[Marnew] Parking lot session minutes

["Salz, Rich" <[e-mail]>]

Please post corrections and omissions.

Two initial items
                How can we get information out from operators, etc, collected at GSMA, etc, anonymized, etc, to provide metrics for real data so we can test and evaluate hypotheses.

                Layer of applicability (network<>UE at radio/MAC layer, or higher); where is the right place to do the work

ITEM ONE
How is radio resource control impacted by encryption, for example?
ECN notifications, deployment of ECN
Suggestion to thnk about sources of information, not necessarily the items
Interested in hearing about middlebox deployments (numbers/types) in operator networks
We need data, but don't now get tangled up in encodings (e.g.)
How/what can we exchange info so that overall wireless system is best used by everyone
Treat less as "we have a problem" and more as "how can we optimize"
Restrict domain/space to radio, e.g and not boil the ocean for the whole internet where we can't agree on what "fast" means
HOPS research group wants you!
We want virtuous cycle, more data feeds more requests for data, etc. etc
Not unlike IPv6 deployment, such as where competitors met at IETF to exchange best practices
Is there a need for IAB to do liaison work?   Seems to be Natasha, etc. informally is good enough, avoid the red-tape (and cant for research groups anyway)
Get NDA's among various parties, exchange data, and informally bilateral cooperate, bring things forth; don't want for next IETF
Operators can do experiments in a lab about effects of encryption and resource; don't need live production data, should also be easier to share
Use the marnew list to talk about test framework

ITEM TWO
Risk 1: boiling the ocean - don't try to solve 19 year old TCP-over-satellite, e.g
Risk 2: other cases other than UE to LTE network (mobile phones, not wireless uplink on company shuttle, e.g.)
Include 3G
Quality of the info we get?  Can we trust it?
What abstraction of the RAN control loop can we abstract into the transport control loop
Should discuss types of application traffic, to make existing lab tests more "real"
Network doesn't know what the apps need
Apps have local viewport, but getting end-to-end information is important
Want to be able to do real tests on real networks with new stuff being tested
We need to come up with something, like an I-D?, that talks about what a carrier could provide (the a & b for a/b testing); writing things down is helpful
What's the simplest thing we can do that gets us forward progress? Propose an operator run tests with same content plain/encrypted, and see impact and bring that data back.  Don't build formal framework at first.


--
Senior Architect, Akamai Technologies
IM: [e-mail] Twitter: RichSalz