In terms of "running code", I have a few updates for homenet.

1) Codel and FQ_codel in openwrt/cerowrt mainline and upcoming linux 3.5

Van Jacobson and Kathie Nichols new RED replacement, "Codel" is out
and showing great results. (in the fq_codel variant especially).

Their paper, "controlling queue delay", the ns2, ns3 sims, and code
are linked to at:

http://www.bufferbloat.net/projects/codel/wiki

Cerowrt has it as it's default qdisc, openwrt has it in their qos
subsystem only (as yet)

Codel is dual BSD/GPL licensed to give more folk a leg up on it.
(fq_codel is GPL, but multiple
other technologies for fair queuing exist)

Reducing latency and fixing bufferbloat on homenet devices, for
ethernet, seems feasible now, although much work remains at multiple
layers of the stack (and wifi needs major work especially)

We are trying to have some data in time for ietf but I suppose most of
that is of interest to other working groups.


2) jg recently outlined a plan for better aqm and classification for
everything (which applies especially to home gateways)

https://gettys.wordpress.com/2012/06/26/the-internet-is-screwed-up-and-how-to-fix-it/

I've spent part of the last year pushing better classification into
multiple daemons (quagga/dnsmasq/rsync/etc), have patches for
transmission in my queue -
in addition to fixing bufferbloat, etc, but the above link is the
first top level overview of that low-level-incremental thing we've
been doing that has existed and could use some more brain cells on it.

recently we had a debate on classifying DNS queries as CS5

2) On the routing/security front, the first ever (so far as I know)
implementation of secure RIP and babel has landed in quagga-RE,
implementing  RFC4822 (and borrowing some stuff from RFC6506). To me,
this is a huge advance on the state of the art, making possible ip
level security in a routed network.

code: https://github.com/Quagga-RE/

Support for hsls, olsr and batman is also migrating into quagga:

http://lists.quagga.net/pipermail/quagga-dev/2012-June/009438.html

My overall goal for the quagga work is to be able to align with the
competing ospf ideas here on homenet using the same daemon
infrastructure and may the best protocol win.

3) Border detection, have quite a few thoughts, look forward to
reading the draft

4) I have made no progress on naming. Experiments with routed
multicast DNS were less than satisfactory across multiple domains.
getting bind to work well requires nsupdate integration with
everything.

And lastly, Cerowrt-3.3.8-6 is out, testing tons of new ideas, and
proving pretty stable, with -7 adding in the new quagga this weekend.

-- 
Dave Täht

Blatant Plug:

http://www.bufferbloat.net/projects/cerowrt/wiki - "3.3.8-6 is out
with fq_codel!"
_______________________________________________
homenet mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/homenet

Reply via email to