Daniel's been thinking about traffic shaping and net neutrality — specifically, the fact that ISPs can use deep packet inspection to probe traffic at the aggregate level, and that unlike national security agencies, they've been eager to use that ability for commercial gain. The temptations are things like selectively throttling competitor services. And his point is that net neutrality as a concept exists to oppose exactly those techniques. So he wants us to unpack what the term actually means in practice, and what kind of internet it aspires to promote.
The timing on this is almost too good. We're in this strange limbo right now where there is no federal net neutrality law in the United States. The FCC restored Title Two classification of broadband in April twenty twenty-four, then the Sixth Circuit struck it down in January twenty twenty-five. So as of today, August seventh twenty twenty-six, the federal rulebook is blank. The only thing standing between ISPs and their commercial temptations is state law and public pressure.
Which is a thin fence and everyone knows it.
It is. So before we get into the technology — and we will, because deep packet inspection is where this gets genuinely interesting — let's pin down what net neutrality actually demands. The simplest version: all lawful internet traffic gets treated equally. No blocking, no throttling, no paid prioritization. Those are the three core principles from the twenty fifteen FCC Open Internet Order, and they map pretty cleanly onto what most people mean when they say the phrase. Your ISP doesn't get to decide that Netflix loads slower than Hulu because Netflix didn't pay the toll. Your ISP doesn't get to block a messaging app because it competes with their own voice service. The pipe is dumb. The intelligence lives at the edges.
And the phrase itself — "net neutrality" — I think people hear it and assume it's some kind of government power grab, the FCC trying to control the internet.
That's the big misconception, and it's exactly backwards. Net neutrality is about preventing ISPs from controlling the internet. The government's role here is to enforce a rule that says the network operator has to stay neutral. Without that rule, the ISP becomes the gatekeeper. They're the ones deciding which services work well and which don't. The regulation exists to prevent private control, not to impose public control.
So that's what it promises. Now let's look at the tool that makes breaking that promise so tempting.
Deep packet inspection. And I want to be precise about what this is, because the name makes it sound like ISPs are reading your emails, and they're not. DPI looks at the data inside a packet — beyond just the header that says where it's going — but it does this at the aggregate level. An ISP doesn't need to read the contents of your Netflix stream to know it's a Netflix stream. The traffic has a signature. The packet sizes, the timing patterns, the server addresses it connects to, the protocol behaviors — all of that together forms a fingerprint that says "this is Netflix" or "this is YouTube" or "this is a Zoom call."
Like identifying a person by their gait instead of reading their diary.
That's... actually a really good analogy. You don't know what they're thinking, but you know who they are and where they're going. DPI classifies traffic flows by matching those signatures against known patterns. And the technology has gotten extremely good at this. It can distinguish between different video streaming services even when they're both using encrypted HTTPS connections, because the traffic patterns are different enough. Netflix buffers differently than YouTube. The chunk sizes are different. The CDN endpoints are different.
So even encryption doesn't hide what service you're using.
It hides what you're watching. It doesn't hide that you're watching something, or which app you're using to watch it. And that's the key distinction. The privacy concern people jump to — "the ISP is reading my messages" — that's not what DPI enables. What it enables is the ISP building a detailed profile of your behavior as a subscriber. When you stream, what services you use, whether you're a gamer, whether you use VoIP. That profile is itself a commercial asset.
And once you've built that profile, the throttling step is almost trivial.
Technically, yes. Once DPI has classified a traffic flow as belonging to, say, a competing streaming service, the ISP's traffic management system can apply a policy to that flow. Deprioritize it. Rate-limit it. Introduce latency. The packets still get delivered — this isn't blocking — but the experience degrades. The video buffers. The call drops quality. The gamer's latency spikes. And the user blames the service, not the ISP.
Which is more insidious than blocking, because blocking is obvious. You try to open something and it doesn't work, you know something's wrong. Degradation just feels like the service is bad.
And that's exactly why selective throttling is the preferred mechanism. It's deniable. The ISP can always say "we're just managing network congestion." The Comcast versus Netflix case from twenty fourteen is the classic example here. Netflix traffic was being degraded at interconnection points between Comcast's network and the transit providers Netflix was using. Comcast said it was a peering dispute about who pays for the infrastructure. Netflix said it was a de facto toll. The resolution was that Netflix paid Comcast for direct interconnection — a paid peering agreement. Netflix wrote a check, and suddenly the congestion disappeared.
Funny how that works.
Funny how that works. And this wasn't even DPI-based throttling in the classic sense. This was congestion at interconnection points that Comcast had the power to upgrade but didn't, because letting it stay congested gave them leverage. The mechanism was different but the commercial logic was identical: the ISP controls the chokepoint, and the over-the-top service has to pay to get through.
And Daniel mentioned the commercial temptation specifically. This is where it gets structural. ISPs aren't just pipes. They're also competitors to the services that run over those pipes.
Right. Comcast sells voice service, so they compete with Skype and Zoom and whatever else. They sell video service, so they compete with Netflix and YouTube. They're both the road and one of the stores on the road. The conflict of interest is baked in. And DPI gives them the technical ability to quietly tilt the playing field in their own favor — degrade the competitor's service just enough that customers drift toward the ISP's own offering, without ever being told that's what's happening.
There's a current example of this that's worth flagging. In twenty twenty-five, Airtel in India proposed a product they called "Fast Lane" — apps would pay for prioritized delivery on their network. This is about as direct a violation of net neutrality principles as you can design. It's not even subtle. Pay us, your packets go to the front of the line. Don't pay, and your users get whatever's left.
And India's net neutrality rules, adopted in twenty seventeen, are among the strongest in the world. The backlash was immediate and loud. The proposal sparked exactly the kind of public outcry that net neutrality advocates have been warning about — because once one ISP floats this, the others watch to see if they can get away with it.
What happened to the proposal?
It's still being fought over, but the public pressure has been intense enough that they haven't launched it. The fact that it was proposed at all tells you where the industry's head is at. The commercial temptation doesn't go away just because there's a rule against it. It sits there, and companies test the boundaries.
So we've seen how the technology works and why ISPs want to use it. Let's ask what kind of internet we end up with if they do.
The toll booth model. That's the simplest way to describe it. If ISPs can charge for prioritization, the internet shifts from a meritocracy of ideas to a marketplace of access. The content that wins is the content that pays, not the content that's best. A startup with a brilliant new video platform can't afford the fast lane fees that Netflix can. So their service loads slower, buffers more, feels clunky — and users never find out if it was actually better, because the experience was worse from day one.
And the startup doesn't even know why they're failing. They just see high bounce rates and assume their product isn't good enough.
The chilling effect on innovation is real, but it's almost invisible because you can't count the companies that never got started. You can't measure the ideas that died before anyone saw them. What you can measure is the consolidation — the incumbents get stronger, the barriers to entry get higher, and the internet becomes less dynamic over time.
There's a free expression dimension here too, and I think it's the part people are most uncomfortable with. Once an ISP has DPI capability and no legal obligation to stay neutral, they can deprioritize content they disagree with. Not just competitors. News sites. Political speech. Anything.
And the post-twenty twenty-five legal landscape in the US — with no federal net neutrality law — means there's nothing stopping them from doing that at the federal level. Some states have their own rules. California's net neutrality law is still in effect, for example. But if you live in a state without protections, your ISP has the technical ability and the legal freedom to shape your traffic based on commercial or even editorial preferences.
The privacy piece is worth reframing here too. Daniel mentioned it in the prompt — the privacy concern is obvious. But I think most people think of DPI as a surveillance tool, and the privacy threat is that someone's reading your data. The deeper privacy problem is the profiling. The same DPI that enables throttling also builds a detailed map of your digital life. What services you use, when you use them, for how long. That data has commercial value independent of any throttling decision.
And ISPs have every incentive to monetize it. They're not selling your browsing history in a spreadsheet — or at least, the major ones say they're not — but they're using that data internally to inform their own product decisions, their advertising businesses, their content strategies. Your behavior is the raw material for their business model. Net neutrality doesn't directly regulate data collection, but it removes one of the primary commercial incentives for building that surveillance infrastructure in the first place.
If you can't monetize the traffic discrimination, the DPI investment looks less attractive.
Right. Or at least, the use case narrows to legitimate network management — which is a real thing. ISPs do need to manage congestion. DPI can be used for that without being used for discrimination. The question is whether you trust them to draw that line themselves, in the absence of rules.
Let's talk about the international picture, because the US is increasingly the outlier here.
The European Union adopted its net neutrality regulation in twenty fifteen, with the BEREC guidelines providing enforcement teeth, and it remains in force. It's not perfect — there have been complaints about zero-rating loopholes and inconsistent enforcement — but the framework is there and it's binding. India's twenty seventeen rules, as I mentioned, are among the strongest in the world. They explicitly prohibit any form of discriminatory treatment based on content, and they've enforced it. The Airtel Fast Lane proposal is being tested against that framework right now.
And the US, as of today, has... nothing at the federal level.
Nothing at the federal level. The twenty fifteen Open Internet Order was repealed in twenty seventeen under the previous Trump administration. The twenty twenty-four restoration under the Biden FCC lasted less than a year before the Sixth Circuit struck it down. So we're in a vacuum. Some states have filled it. Some haven't. It's a patchwork.
What does net neutrality actually aspire to, then? If we strip away the regulatory mechanics, what's the vision?
An internet where the network operator is a neutral conduit. Where innovation happens at the edges — in someone's dorm room, in a garage, in a small team that has a better idea than the incumbent. Where users decide which services succeed by choosing to use them, not by having the choice made for them by a throttling algorithm they can't see. The network should be like the electrical grid. The power company doesn't care whether you're plugging in a Sony television or a Panasonic one. It just delivers electricity. The value is in the devices, not in the delivery.
And the counterargument — the one ISPs make — is that they built the infrastructure, they should be able to manage it as they see fit, and that regulation stifles investment.
That's the argument, and there's a kernel of truth to it. Building and maintaining network infrastructure is expensive. The question is whether the investment case actually depends on the ability to discriminate. The evidence is mixed. Investment didn't collapse when the twenty fifteen rules were in place. It didn't surge when they were repealed. The carriers' own earnings calls tend to emphasize other factors — competition, demand growth, technology transitions. The net neutrality debate is not really about whether networks get built. It's about who gets to extract value from them once they're built.
The ISP wants to capture value from both sides — subscription fees from subscribers, and access fees from content providers. It's a two-sided market and they want to be the toll collector on both.
And that's the structural conflict. The over-the-top services — Netflix, Zoom, whatever — they're building businesses on top of infrastructure they don't own. The ISPs see that and think, we should be getting a cut of that. The consumer sees it and thinks, I already paid for my internet connection, why is Netflix buffering? The ISP is double-dipping, and net neutrality is the rule that says you can't.
I want to go back to something you said earlier about degradation versus blocking, because I think this is the part most coverage gets wrong. The public imagines net neutrality violations as dramatic — a website is blocked, a service is cut off, there's a clear villain. But the actual mechanism is almost always selective degradation. Things just feel slow. And the user has no way to know whether it's their connection, the service, or the ISP's policy.
There's a whole industry built around measuring this, by the way. Companies that run tests from thousands of endpoints to detect differential treatment. And they do find it. Not always, not everywhere, but the patterns emerge. Certain services get degraded at certain times on certain networks. It's subtle enough that any individual user would never notice — you'd just think Netflix was having a bad day. But in aggregate, the data shows systematic discrimination.
And without net neutrality rules, there's no enforcement mechanism even when you do detect it.
You can sue, maybe. You can file a complaint with the FTC under general consumer protection authority, but that's not the same as a clear rule that says "thou shalt not throttle." The FTC can go after deceptive practices — if an ISP says they don't throttle and they do, that's actionable. But if they're honest about it? If they put it in the terms of service? Nothing stops them.
So what we're left with is a system where the only real constraint is public pressure and the fear of reputational damage. Which works until it doesn't.
It works for the biggest, most visible services. If Comcast throttled Netflix tomorrow, people would notice and there'd be an uproar. But if they throttled some new startup that has ten thousand users? Nobody would know. The startup would just fail. And that's the point — net neutrality isn't really about protecting Netflix. Netflix can pay for peering, they can absorb the tolls, they'll be fine. It's about protecting the next Netflix, the thing that doesn't exist yet and won't exist if the barriers are too high.
The thing that gets me is how gradual this could be. If the internet becomes a marketplace of access, it won't happen with a switch flip. It'll be incremental. A paid prioritization deal here. A zero-rating arrangement there. Over time, the internet just feels different — certain things load faster, certain things don't, and you stop questioning why.
The frog in the boiling water. And we're already seeing pieces of this. Zero-rating — where certain services don't count against your data cap — is a form of discrimination. It's not throttling, but it's still the ISP picking winners. If YouTube doesn't count against your cap but Vimeo does, you're going to use YouTube. The ISP has shaped your behavior without ever slowing anything down.
Before we wrap up the main discussion, I want to hear from Hilbert. He's been sitting on something about this topic.
Hilbert: I built one of these systems. Late nineties. Regional ISP in Ohio, got bought by a larger carrier. They handed me a box and a manual and said make it work.
Wait — you implemented DPI-based traffic shaping?
Hilbert: Before anyone called it that. It was just a packet classifier. The hardware was some rack-mount thing from a company that doesn't exist anymore. I wrote the scripts that matched traffic signatures to policies.
What was the stated reason?
Hilbert: Network management. Congestion control. Prioritize business traffic over residential during the day, that kind of thing. Nobody said the word throttling. Nobody said competitor.
But it became clear pretty quickly what else it could do.
Hilbert: The sales team figured it out before we did. They came down to the network room one day and asked if we could identify voice-over-IP traffic. Specifically, the traffic from a competing VoIP provider that was starting to eat into the company's own phone service. I said sure, we can see the signatures. They asked if we could degrade it. Not block it. Just make it sound bad.
What did you do?
Hilbert: I told them it would break things. Which was true, sort of. The classifier wasn't precise enough to degrade only the competitor's traffic without also catching some legitimate stuff. False positives. We'd have had angry customers. They backed off.
But the fact that they asked.
Hilbert: They asked. And the thing is, the engineers weren't pushing back on principle. We weren't net neutrality advocates. We just didn't want to get paged at three in the morning because the traffic shaping rules had gone haywire and taken down something important. The principle came later, after we'd already built the tool.
Do you still have any of that — the scripts, the configurations?
Hilbert: I've got the original Perl script. The one that did the signature matching. It's about four hundred lines. I kept it on a floppy disk for years, then moved it to a USB drive, now it's just a file on my laptop. I look at it sometimes.
As what — a memento?
Hilbert: As a reminder that we built it before we thought about what it was for. The tool was neutral. The use case wasn't. And once the capability exists, the use case finds it eventually.
That's... that's exactly the dynamic we've been describing, just at the engineering level. The technology gets built for one reason, and then the commercial logic takes over.
Hilbert: It was a different time. The internet felt like a public project. We were just trying to keep the packets moving. Nobody was thinking about monetizing the visibility.
Four hundred lines of Perl.
Hilbert: It wasn't complicated. That's the part that stuck with me. It's not hard to do. Any competent network engineer can build a traffic classifier. The only thing that stops it from being used for discrimination is a rule saying you can't.
Right now, in most of the US, that rule doesn't exist.
Hilbert: I know. That's why I kept the script.
I want to sit with that for a second — the banality of it. The tool was four hundred lines of Perl. The decision to use it for discrimination wasn't made by villains. It was made by a sales team looking at a revenue slide. That's how this actually works in practice. It's not mustache-twirling. It's quarterly targets.
The engineers who built it weren't thinking about the societal implications. They were solving a technical problem. Make the network run better. Classify the traffic. The fact that the same tool could be used to tilt the playing field — that realization came later, and by then the tool was already in the rack.
Hilbert, when you look at that script now, what do you actually think?
Hilbert: I think it still works. The signatures have changed — different protocols, different patterns — but the logic is the same. You could update it in an afternoon.
That's not what I meant, but it's somehow a better answer.
Hilbert: I know what you meant. I just don't have a good answer for it. I built a thing. Other people decided what to do with it. That's most of engineering, honestly.
The question of what kind of internet we're building — it doesn't get decided in policy papers. It gets decided in network rooms, by people looking at a traffic graph and a revenue slide. And the policy either constrains that decision or it doesn't.
Right now, at the federal level, it doesn't. So the question becomes: what happens next? State-level action is the obvious path — California's law is still standing, other states could follow. Congressional legislation is possible but hasn't moved in years. The most likely scenario, if I'm being honest, is a slow drift. No dramatic changes, just incremental degradation of the open internet, until one day we realize we're living in a tiered system and nobody can quite remember when it happened.
Would we even notice?
I don't know. I think about the boiling frog thing you mentioned. The changes are small enough, and the services that get hurt are small enough, that most people won't feel it directly. Their Netflix will still work. Their YouTube will still load. The internet will feel normal. They just won't discover the thing that would have been better, because that thing never got a chance to load fast enough.
Net neutrality isn't a technical regulation, in the end. It's a statement about what kind of internet we want to live in. Whether the network is a public conduit or a private marketplace. Whether innovation happens at the edges or at the toll booths. The technology — the DPI, the throttling, the traffic shaping — that's all just the means. The question is what we decide the ends should be.
That decision is currently unmade in the United States. The Sixth Circuit didn't say net neutrality is bad policy. They said the FCC didn't have the authority to impose it under the legal framework they used. Congress could fix that tomorrow. They haven't. So we wait.
Hilbert, anything else from the network room floor?
Hilbert: The floppy disk was labeled "traffic stuff." That's it. No date, no version number. Just "traffic stuff."
Of course it was.
Thank you, Hilbert. For the producing and for the Perl archeology.
This has been My Weird Prompts. You can find every episode at my weird prompts dot com, or email the show at show at my weird prompts dot com. We'll be back soon.
