Serviceworker: Virtual calls around TPAC time

[jakearchibald]

I'd like to set up some calls to cover particular topics around TPAC time. I'll take a look through the issues and propose topics, but if there's a particular issue you'd like to bring up, please let me know.

@youennf, it feels like we should chat about background fetch and try to come to a consensus there.

@annevk @wanderview @asutherland @asakusuma @mfalken @yutakahirano @youennf @jungkees anything spring to mind?

20-21 Oct, 2020 14:00-16:00 UTC

See the TPAC group meeting schedule for your schedule management.

Meeting notes

Meetings are done! Thank you for your participation. Sorry that we couldn't cover all the topics in the agenda. See the meeting notes.

Agenda

20 Oct (Tue)

• Progress updates - A round-table of what we're up to service worker-wise, and what bandwidth/hunger we have for additional features.
  
  
  
  • Finding out current thinking from Mozilla and Apple on implementing navigation preload. The single most important performance win for service workers heard from devs.
  
  
• Declarative routing: https://github.com/w3c/ServiceWorker/issues/1373#issuecomment-451436409
• Clear-Site-Data: https://github.com/w3c/ServiceWorker/pull/1506
• Background activity and privacy - Resolving the issues @youennf posted around background fetch, which may have implications for push notifications.

21 Oct (Wed)

• Cache metadata - Proposal.
• Client lifecycle with bfcache, frozen tabs - Revisit and close on the frozen document discussion: Add an option to include frozen documents., consider how client API should behave with bfcache, How are unloaded tabs represented?
• Testing to exit CR and recharter
• Service workers on opaque origins - Discuss VS Code's use case on Registering Service Workers for unique origins?
• The timing of header setting in fetch - Sometimes headers are set pretty late, is this something we should/could change?

Not covered

• Fix worker timing - Right now it isn't well described, can we salvage it?
• Determine expected parameter values in Navigation + Resource Timing APIs for responses that have been synthesized using a Service Worker: https://github.com/w3c/navigation-timing/issues/124
• Ye old CSS CORS issue - https://github.com/w3c/ServiceWorker/issues/719
• The proposal for ServiceWorkerRegistration.id in #1512. There is a bit of contention on the issue. @wanderview would like to hear wider input to try to resolve things.
• whether cross-origin-isolated value can be platform dependent or not - https://github.com/w3c/ServiceWorker/pull/1545

[asakusuma]

I'd love to see progress on https://github.com/w3c/ServiceWorker/issues/1373#issuecomment-451436409. Are we at a point where a discussion would help?

Not sure if a discussion is actually needed, or if someone just needs to execute, but we still haven't actually updated the spec to reflect the Clear-Site-Data changes that were made in Chromium:
https://github.com/w3c/ServiceWorker/pull/1506
https://github.com/w3c/ServiceWorker/pull/1440

[youennf]

@youennf, it feels like we should chat about background fetch and try to come to a consensus there.

Yes, this might also have potential implications for push notifications as well.

[wanderview]

@jungkees Do you think MS would like to discuss this cache API proposal at all?

https://github.com/MicrosoftEdge/MSEdgeExplainers/blob/master/CacheAPIResponseMetadata/explainer.md

[jungkees]

Yes, @wanderview. Thanks for bring it up. We'd love to put it on the agenda if folks are interested. In addition to the feature details in the proposal, I wish to also hear feedback around whether it'd be useful for devs. @asakusuma @n8schloss @aarongustafson

[yoavweiss]

I'd love to discuss https://github.com/w3c/ServiceWorker/issues/1542 and potential solutions, if there's still room on the agenda

[mfalken]

Some issues that come to mind:

• https://github.com/w3c/resource-timing/issues/119
• https://github.com/w3c/ServiceWorker/issues/719

[jakearchibald]

I've captured the topics in the OP. Please let me know what topics you're interested in, and what timezone you're in, and I'll get schedulin'.

[makotoshimazu]

Thanks Jake for organizing!
I'm specifically interested in updates, background activity, header's timing, and fix worker timing. I'm in JST (UTC+9).

[annevk]

All, except for declarative routing/cache metadata. CET. Thanks Jake!

[asutherland]

All. EDT. Thank you kindly.

[wanderview]

All, EDT. Thanks.

[jungkees]

All, PDT. Thanks, @jakearchibald! Please add @erik-anderson and @SteveBeckerMSFT from MSFT to the invitee list (all PDT).

[mfalken]

I'm interested in at least updates & the two topics I proposed. I'm in JST.

[tomayac]

All, CET (CEST until Oct. 25). Thank you!

[jeffposnick]

👋 I'm specifically interested in the following, and am in EDT:

• Cache metadata
• Declarative routing
• clear-site-data

[makotoshimazu]

Is there a room to add one more topic regarding PerformanceResourceTiming body/transfer sizes? https://github.com/w3c/navigation-timing/issues/124

[jungkees]

Thanks @makotoshimazu. I added it to the Topics section and also added two more topics: frozen tabs and registering SW on opaque origins. I'll sync with @jakearchibald and schedule our meeting(s).

[jungkees]

I updated the OP with the proposed time and the collected agenda topics. I'm working with Jake to get this arranged as part of the TPAC plan. Please let me know if you have any comments or suggestions.

[jakearchibald]

Thanks for stepping in to organise this @jungkees!

There are some topics that might need more time, but we can identify those and organise additional calls with the folks that are interested.

[wanderview]

Sorry for the late request, but if there is time I wouldn't mind:

1. Finding out current thinking from Mozilla and Apple on implementing navigation preload. I continue to hear from partners that this is the single most important performance win for service workers.
2. Discussing the proposal for ServiceWorkerRegistration.id in #1512. There is a bit of contention on the issue and I would welcome wider input to try to resolve things.

[jungkees]

Thanks @wanderview. I updated the agenda. I think we can talk about the navigation preload as part of the progress update item. For ServiceWorkerRegistration.id, I put it as the last one in the 2nd day. Hope we can cover this. Otherwise, let's arrange a separate call for items later.

[makotoshimazu]

Hi, sorry about the last minute request.
If possible, can you add https://github.com/w3c/ServiceWorker/pull/1545 to the items on day 2?
This is a discussion about whether cross-origin-isolated value can be platform dependent or not because the value might largely depend on internal process allocation logic.

[wanderview]

Are we going to use this issue to schedule the 3rd meeting?

[jakearchibald]

@wanderview yes! I'll start organising that next week. Chrome Dev Summit stuff has owned all my time recently.

[mfalken]

I spoke with Jake. We are planning to hold the next meeting in November. To help decide the day, please fill in this Doodle:
https://doodle.com/poll/gn4ew5n3eni3kyik?utm_source=poll&utm_medium=link

Thanks!

[wanderview]

Could we use something other than zoom for the next meeting? The web interface for zoom is difficult to use for large meetings and we're not allowed to install the native app on our corp devices.

[jakearchibald]

Anyone object to using Google Meet?

[annevk]

I prefer Zoom as it makes it easier to not share your personal space, but still be present.

[jakearchibald]

@annevk fwiw, you can use https://snapcamera.snapchat.com/ to apply those effects to any site, as it acts like a virtual webcam. I think your kids would love it too!

[wanderview]

You can blur your background in google meet: https://support.google.com/meet/answer/10058482?hl=en

In zoom web I can only see a single participant at a time which makes it harder to take part in the discussion. Zoom chrome app supports more layouts, but I cannot get my microphone to work with it. As I said above, I cannot install zoom native.

[asutherland]

You can blur your background in google meet: https://support.google.com/meet/answer/10058482?hl=en

This doesn't seem to work in Firefox Nightly? The page you link seems to suggest this is Chrome specific. It does link to https://webglreport.com/?v=2 and my Firefox Nightly on a Macbook supports Web GL 2.0 and has "Major Performance Caveat: No" which suggests that's not the problem. I also tried in Safari and the background magic didn't seem to be available there, but it appeared my Safari didn't support Web GL 2.0.

[wanderview]

Sorry for that. Its a new feature and I was mainly skimming the doc to make sure it had been released.

[wanderview]

This doesn't seem to work in Firefox Nightly?

Kind of off topic at this point, but in case anyone is wondering, the meet blur feature depends on wasm simd which is only shipped in chrome at the moment. While firefox nightly does support wasm simd, there are some known interop issues in other APIs that would also need to be worked around before it could be used there.

[mfalken]

Thanks for those who filled in the Doodle: https://doodle.com/poll/gn4ew5n3eni3kyik?utm_source=poll&utm_medium=link

If you haven't yet, please do if you have a preference. It's looking like Fri Nov 20 is the popular option now. By the way, you can scroll to the right to see more days. cc/ @yoavweiss @jungkees @n8schloss @erik-anderson @SteveBeckerMSFT

[wanderview]

FYI, due to the time change in the US I think the meeting time is now 1 hour earlier than previously. This still works for me, but wanted to highlight for other folks. Might be hard to move later since folks in JPT are already calling in at 11pm.

[mfalken]

Thanks Ben for pointing this out. Let's go ahead with November 20 based on the Doodle responses. I've created #1553 for this meeting.

[wanderview]

@youennf can you share the slides you presented on background_fetch?

[youennf]

@youennf can you share the slides you presented on background_fetch?

They are currently available at https://docs.google.com/presentation/d/1bFufsntEd8WV__ipfXUS5T2j02C6_JDJRTA0rvlN-qw/edit#slide=id.p