Packets discard and Impact on Quality of experience

[boucadair]

FWIW, https://dspace.mit.edu/handle/1721.1/99941 revealed that: "63% of the external communication made by top-popular free Android applications from Google Play has no effect on the user-observable application functionality"

May be cited as an example of packets that may be tagged as "may-discard".

[danwing]

Based on the abstract (as I can't seem to access the paper itself), it seems to have analyzed blocking some communications from apps -- likely to advertising servers or to other non-primary servers, and said that most apps continued working fine when that communication was blocked.

[boucadair]

If we extrapolate the reasoning and consider that the app itself tags these packets as "may-discard" ones, the app will continue to function if its happens that some of these packets are dropped as a side effect of a reactive policy.

[danwing]

I expect lots of the traffic in that study is advertising traffic. Product Managers and engineers need to see ad impressions and would consider advertising traffic as important (even more important!) than the application's primary traffic (such as retrieving weather information).

In any event, being able to say "this flow is un-important" is just as difficult as "this flow is important": relative to other flows to/from the same host, or relative to other users? And who is to decide the importance, because I might assign my software updates to super low priority but if my neighbor has assigned a slightly higher priority to downloading a few petabytes of home movies from their cloud storage, it could take days and days.

That's why I hope to only have metadata describing relative importance within a flow, not between flows.