[Battlemesh] glad to be here
dave.taht at gmail.com
Mon Jul 8 12:35:02 CEST 2019
I just wanted to say how happy I am to be surrounded by like minded
folk here at battlemesh. I am still recovering from an epic battle
the bufferbloat.net folk had with the US cable industry at the march
ietf meeting over the proper use of the "last bit of the internet":
industry wanted to use up the ect1 bit in the IP header for a specialized
service the basically imposes "fast and slow" lanes on the internet. This
proposal had meandered through the ietf for years, seemingly going nowhere,
when in december I got word (through an ECO) that they were actually going
to go ahead with it, and expected the IETF to bless it, lacking a
counterproposal and an organized group opposing it.
Their "L4S" specification is wrapped up in a pretty little bow, chock
full of mis-reprensentations and outright lies about how good it will
be for the internet, with the actual details buried in appendixes to
the relevant standards drafts.
Thankfully my group had an alternate proposal nearly ready to go,
called "SCE" or "some congestion experienced", which was network
neutral and integrates great with the existing fq_codel deployment. So
we burned the midnight oil and 2 months preparing that proposal and
code, and I melted my credit cards flying everyone relevant in, we did
a great preso at the ietf tsvwg wg meeting, and basically did a DDOS
at every other working group, asking difficult questions of the
opposition that reduced them to babeling incoherently at several
(joyful) points. So we stopped a 40b/yr industry, in it's tracks, from
screwing up the internet, at least temporarily.
But it was still essentially a guerilla action against an army.
The major stage of that struggle within the IETF got written up in
lwn, here: https://lwn.net/Articles/783673/
Which also goes into more technical details than I present here.
I got wiped out physically, financially, and mentally, and spent a few
months on a beach in barcelona, unable to use the internet without
falling into a rageknot, playing guitar endlessly until I started to
feel better, which
wasn't until a few weeks ago. So glad Reinhard sent me a plane ticket!
The "low latency docsis" spec, still exists. The attempt to take over
the "last bit of the internet" continues.
The network neuatrality folk have not woken up (yet), the political situation
in the USA is still friendly to the attempt and the battle within the ietf
takes place again next week, with a raft of new (and hopefully technically
superior) folk, alerted to the problem, all over it, and I never intend
to go to ietf again, but to keep working the technical problems and...
... playing guitar. which the airline lost on the trip from SF, at
least temporarily. Anyone else have one?
And all I want to do is fix bufferbloat everywhere, this battle is a
distraction from that - and it still remains to be seen if our
counterproposal for a better form of lossless congestion control (SCE)
will work well enough or not for worldwide deployment as big as
fq_codel has become. I kind of hope to test some of that this week,
Also I've had some adventures in adding 240m new IPv4 addresses to the internet
for general use in the "unicast extensions project", with john gilmore
(of eff fame) and a recent encounter with spacex's starlink satellite
internet project I'd like to talk about at some point.
CTO, TekLibre, LLC
More information about the Battlemesh