They ignore QoS tagging from non-Comcast devices, FYI. I've tried that trick of tagging all my traffic as DSCP 30 (Flash High), and it gets reset to 0 (Best Effort) as soon as it crosses the head-end. This is unfortunately also commonplace for providers and an annoyance to businesses in our new post-pandemic world where they need to appropriate tag collab/voip traffic and it gets reset along the way. Lumen (formerly Level 3) often resets to best effort as soon as you cross into their transit edge from your ISP. These values often do not get carried across peering relationships. QoS across the Internet is not guaranteed.

In the case I'm talking about, it really only matters how Comcast's network treats their direct customers, not how you get treated after, and they absolutely ignore customer QoS (at least on residential links).

That makes me wonder how it can distinguish between a Comcast and non-Comcast device on the network. If this were built into the Cable Modem it would be reasonably doable, but if it's a separate box entirely I'm not sure.

The CPE (Customer Premise Equipment) /is/ the cable modem, as well as a WiFi router. Another thing it does is provide an authenticated SSID for WiFi that is available to other Comcast customers and utilized to boost coverage for Comcast's mobile offerings, via built-in WiFi profiles. The basic understanding I have is that you get free "unlimited" with the Comcast CPE, because they got sued and lost due to the provided authenticated WiFi being used by other customers to utilized what would otherwise be part of your data cap, so they had to lift the cap when they provided the equipment. Since your own equipment doesn't provide the Xfinity WiFi SSID, you must pay to lift the cap.

I did a quick search and couldn't find what I was looking for, but it looks like in 12 states they were prevented until 2022 for enforcing the data cap, and there's a pending class action suit in California related to the WiFi impact on data caps: https://topclassactions.com/lawsuit-settlements/lawsuit-news...

Anyway, that's my understanding. The net-net is I coughed up the $30/mo because I have no alternatives to Comcast (even at lower speeds) because my landlord has an exclusivity contract with them, and I don't want their shit-tier CPE anyway, especially throttling my traffic and broadcasting a WiFi SSID I don't control. My entire network is proper SME enterprise-grade rackmount gear and I have WiFi 6E throughout my house that is secured to only allow devices I control/allow to connect.

So correct me if I've misunderstood, but you originally said "customer hostile QoS" which would imply traffic shaping being done on the CPE router, say perhaps with the goal of decongesting the cable segment. This comment you've seemed to walk it back saying the problem with their CPE is that it facilitates service to other customers. If that second thing is really the only problem, then couldn't you just stick their CPE in a Faraday cage and/or unplug its antennas?

Also if they are doing traffic shaping, where does the default unclassifiable (eg VPN) traffic fall? It should be easy enough to get an unlimited VPS at a close enough data center and relegate Comcast to a dumb pipe (as they should be).

I don't think it does sophisticated traffic classification. It's just a way for Comcast to say "you've used more than 10GB of traffic this month, all of your packets get sent down to the best effort Thunderdome until the next billing cycle".

That conflicts with OP's description saying their Steam downloads were 20% slower, which presumably was tested before hitting any soft transfer cap.

I'd guess the goal of traffic shaping on CPE would be to keep upload bandwidth on the cable segments less congested. Or perhaps conntrack is easier to do in a distributed manner?

I wonder how Hulu streaming traffic is QoS'd on Comcast. Someone should check that out and file a lawsuit.

I haven't walked anything back, thank you. There's more than one bad thing Comcast does to their customers, we haven't got close to enumerating the full list. The two issues with the CPE are orthogonal to one another. Whether or not they lift the cap, they still perform traffic shaping when using their CPE (they also perform shaping /without/ their CPE, but it's much cruder).

Sorry, I missed that you addressed the request for details in another branch. I was hoping to see a description of what their CPE was actually doing - metadata-based traffic shaping, throttling after a certain amount of usage, too small NAT table like ATT routers were famous for, just underpowered performance, or what. But now I understand you just investigated enough for your own purposes.

FWIW I think the DiffServ RFC explicitly acknowledges that the QoS marks are generally ignored across administrative boundaries. It was an extension of the earlier 4-level QoS priority, so by that time it was clear that transit networks weren't paying them much attention. What would be surprising is finding a consumer ISP that actually did something different when a customer set them.

Take the Comcast CPE, put wifi router in small faraday cage or open it up and unplug the antennas?

Most of Silicon Valley is in the clutches of Comcast as well, unfortunately. It's 1.2gbps down / 35mbit up or ATT 768K DSL (and we'd have to pay to get phone lines) where I'm at. Of course, if you check the FCC Broadband Map, there are 7 providers here, with ATT showing 5gbps FTTH, which I've confirmed with their reps multiple times is not possible (there's no fiber on the poles in our backyard and they won't install any anytime soon).

Apartments I've lived in around SV as well are also under the Comcast exclusive contract, with restrictions on any antennas or dishes, so the only other choice would be tethering an LTE or 5G device.

We sold out the critical infrastructure of our country to investors and this is the result.