[Ecn-sane] The curious case of "cursed-ECN" steam downloads

[Sebastian Moeller]

Hi Jonathan

On 4 September 2023 00:36:03 CEST, Jonathan Morton <chromatix99 at gmail.com> wrote:
>> On 3 Sep, 2023, at 9:54 pm, Sebastian Moeller via Ecn-sane <ecn-sane at lists.bufferbloat.net> wrote:
>> 
>> B) Excessive ECT(1) marking (this happened with a multi-GB download)
>
>This *could* be a badly configured middlebox attempting to apply a DSCP, but clobbering the entire TOS byte instead of the (left justified) DSCP field.  

[SM] Mmmh, would that likely be DSCP decimal 1, aka LE? Or someone mapping e.g. a 3bit VLAN priority PCP directly onto the TOS field? What puzzles me is that from the same IPv6 I got one more data flow, 182 MB using ECT(0) and ECN... odd...

>Apparently Comcast just found a whole raft of these in their own network as part of rolling out L4S support.  Funny how they didn't notice them previously.

[SM] In retrospect re-dedicating bits out of an already used header field might have been a bit optimistic....

Regards
        Sebastian


>
> - Jonathan Morton
-- 
Sent from my Android device with K-9 Mail. Please excuse my brevity.