Re: [Webpush] Urgency of Messages

[Martin Thomson <[e-mail]>]

I think that you have the right idea here, picking urgency.

On 5 February 2016 at 07:29, Brian Raymor <[e-mail]> wrote:
>> A start would be to define a header for the 'monitor' request, with a list of flags, and
>> define common ones ( charging, network type ).
>
> I'm curious what the working group thinks about these proposals. Are there alternative approaches that we should consider?

I think that it would be better to simply have the client indicate the
urgency of the messages it wants to receive.  That way, there's a
clear understanding of what it will receive without being subject to
some unspecified algorithm in the push service.

We can make some suggestions perhaps.  For instance:

  On power and wifi: urgency = very-low
  On one but not both: urgency = low
  On neither: urgency = normal (or absent)
  Low battery: urgency = high

Setting a value indicates that the UA wants "normal" delivery of
messages at that urgency.  Higher urgency values than what was
advertised would get accelerated delivery.  Lower values would get
stored and delivered later.

We can also try to set some guidelines around what each urgency means:

  high = incoming phone call, time-sensitive alert
  normal = chat message, first email notification, calendar message
  low = subsequent email notification
  very-low = advertising

That might help establish some consistency between applications in
terms of what they use, so that user agents can set values with some
confidence that they will be respected.  I tried that with -nice, but
it could definitely be improved.