[NNagain] Internet Education for Non-technorati?

Wed Oct 11 16:59:13 EDT 2023

Hi Bob,

> On Oct 11, 2023, at 20:49, rjmcmahon via Nnagain <nnagain at lists.bufferbloat.net> wrote:
> 
> Yes, EDCAs are multidimensional. The coupling of EDCA to a DSCP field, Access Class (AC) or MAC queue is just an engineering thing. The EDCA is really just for an upcoming access arbitration and doesn't have to be held constant. And the values are under control of the WiFi BSS manager.
> 
> What's a WiFi BSS manager one might ask? It's an unfilled role that the standards engineers assumed would occur, yet are networking roles are under staffed all over the planet. The default EDCAs are just some made up numbers that have no simulation or other backing - though many think they're gold or something - which they're not.
> 
> There are so many things at play just for WiFi performance yet alone e2e. I wouldn't know where to start for a consumer label. Even marketing terms like WiFi 6, 6e and 7 seem to mostly add confusion.
> 
> Then engineers design for the tests because what else can they do? And the tests struggle to represent any kind of reality. Labels are likely going to have a similar affect.

	[SM] And this is why it matters how such labels are to be enforced... if the capacity numbers can be checked by end users against reference servers in a adifferent AS than any good-faith effort of engineers to make the test work sufficiently well will also help with general internet access. With good-faith I mean I exclude stunts like ISPs disabling the per user traffic shaper for the duration of a detected test, or treat test traffic with higher priority...

> Even the basics of capacity and latency are not understood by consumers.

	[SM] There are efforts under way to make end users more conscious about latency issues though, like apple's RPM.

> The voice engineers created mean opinion scores which I don't think consumers ever cared about.

	[SM] Why should they if they can directly judge the quality of their VoIP calls? Sure engineers need something easier to come by than asking end users about perceived quality and hence invented a system to essentially predict user judgements based on a few measurable parameters. But as far as i understand MOS is a simulation of end users that is better suited to the normal engineering process than doing psychoacoustic experiments with end-users, no?

> Then we talk about quality of experience (QoE) as if it were a mathematical term, which it isn't.

	[SM] Measuring experinence, aka subjective perception, is simply a hard problem.

Regards
	Sebastian

> 
> Bob
>> -----Original Message-----
>> From: Nnagain [mailto:nnagain-bounces at lists.bufferbloat.net] On Behalf
>> Of rjmcmahon via Nnagain
>> Sent: Wednesday, October 11, 2023 11:18 AM
>> To: Network Neutrality is back! Let´s make the technical aspects
>> heard this time!
>> Cc: rjmcmahon; Nick Feamster
>> Subject: Re: [NNagain] Internet Education for Non-technorati?
>> I've added many metrics around latency and one way delays (OWD) in
>> iperf
>> 2. There is no single type of latency, nor are the measurements
>> scalars.
>> (Few will understand violin plots or histograms on labels)
>> On top of that, a paced flow will have a different e2e latency
>> histogram
>> than an as fast as possible (AFAP) flow. They also drive different
>> WiFi
>> behaviors. Hence, it's not just a simple arrival rate and service time
>> anymore, even for queuing analysis. (Though Little's Law is pretty
>> cool
>> and useful for displacement ratings) Throw in BSS managed EDCAs and
>> all
>> bets are off.
>> _[RR] Wouldn’t the issue of EDCAs (i.e.different queues for
>> different priority classes with different tx parameters for each),
>> just make the analysis (more) “multidimensional”?  Might it be
>> possible to model such scenarios as N different collocated
>> bridges/routers), one for each access category?  Does any of what I
>> just said make any sense in this context? __J __J_
>> _ _
>> _RR_
>> Bob
>>> I think y'all are conflating two different labels here. The
>> nutrition
>>> label was one effort, now being deploye, the other is cybersecurity,
>>> now being discussed.
>>> On the nutrition front...
>>> We successfully fought against "packet loss" being included on the
>>> nutrition label, but as ghu is my witness, I have no idea if a
>> formal
>>> method for declaring "typical latency" was ever formally derived.
>> https://www.fcc.gov/document/fcc-requires-broadband-providers-display-labels-help-consumers
>>> On Wed, Oct 11, 2023 at 10:39 AM David Bray, PhD via Nnagain
>>> <nnagain at lists.bufferbloat.net> wrote:
>>>> I was at a closed-door event discussing these labels about two
>> weeks
>>>> ago (right before the potential government shutdown/temporarily
>>>> averted for now) - and it was non-attribution, so I can only
>> describe
>>>> my comments:
>>>> (1) the labels risk missing the reality that the Internet and
>>>> cybersecurity are not steady state, which begs the question how
>> will
>>>> they be updated
>>>> (2) the labels say nothing about how - even if the company promises
>> to
>>>> keep your data private and secure - how good their security
>> practices
>>>> are internal to the company? Or what if the company is bought in 5
>>>> years?
>>>> (3) they use QR-codes to provide additional info, yet we know
>> QR-codes
>>>> can be sent to bad links so what if someone replaces a label with a
>>>> bad link such that the label itself becomes an exploit?
>>>> I think the biggest risks is these we be rolled out, some exploit
>> will
>>>> occur that the label didn't consider, consumers will be angry they
>>>> weren't "protected" and now we are even in worse shape because the
>>>> public's trust has gone further down hill, they angry at the
>>>> government, and the private sector feels like the time and energy
>> they
>>>> spent on the labels was for naught?
>>>> There's also the concern about how do startups roll-out such a
>> label
>>>> for their tech in the early iteration phase? How do they afford to
>> do
>>>> the extra work for the label vs. a big company (does this become a
>>>> regulatory moat?)
>>>> And let's say we have these labels. Will only consumers with the
>> money
>>>> to purchase the more expensive equipment that has more privacy and
>>>> security features buy that one - leaving those who cannot afford
>>>> privacy and security bad alternatives?
>>>> On Wed, Oct 11, 2023 at 1:31 PM Jack Haverty via Nnagain
>>>> <nnagain at lists.bufferbloat.net> wrote:
>>>>> A few days ago I made some comments about the idea of "educating"
>> the
>>>>> lawyers, politicians, and other smart, but not necessarily
>>>>> technically
>>>>> adept, decision makers.  Today I saw a news story about a recent
>> FCC
>>>>> action, to mandate "nutrition labels" on Internet services offered
>> by
>>>>> ISPs:
>> https://cordcuttersnews.com/fcc-says-comcast-spectrum-att-must-start-displaying-the-true-cost-and-speed-of-their-internet-service-starting-april-2024/
>>>>> This struck me as anecdotal, but a good example of the need for
>>>>> education.  Although it's tempting and natural to look at existing
>>>>> infrastructures as models for regulating a new one, IMHO the
>> Internet
>>>>> does not work like the Food/Agriculture infrastructure does.
>>>>> For example, the new mandates require ISPs to "label" their
>> products
>>>>> with "nutritional" data including "typical" latency, upload, and
>>>>> download speeds.   They have until April 2024 to figure it out.
>> I've
>>>>> never encountered an ISP who could answer such questions - even
>> the
>>>>> ones
>>>>> I was involved in managing.  Marketing can of course create an
>>>>> answer,
>>>>> since "typical" is such a vague term.  Figuring out how to attach
>> the
>>>>> physical label to their service product may be a problem.
>>>>> Such labels may not be very helpful to the end user struggling to
>>>>> find
>>>>> an ISP that delivers the service needed for some interactive use
>>>>> (audio
>>>>> or video conferencing, gaming, home automation, etc.)
>>>>> Performance on the Internet depends on where the two endpoints
>> are,
>>>>> the
>>>>> physical path to get from one to the other, as well as the
>> hardware,
>>>>> software, current load, and other aspects of each endpoint, all
>>>>> outside
>>>>> the ISPs' control or vision.   Since the two endpoints can be on
>>>>> different ISPs, perhaps requiring one or more additional
>> internediate
>>>>> ISPs, specifying a "typical" performance from all Points A to all
>>>>> Points
>>>>> B is even more challenging.
>>>>> Switching to the transportation analogy, one might ask your local
>> bus
>>>>> or
>>>>> rail company what their typical time is to get from one city to
>>>>> another.   If the two cities involved happen to be on their rail
>> or
>>>>> bus
>>>>> network, perhaps you can get an answer, but it will still depend
>> on
>>>>> where the two endpoints are.  If one or both cities are not on
>> their
>>>>> rail network, the travel time might have to include use of other
>>>>> "networks" - bus, rental car, airplane, ship, etc.   How long does
>> it
>>>>> typically take for you to get from any city on the planet to any
>>>>> other
>>>>> city on the planet?
>>>>> IMHO, rules and regulations for the Internet need to reflect how
>> the
>>>>> Internet actually works.  That's why I suggested a focus on
>> education
>>>>> for the decision makers.
>>>>> Jack Haverty
>>>>> _______________________________________________
>>>>> Nnagain mailing list
>>>>> Nnagain at lists.bufferbloat.net
>>>>> https://lists.bufferbloat.net/listinfo/nnagain
>>>> _______________________________________________
>>>> Nnagain mailing list
>>>> Nnagain at lists.bufferbloat.net
>>>> https://lists.bufferbloat.net/listinfo/nnagain
>> _______________________________________________
>> Nnagain mailing list
>> Nnagain at lists.bufferbloat.net
>> https://lists.bufferbloat.net/listinfo/nnagain
> _______________________________________________
> Nnagain mailing list
> Nnagain at lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/nnagain