3 ms·

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

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).

jandrese 3y ago
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".
- mindslight 3y ago
  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?
- seanp2k2 3y ago
  I wonder how Hulu streaming traffic is QoS'd on Comcast. Someone should check that out and file a lawsuit.
tristor 3y ago
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).
- mindslight 3y ago
  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.

jandrese 3y ago

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".

mindslight 3y ago
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?
seanp2k2 3y ago
I wonder how Hulu streaming traffic is QoS'd on Comcast. Someone should check that out and file a lawsuit.

tristor 3y ago

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).

mindslight 3y ago
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.