W3C home > Mailing lists > Public > public-webapps@w3.org > January to March 2019

Re: Web Push Notifications

From: Peter Beverloo <peter@chromium.org>
Date: Mon, 25 Feb 2019 11:55:11 +0000
Message-ID: <CA+tae9_OyQr5VZkijJ1A7KrruUGS=gqapwLWaa5hDUnF6XOt2A@mail.gmail.com>
To: Michaela Merz <michaela.merz@hermetos.com>
Cc: public-webapps@w3.org, ajv-efurgetonrit@vsta.org, stolossgreatm@outlook.com
Hi Jim,

Unfortunately I'm not aware of any public sources for such statistics.

Andy, Michaela: This is an appropriate forum for discussing the W3C Push
API, the design choices for which are outlined in the following document:


Specifically, while both the client side and server side (through RFC8030)
are standardized and have multiple implementations, including open source
ones, the user agent is able to (and usually does) select which push
service to use. Section 4.4. of the W3C Push API explains why this is the


On Sun, Feb 24, 2019 at 8:42 PM Michaela Merz <michaela.merz@hermetos.com>

> I think the term "web push" defines a process in which data is pushed to
> a client regardless of the clients current connection status. In other
> words - as long as the client maintains a connection to it's server,
> it's not really web-push - at least not in the sense the term is used
> today. This having said: Apple, Google and Mozilla maintain their very
> own web-push environments. It is convenient to use FCM (Firebase) as it
> provides an API that enables us to send messages to all operating
> systems - but it also gives a single entity control over all
> environments. In our work, we try to push to the individual endpoints -
> if we have data for Apple devices, we push to Apple, the Mozilla data
> goes directly to Mozilla's servers and everything Google is delivered
> via FCM.
> Unfortunately - web-push is also highly useless for anything that would
> really benefit from urgent notifications - such as important messages,
> incoming Web-RTC calls or the like because a) Google / Chromium devs
> seem not to be interested in fixing a bug that prevents web-push from
> cutting through doze on Android-devices and b) Apple doesn't have it
> (yet) on iOS. And even if the push reaches a device, all one can
> currently do is to display a notification. It won't be delivered to your
> app or web-site so you can't fetch data in the background and display it
> whenever the user is ready.
> It's sad - because web-push really could be a cool way to distribute
> content and information, but it is not - at least not as it is today. As
> to the question of whether or not there are overall volume stats
> available ? You will need to get those from all providers and add FCM
> distribution channels to get a good picture.
> /rant
> Michaela
> On 2/24/19 9:01 AM, Andy Valencia wrote:
> >> I am doing some analysis on web push notifications and I would love to
> know
> >> if there's any list of websites that are using this technology, as well
> >> as overall volume of web push notifications per platform. Is that
> >> information available publicly somewhere?
> > Of course, proper web push notifications are a decentralised technology,
> > thus problematic to systematically measure.  I have deployed push
> > notifications based on JSON-over-UDP (the most efficient), long polling
> > (the next most efficient), and also Firebase Cloud Messaging.
> >
> > Note that push messaging has two components; the event on the original
> > server and where it pushes, and then how that recipient gets word to one
> > or more mobile devices.  I'm assuming, given this mailing list's charter,
> > that it's the latter process which interests you.
> >
> > I suppose you could use Google's Firebase FCM as representative of all
> > push notifications.  It's proprietary and closed source, so (like oh so
> > many other things) you'll need to find out if Google thinks you should
> > know the answer.  But it does re-centralize the web, thus providing
> > a one-stop shop for a partial answer.
> >
> > Andy Valencia
> >
Received on Monday, 25 February 2019 11:55:47 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 13 March 2019 19:09:55 UTC