You do not have permission to delete messages in this group
Copy link
Report message
Show original message
Either email addresses are anonymous for this group or you need the view member email addresses permission to view the original message
to pwa-dev
Tue,
Mar 29, 2022 (Hoch is MCing)
Look for scribe volunteer(s)
Primary: dmurph@
Secondary: Hoch
[penny] Changing PWA install
criteria
Removing installability check
requirements
Thoughts on adding an imperative
call to install such as navigator.install to the spec
How to protect users from spam:
gate on user gesture?
Looking at standardizing install
as a capability of a browser with user control, as opposed to a feature
that is locked to developer under certain criteria
Goal
Make install something that is
under the user's control. Instead of a developer carrot.
"capability a user can
choose to exercise on any webpage that they would like to make
permanent"
reconcile create shortcut
behavior with install behavior
offer something more imperative
as opposed to the beforeinstallprompt event
something like
navigator.install
Significant change!
Pages that do not have
information required today for install, we would fall back to
"create shortcut" behavior like
favicon for icon
page title for pwa name
Q: Should we preserve the
beforeinstallprompt API to help with promotion?
We should keep it around for
some time as sites have implemented it, and we can still use it for
install promotion on browser surfaces
Open question: do we show the
install icon in omnibox all the time? Probably not. We could use
promotion to show that (like we do now), and beforeinstallprompt works
with promotion.
Having install criteria for
further promotion for product help or omnibox is useful. This might not
imply total removal of install criteria. But having the install
capability accessible for users is a net win allowing users to choose to
install something.
Q: On the Android side, where
the install banner is a bit more intrusive.
it prevents promotion entirely
on Android. On desktop, it doesn't prevent anything, it just
allows the dev to show the install dialog on demand
Android - we can look at
preferred related apps to know if we should show that banner as opposed
to install criteria
We should act on behalf of the
user's interests here instead of relying on developers.
We should take the hint of
prefer related app.
We shouldn't show the
short-sheet if the manifest has preferred related apps.
interested in how we 'gate' the
imperative install API so we don't end up with situations like
notification & location right now
we can disallow this API if too
many users decline
we can start with only user
gesture gating, and add more restrictions if there is abuse.
Alancutter@ - we prevent showing
notification prompt if too many users decline?
it's a bit more complicated, we
do something called silent notifications
but yes, we will auto-block
notifications (user is notified of this) for really bad sites
if the site is maybe-good
maybe-bad, the notification is no longer interruptive
trick is because install isn't
a permission, we can't really take this exact same approach.
"ask to install" could
be a permission
could use site engagement score
- but it basically means (did user click on site)
Q: where would install option
be?
There would be an option in the
overflow menu to install any site
How do we distinguish the
"create shortcut" and "install" behaviors for users?
We might want to have it be
developer-controlled, they must use a manifest for certain behaviors
FYI - create shortcut behavior
change in chromium is being considered
creating shortcut for specific
url gets overriden when that site has a manifest & gets upgraded
If we add a generic install for
all sites, it gets used enough for non-web-app sites for us to figure out
how to deal with that
make change first, maybe folks
don't really install non-web-app-sites
[Daniel] dPWA Infra improvements
planned for Q2
Richer install modal
Install should be like a web
store install - screenshots, capabilities, etc.
There is a lot of async
reading/writing to 5? different locations, leading to a lot of
operations potentially racing each other (installs, uninstalls, etc.).
We want to add more metrics to
be able to measure improvements.
We are hoping this will solve a
lot of the icon issues people have been seeing.
General idea - reading/writing
will only be available inside a Command.
Creating an exclusive lock on
app ID (or multiple app IDs if needed).
[Alan] This sounds a lot like
problems with WebContents and wanting to prevent multiple background
ones from consuming lots of resources.
[Daniel] That seems related,
though slightly more related to the underlying global queue.
Lots of layers of overrides
exist today regarding OsIntegration. This is an attempt to cleanly
structure it all in one spot, which should increase execution speed (as
well be easier to maintain).
Might not be done in the
quarter, but the important parts will be getting addressed first.
[Daniel] Verify that Edge doesn't
have any "Chrome App"s supported.
extension->is_app()
alancutter: Are there flags Edge
can enable to try to see if Chrome app installation is supported?
Currently being aggressively
deprecated. If Edge has any, be ready.