If you send a notification that is accepted by APNs, nothing is returned.
If you send a notification that is malformed or otherwise unintelligible, APNs returns an error-response packet and closes the connection.
"Unintelligible" here includes of course invalid tokens, e.g. when the user uninstalls your app.
One of the upgrades was actually returning the last notification that made it through before closing the connection. Before that all you could do was add a sleep() and pray.
I can see that this is a library to send Push Notifications via Apple's servers, but what does it have to do with HTTP/2? Did Apple make a new version of the notifications API that uses HTTP/2? Why wouldn't they just silently upgrade the transfer protocol on the existing service?
Wow, that title. What this actually links to is a Go wrapper for Apple's new APNS interface, which utilizes HTTP/2. While this looks like a useful tool for the toolbelt, I would have much rather read something about how Apple got to the new protocol, or an interesting blog post about experiences using it. Having come from a shop that worked on the older protocols and been the tiniest part of an implementation with one of the brightest developers I've ever met, this new protocol is brilliant and a huge improvement, and it's a shame that this post, despite the title, doesn't do that justice.
APNS2 has a simpler interface and doesn't do things like generate website packages etc.
APNS2 has better test coverage, and handles errors in a more go-like way.
(Returns a response if a documented response was received from Apple, or an error where an unexpected error occurs.)
We are running this in production at http://carnival.io to send millions of notifications,
to my knowledge buford has not been tested in production yet.
From a first look APNS2 seems to add very little value over bufort which we are running in production very successfully. I really wonder how you manage to send millions of notifications (over which period of time?) without much boilerplate code.
I guess the unpopular bit is the insinuation that the following happens with Apple repeatedly: Somebody develops an industry standard, Apple implements it, people start claiming that Apple invented it.
If you send a notification that is accepted by APNs, nothing is returned.
If you send a notification that is malformed or otherwise unintelligible, APNs returns an error-response packet and closes the connection.
"Unintelligible" here includes of course invalid tokens, e.g. when the user uninstalls your app.
One of the upgrades was actually returning the last notification that made it through before closing the connection. Before that all you could do was add a sleep() and pray.