[NNagain] On "Throttling" behaviors

[Sebastian Moeller]

Hi Dan,

please see [SM2] below...


> On Oct 2, 2023, at 18:29, dan <dandenson at gmail.com> wrote:
> 
> 
> 
> On Mon, Oct 2, 2023 at 1:28 AM Sebastian Moeller <moeller0 at gmx.de> wrote:
> Hi Dan,
> 
> 
> 
> > On Oct 2, 2023, at 03:34, dan via Nnagain <nnagain at lists.bufferbloat.net> wrote:
> > 
> > I think a big problem with Net Neutrality is it's sort of a topic for providers and not for end users.
> 
>         [SM] I respectfully disagree. If my ISP sells me access to the internet, I expect being able to access the whole internet, at least as far it is with the reasonable power of my ISP. So I would argue it is a topic for both parties.
> 
> You missed my point.  That's a practical concern of the end user, not a technical one.  End users (~99% of them) want to 'turn on internet and it works as advertised' 

	[SM2] Ah, yes, I agree, end-sers IMHO are interested to experience no noticeable discrimination, but really do not care all that much how that is implemented. E.g. an ISP might use QoS/priority scheduling for VoIP (or FQ scheduling as long as VoIP flows stay well below the 1/N capacity share for N flows) and hence treat discriminate on a technical level without violating end-user expectations. I add however that such a scheme would also be in-line with well-balanced NN regulations. (But see Roland's example how this might result in less in-the-clear situations when an ISP only does this for their own VoIP traffic; though there are considerable obstacles to do this generally (though VoIP traffic has pretty clear patterns so one could simply do behavioral classification here, not that that is all that easy either)).


> 
>  
> >  Most end users have no concept of what it is, why it's good or bad, and if they have a compliant service or not.  They confuse 'speed' with packet loss and latency.
> 
>         [SM] I would agree that the majority of users might not have actionable concepts of how the internet works, let alone what to expect precisely, but I predict if we ask them whether they think it a good idea for their ISP to pick winners and losers of content providers I expect most of them to say no.
> I agree, again, the end user wants it to work 'as advertised'.  If they buy 100M service, they want to see 100M service. 

	[SM2] +1; this is BTW why the European regulators opted for recommending to implement "official" capacity-tests (erroneously called speed tests) as IP/TCP/HTTPS goodput tests over a small number of parallel flows to well connected measurement servers outside of all ISAP-ASs, as that is the number that end-users expect and are accustomed to compared to something like gross throughput (which would be harder to measure anyways). Side-note, they also recommend to use IPv6 if available... This initially raised quite some objections by ISPs (all technical in nature mostly concerning the precision and accuracy of such measurements) but at least in Germany in spite of the objections the system simply works and as far as I know neither end-users or ISPs are unhappy.


>  
> 
> 
> > This is why my arguments are primarily from a consumer protection perspective of forcing providers to be transparent in their NN or non-NN activities.
> 
>         [SM] Transparency is indded an important factor, but IMHO not sufficient.
> I'm being pragmatic.  Building up a NEW legal infrastructure means new challenges.  Any laws around internet access and NN have to co-exist with all the unrelated laws.  Codifying transparency connects that transparency to existing legal infrastructures.  "I bought 100M you gave me 10M when I used a service" falls to a number of legal precedents already such as bait and switch, false advertising.

	[SM2] I agree that trransparency is an important building block here.


> 
> >  If they are going to shape streaming video down, they should have to put that in the big print on the plan.  In doing this, customers would get some practical information that doesn't rely on heavy technical details and various opinions.  Consumers will understand the 100x20 w/ 1080p streams.  This could be on the broadband nutrition label *but* I would say you need to have that up-to-date for the times with key information on there and with strikethrough text.  ie, Download [ 100 ], Upload [ 20 ], streaming [ 720 1080p 4k ] and that list can be updated yearly during the BDC process or whatever.  
> 
>         [SM] Here we get into the weeds:
> a) for customers only having access to a single (broadband) ISP this information would hardly be actionable
> b) personally I could agree to this, but e.g. Netflix [720], Amazon [4K], GoogleTV [720] (or what ever google calls this service right now) would IMHO be problematic.
> c) I think that this gets really hard to enforce unless the ISP also throttled VPN/encrypted traffic and there I see a line crossed, no amount of transparency would make it palatable if an ISP would heavily throttle all encrypted traffic.
> 
> I agree, this one muddies the waters. 
> a) except that this can trigger other competitors, RDOF, BEAD, etc.  

	[SM2] Ah, I guess we agree that local access network competition is necessary (either via different competing infrastructures, but IMHO even more important via multiple ISPs competing on the same access network; I base this on the fact that the later will much more competition than the former).


> b) I would simplify this and categorize 'streaming video/entertainment' together and disallow any brand preferences.  Granted, limited JUST netflix (in my model) would allow Netflix to sue the provider if they really wanted.  

	[SM2] As long as this affects a whole traffic-class without exemptions (at least purposeful exemptions based on a profit motive by the ISP) I am less concerned about this... This also matches with what the EU did during the covid10 lock-downs, gave explicit interpretation that prioritization of vide conferencing/VoIP over video streaming was permissable with in the rules (I think that this was not stated as general interpretation but specific to the congestion experienced during the lock downs, but still it gives some precedence how NN rules can be adaptive and IMHO reasonable).


> c) VPN traffic is another beast and since it more or less bypasses DPI (to some degree, encrypted netflix still behaves like netflix) but this would be an easy carve out to say "once you're in a VPN, it's the wild west" because frankly it is, you have at least 2 internet providers in the mix in some way.
>  
	[SM2] True. However it is the one way end-users can use to do their own traffic engineering; e.g. for that otherwise pretty competent german ISP that decided to let its links from other T1-ISPs run too hot during primetime: users affected by this (which is not all that many, at least not all that many notice and correctly root-cause the observed issues) can use VPNs that are both well connected to the ISP as well as the target servers/AS that are consciously under-peered by the main ISP. However the fraction of users doing this will be miniscule.


> 
> > 
> > I'm really a proponent of creating a competitive environment, not controlling the technical details of those products.
> 
>         [SM] +1; but that only ever helps in a competitive market environment and that means there need to be enough buyers and sellers on both sides that no single one gets an undue influence over the market. At least in Germany we at best have an oligopoly market for internet access, so the "free market" by itself does not solve the discrimination problem. That might well be different in other parts of the world.
> We mostly have a competitive market in the US.  Really the only thing interfering with that is government funded monopoly either in funding designed to create monopoly or city franchise monopolies.

	[SM2] And this is why discussions like this here are so important, they help widen the perspective outside of one's own environment. Now back in 2009-12 when I lived in CA I did not have the feeling of really broad ISP alternatives, but I admit there were more than one option.


> >  And making sure key parts of the rules require transparency, because that can help drive competition.  If a local competitor is running a DPI's QoE box shaping down streaming, I think they should have to share that conspicuously with the end user.  I can then market directly against that if I want.
> 
>         [SM] Again I fully endorse that, but feel that by itself would not be enough.
> 
> I'm of the opinion that market tampering is 2 steps forward and then a full stop for a decade+.  The less (without ignoring ) the government gets involved in individual markets the better those markets are over time.

	[SM2] Mmh, I think this might be more subtle than we are treating it here, almost all markets depend on the government to act as referee to make sure "players" stick to the rules. Governments ince the neo-liberal explosion in the 90s have IMHO not done the best job as referees, but are slowly improving and are the only option anyway ;)


>  Yes, gov money can get build outs done, but in the US we've seen it time and time again that those build outs come and then absolutely nothing happens for a decade as those same companies wait for their next round of funding.

	[SM2] I might be too naive here, but if public intervention ends up financing FTTH to all premises with a reasonable topology (PtP or PtMP both with reasonably short distances) I would not care much if those build put would be stuck with e.g. 2.4/1.2 segment capacity GPON for the next decade. The big thing is going to get the old copper network replaced with something suitable fr the coming decades.

> This holds true is basically all markets, not just internet.  It's been a foundational argument in American economics since basically the beginning.  It's very hard for startups or small companies to form up in markets that were built by government money and those markets stagnate.  It's accidental protectionism essentially.  

	[SM] Well, look at the content provider market, as an example where government does not interfere all that much, here we see the big players use active measures to side-step a functioning market (like swooping up potential competitor early only to ket their products whimper and die). I happen to believe (based on some evidence like Amsterdam, Zuerich, Stockholm?) that it is possible to combine fair/equitable internet access (publicly financed and owned access networks) with a thriving market of access services where real competition keeps the providers honest and on their feet.


> 
> > I would say that I think that any government money should require NN plans.  If a company is taking government bribes for 100x20 speed, that should be a NN 100x20.  They can offer 100x20 w/ 1080p streaming also, but that gov money should have NN baked into what they/we are paying for.
> 
>         [SM] Interesting approach that I agree with, but again do not think it to be enough.
> 
> 
> > 
> > I know that one of the initial ideas of NN was to say that there are no fast lanes, but that's ridiculous because just having different speed plans IS having fast lanes for premium pricing.
> 
>         [SM] I think this is a misunderstanding. NN is not against ISPs offering different access capacity tiers at different prices, but that within the transparent customer-known "capacity-limit" the ISP does not pick winners or losers amount flows and especially not based on financial considerations. If an ISPs sells internet access they better do so, and they better not try to sell the same traffic a second time to the content provider. So the fast lane "fast lane" argument really means that the ISP does not built an unfair fast-lane for content provider A (capA) over content provider B (capB), and even that is subtle, if cabA puts caching nodes with the ISPs network, but capB does not a result in accessibility is not "on the ISP" however if the ISP would additionally slow down capB traffic that would be a problem. In a sense the thing that is problematic is building fast-lanes by slowing down all the rest.
> devils advocate (yes, non-NN is the device here haha) What's the practical difference between selling someone a 100M plan with 1080p streaming vs a 10Mbps plan for the same price?  Marketability mostly right?  ie, 100M looks better than 10M even though the customer practically only gets 10M for the streaming they are buying the connection primarily for. 

	[SM2] Technically you are right and game theory tells us the "rational" choice would be the 100M option. But game theory only works as long as you accept the premise, the moment you offer the alternative "change the rules" it is less clear what the best individual option would be, no? 


> If the ISP is primarily providing that 10Mbps '24x7' for streaming, then most of the cost to deliver this is in that first 10M of service.  Throwing in 90M of 'other internet stuff' is very cheap for the ISP.

	[SM2] That probably is a bit of a gamble by the ISP. A typical user will not actually create noticeable more traffic on average just because the access speed would be higher. Yet, networks are not tightly tailored for average speed anyway, as that will not help much during prime time and I believe inter-ISP traffic is typically not priced by the average but more like peak usage capacity (95%-ile billing). But I am rambling mostly, while you know real numbers so I accept your premise.


>  So which plan is actually better for the consumer?  100M w/ 1080p streaming or 10M NN?  100M NN would certainly cost the ISP more on average as more and more people pull 4k streams. 
> So if the plans are $100 100M NN, $65 100M w/ 1080p streaming, and $65 10M NN, is this better or worse for the consumer than $100 100M NN & $65 10M NN?

	[SM2] I guess that really depends on the consumer in question, some will cherish 4K over 1K enough to shell out more per month, some will not.


> I would argue that having the transparently shaped 100M w/ 1080p streaming in the mix is likely better for the consumer.  I will bring back your single provider area statement above as the weakness in this particular argument though.

	[SM2] Real options are always good. As an end-user however I am not unhappy if the options are less drastic (like with or without telephony flat rate) ;)


> 
> 
> 
> >  The only possible way around this is the allow or even enforce a price per Mbps and make it like a utility which I don't think anyone wants.  Any other model eliminates the sane enforcement of a 'no fast lanes' policy.  
> 
>         [SM] I think this is an attempt at reductio ad absurdum, that does not really cover the NN position well enough to be useful here. Think about it that wat, ISPs are not supposed to pick winners and losers (or rather tilt the table to create winners and losers).
> This statement was meant to tie with the 'no fast lanes' reductionism above.  If you can pay for faster service, then there are fast lanes based on money.    And if there are fast lanes based on money, why is a hybrid of 2 plans worse.  Buy the cheap plan for your streaming and the fast plan for your everything else.

	[SM2] I think "fast-lane" implies the wrong thing... it is HOVs that are the problem, users can still decide whether they want to access the internet via country roads or interstates as they should, but if push comes to shove a user paying for a service should be the one deciding how to prioritize traffic (the user already pays for the delivery), not the ISP and especially not the ISP selling the same traffic twice ;)

> 
> > I believe that transparency is the key here.
> 
>         [SM] THe EU actually agrees., an impoertant part of the NN-relevant EU regulation covers transparency.
> 
> 
> >  Transparency in how bits are delivered and managed to the end user allows the consumer to make choices without demanding unrealistic education of those consumers.

	[SM2] Transparency is good and is actually a precondition for markets to operate as efficient resource allocation tools....


> 
>         [SM] That however seems quite a challenge... though it is not unheard of to offer multiple parallel explanations at different levels of abstraction/complexity...
> Sometimes you need to layer your analogies or simplify things down to receivables.  Finding the best language for the consumer for the bold print is difficult. 

	[SM2] I agree, also there will not be one-size-fits-all explanation here, different user sets can "digest" different level of detail.


> 
> 
> >  They know if they want more than 1080p streaming.  They know if they want a gamer or home work focused service that does that by shaping streaming down to 1080p.  There are consumer benefits to a non-NN service.  
> 
>         [SM] Really only if the "unfairness" of that service are exactly tailored to a specific users needs and desires, not sure that any single such a policy would be desirable for the majority of end-users, but I might be insufficiently creative here.

> That's basically my claim, if an ISP can offer a 'bespoke' shaping model for the customer's goals then non-NN is better for them.

	[SM2] If it is bespoke and under user-control, it is not a NN violation anyway as NN does not affect what end-users do in their own networks ;)


>  This context came to be through servicing a ton of migrating work from home people.  They really want their zoom and voip and VPN sessions to work and they've had to ban their children from watching Disney+ while the work.

	[SM2] Yes, in that context offering VC over DASH is a marketable maybe even monetizable feature for an ISP to offer, it is also IMHO not an NN issue as the gist of NN is while not having the ISP pick winners and losers, end-users are free to do so (and the actual implementation of such a policy might well be outsourced to the ISP).


>  I have a number of customers with mikrotik routers as their dmark and wired ports and completely separate plans for their home office for this reason.  We don't offer any DPI shaping, trying to be NN-in-spirit with AQM.  Cake has met 90% of this need for us but there are circumstances where there's just too much demand by a houseful of kids for cake to really make this common scenario work well. 

	[SM2] My observation as well, when ever reasonably fair is suitable cake hits it out of the park, but once targeted unfairness is desired/required we need other tools to at least complement cake.


> 
> > Those benefits dont outweigh the harm from secret shaping, but if all of this is clearly shown in a conspicuous way like the broadband nutrition label and there were teeth in that label, then NN is a feature that can be advertised and DPI shaped is another feature that must be advertised.
> 
>         [SM] The only way I see DPI shaping ever to be in the end-users interesst is if that feature is under the end-users control. E.g. on a very slow link the administrator might decide that interactive VCs are more important than streaming video and might desire ways t achieve this policy and might appreciate if the ISPs offers such de-prioritization as a self-serve option
> I like the idea of a customer portal to set the DPI shaping to their preferences.  ie, sell them 100M service 'cooked to preference'.  outside the scope here but I wouldn't want NN legislation to prevent this option.

	[SM2] My understanding of European NN regulations make me believe you could offer such plans without any NN problems at all, as NN is not supposed to interfere with end-user (as in operators of private leaf networks) choice, but make it possible for end-users to chose by not having someone else make choices.


>  
> 
> 
> > Now for my ISP version of this.
> 
>         [SM] And now it gets interesting ;)
> 
> 
> > 
> > I WANT to compete against the companies that have to put 1080p and no 4k in their advertisements.  I want to offer full NN packages for the same price as their 'only 1080p' package.  I want the competitive environment and I want to know if the service area next to mine with just 1 or 2 providers doesn't offer NN packages so I can move in on them and offer them.  
> 
>         [SM] I wish all ISPs would have such a "bring it on" let's duke it out on the merits attitude! Yest the big mass-market ISPs in my home market do not really qualify for that all they compete about are nominal "speed"/price (really capacity/price) and some additional features like flat rate for telephony to fixed line and/or mobile networks. No details given about specific properties... for these one needs to scour rge internet fora to figure out what a specific ISPs customers mainly kvetch about (and then one needs to abstract over the fact that not all complaints are created equal and some say more about a user's deficiency in expectation than about the ISP's service delivery).
> Facebook and google reviews are awesome for figuring out where to expand.  Transparency rules and nutrition labels would open up bad ISPs to competition.  To be fair it would open up good ISP to competition by great ISPs also....

	[SM2] Over here the observation is that a lot of customers will stick to their ISP (often the incumbent ex-monopoly telco) even if that is more expensive than the competition... it seems some users really do not want to think much about telephony/internet and just have it work well enough.


> > 
> > So from my perspective, transparency in shaping is all I really think should be done.  And teeth in resolving a lie, ie false advertising, with a rulebook already in place for that so no new legal theory needs worked out.  "I bought the 1G NN package and I'm running the NN test suite and it says I'm shaped to 1080p video", rectify this ISP or I'm going to throw a legal tantrum to FTC, FCC, and maybe my own lawyer and get some penalty claims.  And that test suite is pretty darn simple.  speed test, 1G.  Netflix, 1080p.  DPI shaped.
> 
>         [SM] Alas, that NN test suit is currently made of unobtainium. Though some individual parts exist already. For example kudos to netflix for fast.com, which as far as I understand uses essentially the same of its CDN nodes for speedtests it would likely pick for content delivery so these are pretty "realistic" testing conditions. Alas the other big streaming platform have as far as I know not followed suit.
> As I tried to brain storm in a different context, what we would need at the very least would be speedtests that houses its servers in at least the networks of the big transit providers and runs comparative tests against these (either all or random pairs selected for each test). And while this would already be pretty ambitious to implement this would only scratch the surface of what would be required. However one redeeming fact is IMHO that NN-violations that achieve financial goals, by necessity need to be perceivable by the eye-balls so the thing we are looking for is likely not subtle but "in your face".
> 
> > 
> > End rant.
> 
>         [SM] I rally wish more ISPs would have your spirit (and that we had more ISPs competing for eye-balls)! Also I think that your stance let the market settle this is conceptually in-line with our current macroeconomic system and theory (I am just not sure whether yje internet access market generally is healthy enough to relat on this mechanism to sort out bad apples).
> 
> Regards
>         Sebastian
> 
> 
> 
> I think that you and I have a different level of faith in capitalism.  Your 'Not Enoughs' suggest this to me, though I might be reading too deeply between the lines. 

	[SM2] Looks like we do differ ;) I note that I am an end-user only so my influence on companies is negligible while for local/state/federal government I least have a vote....


>  I've worked in a number of industries from being near the top of the technical bits of a fortune 500 to running my own startups.  The softer and wider the net government rules casts, the less effective and more intrusive they seem to be with more non-sense.  The simpler and sharper toothed rules allow for a more vibrant level of competition.  

	[SM2] I generally agree, simpler, clearer rules are easier to follow/interpret, so the goal should be as simple as acceptable.



> I want very simple and clean rules with a very small service area and a lot of clamping force when necessary... so ISPs obey the rules because consequences are capital but have the freedom to innovate and expand.


	[SM2] I think we agree, the idea behind NN to me is not about hindering innovation at all, but I also see no macro-economic utility in "innovation" in the direction of more discrimination by ISPs (unless implemented as opt-in service for the end-user).