Closed LJWatson closed 5 years ago
@mgiuca, @dominickng, would be great to hear your perspectives also...
What progress has your spec made in the last 12 months?
Limited. We are mostly waiting on implementation feedback.
Mozilla implemented the manifest processor around six years ago but it's only now making its way into products: namely Firefox Preview on Android. As part of that effort, we identified some non-blocking issues with icon's purpose (fixed), and also with CSS color parsing. Mozilla is hoping to provide further feedback as we gain implementation experience and put this out into products and and in front of real people.
The Manifest spec is also shipping in Safari, and we'd really like to get more feedback from Apple about their experience.
Is anything blocking your spec from moving to CR?
Yes. Because the spec lacks any programmatic means to install a webapp, it makes the spec notoriously difficult to test: basically, we need to do a bunch of manual testing.
There is also some disagreement on the a BeforeInstallPrompt()
event (BIP). Despite BIP having been in the spec for a few years, neither Safari nor Firefox have opted to support it. As it stands, Mozilla does not plan to support BIP. We are unsure what WebKit's plans are - @hober maybe could let us know? If WebKit doesn't plan to support it, then we should probably remove it from the spec.
We also made a mistake trying to use WebIDL to model the data structures of Web Manifest. Although it seemed like a good idea at the time, this turns out to not work well in practice: no implementation does processing of Web Manifest data through the a Web IDL binding layer. Some work is underway to change the spec to use Infra types instead.
There is also a proposal to add "shortcuts" to icons #768 from @aarongustafson, but lacks implementation interest from a second engine. We might need to move to incubation?
If yes, what is your plan to unblock it and do you need any help?
To unblock, we might need to make some hard choices about BIP and new feature proposals. We should try to get agreement on a limited number of things (i.e., maybe 3) to work on over then next year.
What do you want to achieve for your spec in the next two days (of breakout sessions)?
Decide what's in scope, and what is not. Hopefully hear from @hober or other folks from WebKit about what they would like to see in the spec. At this point, Mozilla is happy with the base set of manifest members.
There’s actually a lot of interest from developers for Apple to implement icon support (https://bugs.webkit.org/show_bug.cgi?id=183937) and beforeinstallprompt
support (https://bugs.webkit.org/show_bug.cgi?id=193959).
Thanks @marcoscaceres. Your summary looks pretty good to me. It would be a bit sad to remove BIP since it's a pretty core part of what we're doing on Chrome, and we've seen a lot of demand from developers for it (as @tomayac has pointed out, so have other vendors).
We've not removed it yet :)... we just need to see if we can sell it better and get other folks to understand the utility of it.
There’s actually a lot of interest from developers for Apple to implement icon support (https://bugs.webkit.org/show_bug.cgi?id=183937) and
beforeinstallprompt
support (https://bugs.webkit.org/show_bug.cgi?id=193959).
I don't know if any of the replies on those bugs are from Apple / WebKit engineers. Having a handful (tens) of +1s from external developers doesn't really give us much of a signal.
It would be good to hear from Apple on this.
We (Chrome) would be open to deprecating BeforeInstallPrompt in favour of a different programmatic install API (I don't think there's a lot of love for the API we settled on), but it is fairly heavily entrenched on the open web, having been available in Chrome (and I believe other browsers like Samsung and Opera) for around 5 years.
Chrome's usage metrics for BIP show that this is currently used on 1.5% of all page loads (though I'm not exactly sure what we're measuring here; possibly just any mention of the global object BeforeInstallPromptEvent
), which is too high to remove outright.
Perhaps we can chat with Apple @ TPAC. Are you here @mgiuca?
No I'm not. Happy to hear about the results of any discussions.
Me and @marcoscaceres are here
I spoke with a bunch of folks with Samsung at lunch and they seem very interested in Shortcuts. Kicking off some deeper conversations with them.
A brief rundown of what we talked about at TPAC based on my recollection. Please correct if I messed anything up.
rel
value for link
elements to the Microformats registry for inline declaration of shortcuts (browser interest and UI exposure TBD).ImageResource
- Agreed to break out ImageResource
to a separate spec which would be referenced by Web App Manifest (and other specs). ImageResource
will be extensible in context to allow for manifest-specific features (and other specs, naturally). Once the new spec is in place, we will migrate ImageResource
issues to that repo. Agreed to add description
(#773) and color_scheme
(#758) members to ImageResource
upon migration.beforeInstallPrompt
- moving this feature to optional (PR inbound: #797)allow_list
- Discussed bringing the concept of Feature Policy-esque capabilities declaration into the Manifest for auditing, user awareness, etc. @aarongustafson took an action item to write an explainer. (#798)getInstalledrelatedApps()
- Agreed to move forward with thisisInstalled()
- Into to the idea of a window being able to determine if a PWA for that domain & scope is installed so that it may be launched. Seems to overlap some with prefer_related_application
(where that app is the PWA), Chrome Apps’ url_handlers
, and Windows’ web-to-app-link concept. General agreement that we need to solve this. @raymeskhoury to look for more use cases.ImageResource
is already set up to handle the MIME types, but leans on HTML for sizes
definition. Suggested that the proposal for a third dimension be taken to the WhatWG and, once cleared, we can look to add it into the spec.Agreed to break out ImageResource to a separate spec which would be referenced by Web App Manifest (and other specs).
@aarongustafson ImageObject
from Payment Handler API (https://www.w3.org/TR/payment-handler/#imageobject-dictionary) could be changed to ImageResource
.
Please can you respond to this comment with a brief specifications status report for the WebApps meeting at TPAC. The report should address the following:
We're tracking these status reports in WebApps issue #19
Thanks.