Comcast Says Should Be Able To Create Internet Fast Lanes For Self-Driving Cars (theverge.com)

An anonymous reader quotes a report from The Verge: Comcast filed comments in support of the FCC's plan to kill the 2015 net neutrality rules today. And while pretty much everything in them is expected -- Comcast thinks the rules are burdensome and hurt investment, yet it says it generally supports the principles of net neutrality -- there's one telling new quirk that stands out in its phrasing: Comcast now says it's in support of a ban on "anticompetitive paid prioritization," which is really a way of saying paid prioritization should be allowed. "The commission also should bear in mind that a more flexible approach to prioritization may be warranted and may be beneficial to the public," Comcast says in its filing. The key qualification is "anticompetitive," which is a term that could be interpreted in a lot of different ways depending on who's defining it.

Comcast doesn't just see paid fast lanes being useful for medicine, however. It also thinks they might be fair to sell to automakers for use in autonomous vehicles. "Likewise, for autonomous vehicles that may require instantaneous data transmission, black letter prohibitions on paid prioritization may actually stifle innovation instead of encouraging it," the filing says. This makes Comcast's position pretty confusing. Comcast says it opposes prioritizing one website over another. It even suggests the commission adopt a "strong presumption against" agreements that benefit an ISP's own content over competitors' work, but it's not clear how benefiting one car company or telemedicine company over another is any different.

Comments

What they are really saying

[reboot246]

We're against net neutrality when it hurts our bottom line and we're for it when it helps our bottom line. They don't care about customers; they care about profit.

Huh? That takes a special kind of stupid.

[dgatwood]

No engineer in his/her right mind would ever even consider designing a self-driving car in such a way that it required instantaneous communication. There's too much potential for network failures even under ideal circumstances with a perfect signal, just from routing problems alone. And that's before you consider vehicles driving through tunnels, rain fade, spectrum congestion, deliberate interference, etc.

Basically, the FCC asked, as part of people's filings, to come up with ideas for innovation that would be made impossible without paid prioritization. As expected, Comcast tried, and as expected, failed.

Fundamentally, Internet service either works or it doesn't. If slowness causes something to fail, then the service doesn't work, and therefore the best that paid prioritization can do is give the customers the service that they paid for. If slowness does not cause something to fail, then paid prioritization serves no beneficial purpose.

Therefore, there is no plausible situation in which paid prioritization can possibly be beneficial to consumers. Period. At best, it can only increase the potential for consumer harm, and at worst, it is the direct cause of consumer harm.

Re:Huh? That takes a special kind of stupid.

[dgatwood]

It is highly likely that you're wrong. And by highly likely, I mean absolutely certain. I can guarantee you with absolute certainty that no self-driving car system will ever be a centralized control system, because that would be fundamentally unsafe, for several reasons:

• Many driving decisions are made by humans in on the order of tens of milliseconds. Even under the most ideal circumstances, cellular networking is almost never that fast. So even if cars were the only devices on the Internet, they would still be less safe than a human driver if the decisions were being made by a remote server on the other side of a cellular hop.
• A centralized server is a single point of failure that could bring the traffic grid to its knees. If a single computer fails and the car decides to limp off to the side of the road, it's a minor nuisance for the other cars. If a million cars all fail and limp to the side of the road, the company goes out of business, because nobody is ever going to trust that company's self-driving cars again. That's what we call a company-limiting decision.
• Any self-driving system that can only work when in a major metropolitan area would completely eliminate the biggest, most important use of self-driving tech, which is to eliminate the need for drivers on long-haul trucking routes.

It is simply not realistic to believe that anyone would design a self-driving car system that is controlled from outside of the vehicle itself. That's why nobody is doing that. Nobody.

Note that self-driving cars do periodically use the Internet for things like asking for road condition updates, both to avoid closed roads and to alert it ahead of time about lane closures that might require special attention. None of that functionality, however, is life-critical, and any self-driving car must be able to cope without that information (both because it might not be kept up-to-date by local authorities and because the network might not always work). And in any case, that data communication is not continuous. A single data burst every couple of hours would likely be perfectly fine, and when you're talking about something that infrequent, you have a lot of opportunities to retry before it becomes important. That makes autonomous vehicle communications quite possibly the single least important data flowing over the Internet, priority-wise.

In other words, it's hard to imagine how you could possibly be more wrong.

Re:Already have 100 Gbps Internet3

[Joce640k]

Yep.

And hands up who wants their automotive safety to depend on a Comcast Internet connection?