Nathan Youngman

Of course we also need to get a better feel for how errors from Apple occur in real life (readErrors). Using a production device token with a sandbox cert setup...

Validations before sending data should be able to prevent a good number of these. Still, there are errors that can occur at any time.

True. We should be able to use validations to avoid errors 2-8. I would like to separate out the preparation of the data to send Apple from the actual sending.

Wow. Thanks everyone. :smiley:

Since Apple dropped a new HTTP/2 API #57, I've been working on a new library that requires Go 1.6. https://github.com/RobotsAndPencils/buford I've used some of the ideas I saw in other...

It's not your setup, see #27 and #12. This is fixed on the develop branch, but that branch isn't quite ready to move to master.

In light of #57 (HTTP/2), maybe we should pull back this change and just make a few minor changes to fix broken tests for now.

There shouldn't be a batch limit. It's quite odd that it stopped sending. No error message? Has anyone else seen this? Apple does have a limit on connections opened to...

I'd be happy just to remove the "set an identifier if not set" logic and always require it from the caller.

I still think removing `determineIdentifier` entirely would be the best bet. Just require the identifier from the client.