Re: [Webpush] When UA should send an acknowledgement?

[Martin Thomson <[e-mail]>]

On 9 June 2016 at 11:38, Benjamin Bangert <[e-mail]> wrote:
> On Wed, Jun 8, 2016 at 5:58 PM, Martin Thomson <[e-mail]> wrote:
>> On 8 June 2016 at 23:42, Benjamin Bangert <[e-mail]> wrote:
>>> I'm in support of at-least-once processing, so the ack would occur last per
>>> the DOM API. The difference would be that the DOM spec should require
>>> ack even if decryption fails or the SW fails for some reason.
>>
>> So I get the point about decryption: if it is bad, then that won't
>> change next time.  But a SW failure makes the application vulnerable
>> to random chance.  Sure, the SW might be bad, but badness isn't a
>> terminal condition there in the same way that crypto is.
>
> Agreed, I think its sensible for a UA to try the message once more. If
> a crash or something else bad still occurs, the UA can ack instead of
> trying for a third time. Given the UA's manage to remember tabs open
> right before a crash, it doesn't seem far-fetched for a UA to count
> process attempts for a message-id where a crash occurred afterwards.

That I can get behind.  I have opened a PR for the decryption failure.
I will open a second one for the case that the SW rejects multiple
times.  I think that we have a bit of leeway on that one though.

(Note: all of these are for the push API spec.  Do you think that we
need any changes to the protocol spec as a result of all this?)