Push API – W3C Working Draft | Hacker News Reader

Push API – W3C Working Draft

w3.org

162 pointssinak55 comments

This should have mandatory crypto built in from the beginning so that third party push providers can be used without causing privacy/security concerns.

The browser and the web server should negotiate an encryption key, or set of encryption keys during setup, and any push messages signed and encrypted. That way, any third party push provider can't inpect the messages.

Also, pigs should learn to fly.

I think the whole point of the spec is third-party push servers, a la GCM.

Encryption is still totally doable, as the client can negotiate directly with the service provider and then hand off to the single-socket push provider who would only see encrypted messages. You'd need metadata readable by the push server, however, so it can deliver messages to the right client.

Push messages already exist via Server-Sent Events [1]. This would simply standardize a less resource intensive single-connection model that third parties can interact with.

[1] http://www.w3.org/TR/eventsource/

Yes, you can already add your own encryption on top of the push system. What I was proposing is that it is standardised and doesn't require any extra code from the application developer.

So when I want to push a message out to one of my users, the push software I am using will handle encrypting it, and the push software on the users device will handle decrypting it. They will both be aware of the encryption keys involved because they will have negotiated them during the initial direct connection and will have them stored.

The push API is pretty awesome, the key difference between it and websockets in term of what it offers is this bit:

> A push service allows a webapp server to send messages to a webapp, regardless of whether the webapp is currently active on the user agent

So you can send push notifications to a user even if they don't have a tab open to your website.

Is that the bit (step 1) where they install the Service Worker? I.e. regardless of whether or not the page is open, the Service Worker is active?

Yes, and it has to maintain TCP connections with every push source you've registered, including re-connecting whenever your network situation changes. IMHO, this will lead to a lot of wasted bandwidth. Might not be a problem on desktop machines, but on mobile devices…

It already happens on mobile devices through APNs, GCM, and the like. Maybe mobile web browsers will have backing cloud services that use the mobile OS's recommended push server as a frontend to the mobile device? That doesn't sound like an ideal scenario though, considering that then the cloud service would have an enormous amount of user data going through it.

APN and GCM multiplex messages from all apps on a single TCP connection. That makes a significant difference.

That is how the Push API works as well. In fact usually developers will be using APN or GCM when they use the Push API.

well how do smartphones and apps deal with push notification in general? (i mean native apps)

The OS provides a messaging service. Applications can request differing levels of latency guarantee. The default on Android is to deliver immediately unless the device is off or sleeping, in which case it'll be queued for up to 4 weeks or until the device wakes up, whichever comes first. There is also throttling functionality, where an app that sends too many messages in a short period of time will find that its messages are only delivered when the device radio has to wake up for other non-throttled messages.

With iOS, Apple provides a service that event sources push their messages to, and a single TCP connection is maintained with the device regardless of the number of event sources. Google has something similar, if not identical.

I'm... not sure I want this. It sounds pretty useful/fun form a programmers perspective.

But from an end-user perspective I fear that it will lead too various "features" hogging resources. I would assume I could block them client-side, but I'm also the family-tech and quite frankly I don't expect my grandparents to do the same.

"User agents must not provide Push API access to webapps without the express permission of the use"

I share your concern. but I share a great fear that web apps will become increasingly irrelevant if they remain just documents that can't do anything when they're not explicitly in a running tab/window.

The permission requirement is the same one that allows for local storage, which could have caused great problems, but has so far held up well.

And the Ask toolbar doesn't get installed without expressed permission, either.

> The permission requirement is the same one that allows for local storage

Assuming this does not allow the push of UI interaction, then yes. If it would allow UI responses to be pushed then we would be one XSS vulnerability or "uneducated users" away from a new set of exploits for forcing pop-under advertising , and other potentially more insidious problems, through.

Local Storage and Index DB are storage only so the only attack routes are DoS-due-to-disk-full-errors which is generally less attractive (you can make more money pushing stuff to people than pushing people off the network).

Safari supports push notifications from websites for a while. I've found it convenient for the news websites I often visit just to check for new content. It's like RSS integrated into OS. One can misuse it, but generally it's useful.

We all get phone calls or text messages now when we don't have the associated binary apps open. Isn't this just the same thing for web apps?

It's the same things for web apps, with added near-infinite possibilities for spam.

It's permission based. Read the linked spec.

Are the permissions fine-grained enough to give users the power to say Yes to useful content and No to intrusive marketing messages?

Or is the option to include that choice going to be left to site designers?

I can't see that in the spec, so no. Like most things, if you don't like an app, you can revoke it's privileges and uninstall it.

The questions you keep asking are basic things solved by reading the article - you seem to have concerns, but you haven't done the minimum amount of work necessary to determine whether your concerns are valid.

Permissions are per-origin.

> The push API is pretty awesome

The idea behind this is awesome (although we already have this for some platforms - e.g. https://developer.apple.com/library/mac/documentation/Networ...)

But the API looks over-engineered. e.g.

navigator.serviceWorker.ready.then(function(serviceWorkerRegistration) {

serviceWorkerRegistration.pushRegistrationManager.register().then(...

I believe this spec is aiming to standardize the service offered primarily by Google (GCM) and Apple (APN), and open/standardize it for other people like pubnub/parse/pusher/...

The new web APIs are promise-ified by default rather than using the callback model of the past. So rather than register(callback), it's register().then(...).

Whether this is a good thing or not is another question I don't know the answer to.

My comment was mainly about serviceWorker.ready.then. I've never seen a JS API using such pattern - the name of the event as a property which is a promise.

A manager object/class also usually rings a bell.

Not that I'm an API specialist but I would rather implement it as:

document.addEventListener("serviceready", function() {

window.pushNotification.register().then(.....

For an event that only happens once, then the

    navigator.serviceWorker.ready.then(function(){...

version is preferable to

    document.addEventListener("serviceready", function() {

because A) you don't have to worry about removing your event listener after the fact, and B) if the event has already happened, then your callback will be called immediately.

>The push API is pretty awesome, the key difference between it and websockets in term of what it offers is this bit:

> > A push service allows a webapp server to send messages to a webapp, regardless of whether the webapp is currently active on the user agent.

Actually, the difference you describe is what Server-Sent Events [1] already provide: one-way notifications vs bidirectional communication with websockets.

This spec, rather, would standardize an API for a single-connection (read: better battery life) push server that third party webapps can use in the same way Google Cloud Messenger is used by third-party application servers for android apps.

[1] http://www.w3.org/TR/eventsource/

It's great that there are some standardization efforts in this space but for this to be useful we would need wide coverage of clients implementing it. Stuck using websockets for the next decade I suppose ....

This (and current Safari support) is what we (Roost - https://goroost.com - YCS14) are building off. Moving toward a complete web-push suite that supports all browsers on both desktop and mobile. Long road in front of us, but it's going to be an exciting world.

What I find particularly interesting in the spec is the editors list.

This has the potential to create cool and useful stuff, of course, but (IMO) it will be used mostly for advertising. Guess who's around the spec? Google and two telcos, of course.

I'm not sure how this would be used for advertising? This is about pushing messages to the user agent when it is not actively visiting a page. Ads are actively requested when you view a page, there is no need for push there.

So "they" can reach you, even if you don't visit. I guess it will be used to "remind" or notify you of all new posts/entries/content/everything. From there, it can move fast to notifying you about a new sponsor (or just saying who sponsored the new post).

Of course, it will need user approval, so the problem is (I hope) minimized (but just wait for the amount of Google+ "friend invites" notifications you will receive, even if you visited it only once).

The client has to have a TCP connection open to the server for this to work, right? Otherwise you run into all the problems of trying to reach machines behind DHCP, firewalls, etc.

Yup, so there will be push servers that clients can connect to, so clients don't have to maintain a ton of open connections, similar to APNs and GCM for iOS and Android, respectively.

That detail is implementation dependent. The site that wants to push to your browser hits up a third party push service provider that the user uses to get push notifications. That will probably Apple for iOS and Safari, Google for Chrome and Android, and I'm guessing Mozilla might provide one, as well as a host of other smaller companies that might be interested.

the interesting question for me is what this means for the big players and their respective platforms.

For instance, the more traditionally native features webapps take on, the less reason I have to need a native app hence am less coupled to a platform. I have always been a bit apprehensive to native apps because of platform lockin and it feels like this plus some of the push from google and their upcoming iot url push system makes it feel like web standards might really make webapps as common/popular/featured(or more) as native. It has plenty of kinks to work out around not being too spammy/sneaky with permissions but overall it feels like this is a major advancement for the web as a platform - assuming the big platform players implement it well.

It's good to see this in the works. Some of the apps I build are simple enough that a web-app is sufficient for almost everything, except that users really want push notifications. If this API becomes implemented (widely enough) that's one less thing we need PhoneGap and friends for.

Hey, check us (Roost YCS14) out at https://goroost.com - we already do this in a way that will abstract the browser part so you don't have to change anything once the other non-Safari browsers come online.

OT: Does anyone know what software was used to create those sequence diagrams?

Visio is pretty great.

Looks like it could have been OmniGraffle Pro.

LucidChart

If you're looking to build sequence diagrams, a guy I knew made this: https://www.websequencediagrams.com

You define everything with text and it renders a diagram based on what you've written. It's kind of like the dot/Graphviz of sequence diagrams.

If you want free / open source you could look at Dia. http://dia-installer.de/index.html.en

You could also try yED. http://www.yworks.com/en/products/yfiles/yed/

EDIT: This site has a nice list of a bunch of software that you can check. http://alternativeto.net/software/dia/

The list from your edit is great. I didn't even know that alternativeto.net existed.

Any word on when browsers other than Safari will be supporting push notifications?

FWIW, I've hearing "late this year" and "early next year" from some contacts.

There is code in nightly right now for it, I believe it is pref'd off though.

'This section is non-normative'

Like what the hell does that mean?

Please explain.

It means that this section contains example code, but is not actually part of the standard. From an operative point of view, this means that if it conflicts with anything else in the document, the "anything else" takes precedence.

> 'This section is non-normative'

Usually it is a concise or dramatic means to demonstrate something, but not adherent to the norms of that context.

For example, pouring petrol over a sofa and igniting it to demonstrate its fireproof attributes. Not normal behaviour but useful as a demonstration.

Normative is 'what should be normal', or 'ideally'—a normative argument is one pushing for a certain model of how things should be.

http://ln.hixie.ch/?start=1140242962&count=1

Reply on news.ycombinator.com