health's perspective.

Specifically:

- ShadowRoot#olderShadowRoot: Will remove this from IDL.

- Element#createShadowRoot() on a shadow host: Will throw an error.

What's the plan for the <shadow> element? That will now be useless, right?

Compatibility And Interoperability Risk

Edge/IE: Not supported

Firefox: Not supported

Safari: Not supported

The compat risk would mostly be with Blink targeted shadow dom v0 applications using multiple roots?

What's the typical feature detection for shadow dom v0?

Did earlier versions of Polymer use multiple roots?

Alternative implementation suggestion for web developers

No alternative.

Usage information from UseCounter

0.018% for ElementCreateShadowRootMultiple

https://www.chromestatus.com/metrics/feature/timeline/popularity/779

It looks the usage decreased significantly at Apr 2017.

Is that older versions of Polymer contributing most of that?

--

Rune Lillesveen

[Hayato Ito]

Replied in-line.

On Mon, Aug 21, 2017 at 10:55 PM Rune Lillesveen <ru...@opera.com> wrote:

In general non-owner lgtm with some questions in-line.

On Mon, Aug 21, 2017 at 2:04 PM, Hayato Ito <hay...@chromium.org> wrote:

Primary eng (and PM) emails

hay...@chromium.org, ko...@chromium.org

Link to “Intent to Deprecate” thread

https://groups.google.com/a/chromium.org/forum/#!msg/blink-dev/9qDsRePDALE/7c8kOkJfLgkJ

Summary

Remove the support of multiple shadow roots of Shadow DOM v0.

Multiple shadow roots feature enables users to put more than one shadow roots on a shadow host.

See this for an explainer of this feature.

This is a step to deprecate/remove all Shadow DOM v0 features, in favor of Shadow DOM v1.

Given that Blink now supports Shadow DOM v1 as well as Shadow DOM v0, multiple

shadow roots feature has been one of the biggest causes of code complexity. It

would be a huge win if we can remove this feature from Blink, from the DOM code

health's perspective.

Specifically:

- ShadowRoot#olderShadowRoot: Will remove this from IDL.

- Element#createShadowRoot() on a shadow host: Will throw an error.

What's the plan for the <shadow> element? That will now be useless, right?

We won't remove the support of <shadow> elements at this timing because:

- Using <shadow> elements in a single shadow root is a valid usage, where <shadow> elements should behave like <content> elements (except that <shadow> doesn't support select attributes). 

- <shadow>'s usage is still high; 1.93% (https://www.chromestatus.com/metrics/feature/timeline/popularity/1897).

I guess we can remove <shadow> elements together with entire Shadow DOM v0 APIs.

Compatibility And Interoperability Risk

Edge/IE: Not supported

Firefox: Not supported

Safari: Not supported

The compat risk would mostly be with Blink targeted shadow dom v0 applications using multiple roots?

Yes, the situation is very similar to the case of /deep/ combinator.

 

What's the typical feature detection for shadow dom v0?

We are using Element#createShadowRoot for tracking; https://www.chromestatus.com/metrics/feature/timeline/popularity/456.

 

Did earlier versions of Polymer use multiple roots?

Yes, as far as I remember, there was only *one* custom element which uses multiple shadow roots feature in earlier versions of Polymer.

 

Alternative implementation suggestion for web developers

No alternative.

Usage information from UseCounter

0.018% for ElementCreateShadowRootMultiple

https://www.chromestatus.com/metrics/feature/timeline/popularity/779

It looks the usage decreased significantly at Apr 2017.

Is that older versions of Polymer contributing most of that?

Possibly, but we don't have data. We guessed that it might be due to a Chrome Extension, such as AdBlock.

I noticed this usage drop just today.

 

--

Rune Lillesveen

[TAMURA, Kent]

LGTM to remove.

The current deprecation message doesn't have the target removal milestone. So I recommend to add the target milestone to the deprecation message for M62 release, and remove the feature in M63 or M64 just in case.

--
You received this message because you are subscribed to the Google Groups "blink-dev" group.
To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAFpjS_1hXkp286Qqj2dq%2B3Keiux4CjFb5m%3DGJGPBjrBd0c5e0Q%40mail.gmail.com.

--

TAMURA Kent
Software Engineer, Google

[Hayato Ito]

Yeah, that is my plan. I'll add the target milestone to the deprecation message after I get 3 LGTMs.

[Jochen Eisinger]

LGTM2

To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAFpjS_225fUBJHGGE_QwGRR0e3-VAtjkY4mnGBfV7G%2BNcVfpQg%40mail.gmail.com.

[Dimitri Glazkov]

lgtm3

To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CALjhuifmq8KMrpoC0VFaCGdR_wSNRhhMMdi5JWgPDbz-QShsaQ%40mail.gmail.com.

[Rick Byers]

From our experience with the /deep/ removal the case that worries me most is sites which do feature detection for ShadowDOM v0 and fall back to a polyfill when that fails.  Such sites may currently work in all browsers today, but could conceivably break in Chrome when it supports some but not all of the ShadowDOM v0 features.

Hayato, can you verify whether some popular release of Polymer have this property for multiple shadow roots (as they do for /deep/)?  

From internal data it's interesting that this UseCounter seems to be hit exclusively on desktop platforms, it's almost 0 on Android.  Any idea why?  Perhaps there's some popular desktop-specific site dominating the metrics here?

What's the typical feature detection for shadow dom v0?

We are using Element#createShadowRoot for tracking; https://www.chromestatus.com/metrics/feature/timeline/popularity/456.

 

Did earlier versions of Polymer use multiple roots?

Yes, as far as I remember, there was only *one* custom element which uses multiple shadow roots feature in earlier versions of Polymer.

 

Alternative implementation suggestion for web developers

No alternative.

Usage information from UseCounter

0.018% for ElementCreateShadowRootMultiple

https://www.chromestatus.com/metrics/feature/timeline/popularity/779

It looks the usage decreased significantly at Apr 2017.

Is that older versions of Polymer contributing most of that?

Possibly, but we don't have data. We guessed that it might be due to a Chrome Extension, such as AdBlock.

I noticed this usage drop just today.

 

--

Rune Lillesveen

--
You received this message because you are subscribed to the Google Groups "blink-dev" group.

To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAFpjS_3EN1FVpcAk_P8t83eL8zS4Q6nRzROwUUbF2qZg%2B4e2DQ%40mail.gmail.com.

[Hayato Ito]

Polymer 0.5 is the last version which uses /deep/ and multiple shadow roots internally, as far as I know.

That means once we successfully remove /deep/, there shouldn't be an issue in removing multiple shadow roots, regarding sites using Polymer.

We are removing /deep/ at M63.

 

From internal data it's interesting that this UseCounter seems to be hit exclusively on desktop platforms, it's almost 0 on Android.  Any idea why?  Perhaps there's some popular desktop-specific site dominating the metrics here?

Though I don't know any good way to isolate the effect, it is known that the popular Chrome Extensions, AdBlock or AdBlock Plus, use Shadow DOM v0, and they might contribute the desktop usage of multiple shadow roots. They might call Element.createShadowRoot() for an element even if it has already a shadow root.

I guess the most usages of multiple shadow roots today are not *intentional* for users. They would just *overwrite* the old shadow root unintentionally without using the future of multiple shadow roots.

[Rick Byers]

Thanks, LGTM to me as well (with a targetted deprecation message for at least one milestone).

[Takayoshi Kochi]

Thanks for all LGTMs!

The target milestone (M63) is now included in the message.

https://chromium-review.googlesource.com/c/chromium/src/+/640432

--

Takayoshi Kochi

[Takayoshi Kochi]

Notice: The removal of multiple shadow roots slipped from M63 to M64, and it was removed as of r512017.

--

Takayoshi Kochi