Re: [Marnew] endpoint consent

["Joe Hildebrand (jhildebr)" <[e-mail]>]

In the SPUD work, we've been thinking of the path/app communications as "declarations".  Anyone can make any declaration, but until it's an interesting enough declaration (with an interesting-enough trust or incentive model), the other actors won't implement it, and will therefore ignore it. 

Imagine Pete Resnick's example of a declaration that says "I'll give you bandwidth X if you stay below bandwidth Y.  If you go over Y, I'll cut you down to X/2."

You might figure out a way to show that to the user.  You might only trust these indicators from your "local" network (e.g. the stuff between your phone app and your wireless operator), if the local network could prove that somehow.  You might set a minimum X before you got the user involved.  There are lots of potential UI issues here, but also the potential for a lot of interesting optimizations and business models.

-- 
Joe Hildebrand





On 9/23/15, 12:04 PM, "Marnew on behalf of Aaron Falk" <[e-mail] on behalf of [e-mail]> wrote:

>In the interest in warming up the discussion…
>
>
>I’ve seen in a few papers that ‘endpoint consent’ is considered a critical element when discussing middlebox participation.  But I haven’t seen a clear articulation of whether both endpoints must consent or, if not, how to decide which one.  I would say
> that currently at most one endpoint is explicitly involved with middleboxes.  But perhaps that is only because our protocols don’t support such a three-way negotiation.   
>
>
>—aaron
>
>
>