Re: [Webpush] Use Case related to subscription sets

It's even more complicated if you consider that a camera may have it's
own wifi connection ( watches are in this situation). Another example is
'multiple profiles/users' on the same device, sharing a single TCP
connection.

No objections - but it's one of the most difficult problems, implementing
it efficiently is tricky.

I suggest considering the same authentication mechanism that we discussed
for app servers
in the case of UAs - i.e. a JWT, signed with UA-generated key. Would
eliminate some
ambiguity, and make it a bit more clear what happens in cases of multiple
paths to reach
the final device ( camera/watch/etc).

Costin

On Mon, Jan 11, 2016 at 4:33 PM, Hervé Ruellan <herve.ruellan@crf.canon.fr>
wrote:

> The PR brings a good solution for our use case of a client acting as a
> proxy. I've got no objections merging it, on the contrary.
>
> Hervé
>
>
> On 09/01/16 02:05, Brian Raymor wrote:
>
>>
>> Is there any further feedback on Martin's pull request?
>>
>> https://github.com/webpush-wg/webpush-protocol/pull/67
>>
>> If there are no objections, I'd like to merge into the webpush draft
>> early next week.
>>
>> ...Brian
>>
>>
>> -----Original Message-----
>> From: Hervé Ruellan [mailto:herve.ruellan@crf.canon.fr]
>> Sent: Tuesday, November 24, 2015 10:22 AM
>> To: Martin Thomson <martin.thomson@gmail.com>
>> Cc: Benjamin Bangert <bbangert@mozilla.com>; webpush@ietf.org; Brian
>> Raymor <Brian.Raymor@microsoft.com>
>> Subject: Re: [Webpush] Use Case related to subscription sets
>>
>> Thanks for the pull request.
>>
>> I find it a good compromise: it covers our use-case while still
>> encouraging UA to reuse subscription sets.
>>
>> Hervé
>>
>> On 24/11/15 00:19, Martin Thomson wrote:
>> > On 23 November 2015 at 13:46, Brian Raymor <Brian.Raymor@microsoft.com>
>> wrote:
>> >> Perhaps the combination of "user agent decides" with "push service
>> *encourages*"
>> >> by limiting concurrent HTTP/2 streams is the potential compromise.
>> >
>> > That's where I'm headed.  Though I'm also adding "spec *encourages*"
>> > by using the word MUST.  I don't think that we get any gains for the
>> > important scenarios if we don't provide at least some encouragement to
>> > aggregate into a set.
>> >
>> > I've added text to the PR that explains what the push service might do
>> > to punish user agents that don't allow for aggregation.  It's relying
>> > on the same magical anti-DoS stuff again, which is hardly ever
>> > perfect, but often adequate in practice.
>> >
>> > Here I expect that looking at the connection will work to catch
>> > genuine but innocent mistakes.  The bad guys are always going to be
>> > harder to pin down.
>> >
>>
>>
> _______________________________________________
> Webpush mailing list
> Webpush@ietf.org
> https://www.ietf.org/mailman/listinfo/webpush
>