Serviceworker: Create F2F agenda - 7 November 2017

[jakearchibald]

Location

TPAC, Burlingame.

Agenda

Push API

I don't know the details, but I believe the Safari folks have some concerns around the push API, so let's go through that first, timeboxed to an hour.

Changes & bugs

• Implementer update
• Safari & multiple instances - do we want to discuss this since the call was missed? Related issues:
  
  
  
  • #1185 thoughts on exposing multiple-instances to sites via the DOM
  
  
  • #756 consider allowing multiple worker thread instances for a single registration
  
  
• Clients API
  
  
  
  • #1216 Should we expose reserved clients?
  
  
  • #1091 consider Client behavior for windows where initial about:blank is replaced with a loaded document this issue also tackles the naming
  
  
  • If we don't expose reserved clients, the following can be easily resolved:
  
  
  • #1215 Navigating reserved clients
  
  
  • #1035 postMessage() to a reserved client throws
  
  
  • #1034 what should Client.url return for a reserved Worker Client?
  
  
  • #1090 should clients.claim() control reserved Clients?
  
  
  • #1078 clarify that Client "creation order" means when reserved Client created
  
  
  • #1031 reserved Client objects and redirected navigations
  
  
  • #1030 Client.reserved references wrong context state?
  
  
  • #1123 WindowClient.navigate() behavior for reserved clients and in-flight navigations – also asks for clarification on in-progress navigations
  
  
• Should we ditch resurrection?
  
  
  
  • #1204 resurrected service workers don't fire an activate event which may break expectations - should we ditch resurrection?
  
  
  • #1198 ready promise doesn't work right when controlling registration is resurrected
  
  
• Removing foreign fetch
  
  
  
  • Should we remove link rel=serviceworker? Some CDNs like it.
  
  
  • What about InstallEvent? It doesn't do anything beyond ExendableEvent now.
  
  
  • #1207 Remove foreign fetch
  
  
  • #1188 Remove foreign fetch
  
  
  • #1186 How can we use SW module link in backward-compatible manner? – related issue to rel=serviceworker
  
  
  • #1158 ignores nonce/integrity/referrerpolicy attributes
  
  
  • #979 Define procedure to convert options object of sw.register() to Link header
  
  
• #1213 async waitUntil microtask issue
  
  
  
  • Also #1039 async waitUntil() does not allow extension on final promise resolution?
  
  
• #719 "no-cors" CSS SOP violation
• #204 Update algorithm should unregister SW on 404 and 410 errors
• #1189 Does changing updateViaCache mint a new ServiceWorkerRegistration or update them all? - is this kind of state
• #1191 What happens to the request body if the fetch isn't handled?
• #1200 is the browser required to start the ServiceWorker for postMessage() if there is no message event handler? - should we limit the special behaviour to fetch, or all service worker global listeners?
• DOM Nerfing all events on all event targets in a serviceworker after initial script evaluation seems odd
• Range requests
  
  
  
  • Can we treat 206 as a failure if the request wasn't a range request?
  
  
  • https://github.com/whatwg/fetch/issues/144#issuecomment-323453945
  
  
  • https://github.com/whatwg/fetch/pull/560
  
  
  • #1201 should FetchEvent.request ever be a range request if we cannot verify if the underlying resource is the same? - maybe part of the above, but I want
    
    to confirm
  
  
  • #913 cache.match should create partial responses if given a range request - probably nothing to discuss
  
  
• #1167 Proposal: FetchEvent.navigationLoadType - exposing reloads etc
• Keeping the service worker awake during a fetch
  
  
  
  • #1184 error handling for aborted streams feeding a navigation request
  
  
  • #1182 How long can the service worker stay awake during an active fetch
  
  
• #1193 messageerror event should use ExtendableMessageEvent - any objections?
• #1077 Find a way for service workers to get their current state - Bikeshed
• skipWaiting resolve time - I'm not sure we can do anything about this
  
  
  
  • #1187 skipWaiting() promise should resolve after promotion to .active
  
  
  • #1016 serviceworker.skipWaiting
  
  
• #1162 reflect internal "parsed" state in ServiceWorkerState enum
• #1146 "Unregister" operation may be interrupted by ServiceWorker#postMessage
• #1141 Service Worker makes AJAX Progress Listener not Working
• #1117 Should a service worker be allowed to register new service workers?
• #1103 Change the spec so that when a new SW is being registered it's set to not use cache the first time
• #1093 consider how clients.claim() should work with srcdoc iframes
• #1080 clarify Clients.matchAll() focus order for nested frames
• #910 Expose fragments cross-origin for subresources
• #755 consult document CSP in Register algorithm
• Introduce Background Fetch

Interesting feature requests

• Bypassing the service worker in particular cases
  
  
  
  • I'd like to get a feeling for how much of this is an implementation issue, and how much we need an API for.
  
  
  • Drink every time I mention static routes.
  
  
  • #1195 Proposal: Allow addEventListener/removeEventListener for 'fetch' to called at anytime
  
  
  • #1026 Consider mechanisms to bypass the service worker for things that we know won't be in it's cache
  
  
• Service workers in the clients API & adding to windows
  
  
  
  • #1037 Clients.matchAll() should exclude the current environment - OR SHOULD IT? (sorry)
  
  
  • #1036 Support returning ServiceWorker Client environments from Clients.matchAll()
  
  
  • #955 consider exposing self.clients on window and other worker globals
  
  
• #939 Recovering from fetch failures
• #921 Enabling multiple Service Workers for a single scope
• #1028 Consider providing navigation event - only if there's a 'champion' in the room
  
  
  
  • Related: #758 A way to control which client (or type of client) receives a navigation
  
  
• #998 user initiated 'uninstall' event - only if there's a 'champion' in the room
• #886 consider firing a "clientcontrolled" event on ServiceWorkerGlobalScope
• #863 Allow caches to opt-in to granular cleanup
• Preventing / delaying service worker updates
  
  
  
  • #822 Preventing server-forced updates
  
  
  • #1208 Provide a way to delay execution of a new service worker
  
  
  • #761 Allow preventing the update process to finish
  
  
• #820 Save initial files into cache storage instead of browser's cache
• #770 registration.ready promise
• #727 support updating multiple service workers coherently
• #614 A way to immediately unregister a service worker
• #587 consider exposing size for cache entries

[jungkees]

Client API and Navigate-hook

The latest agreement on FetchEvent (https://github.com/w3c/ServiceWorker/issues/1091#issuecomment-311023682):

• clientId: The client that initiates the request.
• resultingClientId: The potentially-reserved client that will house the resulting document/worker.
• replacesClientId: an existing client that will be replaced, or have its document replaced https://github.com/w3c/ServiceWorker/issues/1091#issuecomment-311325493

Proposal of not exposing reserved clients: #1216.

We are here now. I'd like to discuss:

• If the above meets the original use cases?
  
  
  
  • Per-client cache control.
  
  
  • Messaging to reserved clients: this will be excluded by #1091.
  
  
• Will devs be happy?
• Implementable?
• How to write spec?
  
  
  
  • To spec the above, we should resolve https://github.com/whatwg/html/pull/2809#issuecomment-313374002 first.
  
  

Issues:

• Client Id attributes and the initial about:blank case: #1091.
• Not exposing reserved clients (#1216) may resolve #1034, #1035, #1215.
• Navigate-hook: #765.
• Redirected navigation: #1031.
• Spec details: #1042.
• Resolved (not closed yet): #1080, #1090.

[jakearchibald]

@jungkees cheers! I've added the stuff I've been looking at so far to the OP. I'll merge yours into that.

I'm sweeping through open issues today.

[wanderview]

We're trying to implement https://github.com/whatwg/fetch/pull/146 and ran into a potentially thorny spec issue in https://github.com/whatwg/fetch/pull/146#issuecomment-341710548. Might be worth talking about in person.

[tomayac]

How do people feel about discussing “meaningless” Accept headers like */* for images in Firefox that make dealing with such requests in Service Workers hard and encourage relying on file extensions, especially as there is no consistency among how browsers deal with this.

[wanderview]

How do people feel about discussing “meaningless” Accept headers like / for images in Firefox that make dealing with such requests in Service Workers hard and encourage relying on file extensions, especially as there is no consistency among how browsers deal with this.

We can discuss it, but standardizing Accept headers is a bit orthogonal to service workers. I recommend raising a fetch spec issue to investigate alignment between browsers on Accept headers.

[tomayac]

While it applies to Ajax situations like fetch and XMLHttpRequest, the issue applies to all other static “fetching” scenarios as well, be it link, img, a, etc. Dealing with it in the context of the Fetch API alone seems too limited, at least from my point of view.

[jakearchibald]

@tomayac the fetch spec deals with all browser fetching, not just the fetch API.

[jakearchibald]

I've moved the push API stuff to the morning, so @beverloo can join.

[wanderview]

All spec'd browser networking loading goes through the fetch specification (modulo old specs not updated yet). Not just the fetch() API.

[tomayac]

Thanks both for the clarification. I guess fetch spec issue(s) is the way to go then.

[jakearchibald]

Planning to start around 09:00 tomorrow.

[delapuente]

Sorry to miss this F2F. Have a really nice and productive session!

[jakearchibald]

@wanderview points out that we would benefit from better tests for the cache API.

[jakearchibald]

Action: Add notes about cache sizes + opaque responses. But diversity in mitigation is welcome.

[jakearchibald]

Action: Built a test that creates a same-origin worker, but in the service worker fetch a response from another origin. According to the spec it should work, and the base url should be the response url. If it fails, @wanderview would be interested in adding that failing to the spec.

[mfalken]

Minutes: https://www.w3.org/2017/11/07-serviceworker-minutes.html