I have had this patch in the cerowrt kernel for a while and despite
fiddling with it, most of what it attempts to measure is thus far not
useful for bufferbloat, and I can get most of that with a userspace
tool such as netperf and/or some hooks in the kernel.
Now, maybe it can be made useful, but it tinkles all over the tcp fast
path, requires maintenance, and I just don’t feel like porting it
forward to 3.1 or 3.2, or waiting for someone else to do so. In
theory, it looked good. In reality, it’s hard to care.
I’m glad I tried it, I’m glad that it’s out there and I could try it
again, but I’d just as soon live without it.
If jacobo has found a use for it, I’m all ears.