[Cerowrt-devel] back in the bloatlab

[Cerowrt-devel] back in the bloatlab

[Dave Taht]

[-- Attachment #1: Type: text/plain, Size: 2202 bytes --]

I put in some time testing the below version of cerowrt

http://huchra.bufferbloat.net/~cero1/3.3/3.3-rc6-2/

back in the bloatlab, which, among other nice things, has a dedicated, real
/60
ipv6 allocation. I had a chance to fiddle with encrypted wireless (works),
the existing mesh network (works), radvd (fiddly), and also setup the new
quagga as a 'leaf' babel node, which worked pretty good.

Used as a ipv4 gateway, the default aqms start kicking in a little earlier
than
I'd like when running at gigabit... but that was about it.

I also spent some time running a pure ipv6 network - proxying the web
with polipo, using ipv6 enabled sip tools like linphone, using ipv6 based
dns forwarders,
my email server was already ipv6 enabled, chat.freenode.net is ipv6
enabled,
and (aside from missing skype and having to proxy git) worked quite well. I
hardly
noticed there was no ipv4 in the network at all, and in fact plan to stay
pure ipv6
while I do some more tests in this fashion.

There only appears to be one (minor) ipv6 related bug remaining to be
squashed in this kernel release, which was just submitted upstream and I
hope would appear in 3.3rc7.  (all routers multicast fix) So it's my hope
that kernel related concerns can stop being a priority and I can focus on
integrating some packages, and scripting, and making the thing
independently buildable again, and fixing some bugs much further up the
stack, notably dhcp-pd.

I have a concern with a reported bug on encapsulation that might need
addressing.

And I do wish I could get QFQ working in this release. I've begun to
suspect it's not QFQ, but a bug in one of the filters... QFQ was useful in
creating the 1/d shaper code...

And I'm hoping that this:

"ICMPv6 RA: ndisc_router_discovery() failed to add default route."

Is merely an interaction with zebra.

default via fe80::a021:b7ff:feac:e446 dev eth0  proto zebra  metric 1024

Aside from that I'm getting pretty happy with where things stand and hope
that inside of 2 months this sucker will be worthy of running as a default
gw router.

-- 
Dave Täht
SKYPE: davetaht
US Tel: 1-239-829-5608
http://www.bufferbloat.net

[-- Attachment #2: Type: text/html, Size: 2501 bytes --]

Re: [Cerowrt-devel] back in the bloatlab

[Dave Taht]

For those foolish or weird enough to want to try living in a purely
ipv6 world, I've put up a preliminary document at:

http://www.bufferbloat.net/projects/cerowrt/wiki/Pure_IPv6_based_networking

-- 
Dave Täht
SKYPE: davetaht
US Tel: 1-239-829-5608
http://www.bufferbloat.net