I think I've finally found a good IPv6 tunnel provider, SixXS:

Connecting to 2001:4f8:0:2::1e:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 275,666,325 (263M) [application/x-gzip]

100%[========================>] 275,666,325  835.46K/s    ETA 00:00

That's dax pulling a file from mirrors.isc.org.  Also, quick RTT:

[dax:14:57]% ping6 -c4 mirrors.isc.org                         [p8]
PING6(56=40+8+8 bytes) 2001:4830:1500:21::2 --> 2001:4f8:0:2::1e
16 bytes from 2001:4f8:0:2::1e, icmp_seq=0 hlim=59 time=73.616 ms
16 bytes from 2001:4f8:0:2::1e, icmp_seq=1 hlim=59 time=73.894 ms
16 bytes from 2001:4f8:0:2::1e, icmp_seq=2 hlim=59 time=73.475 ms
16 bytes from 2001:4f8:0:2::1e, icmp_seq=3 hlim=59 time=73.810 ms

--- mirrors.isc.org ping6 statistics ---
4 packets transmitted, 4 packets received, 0.0% packet loss
round-trip min/avg/max/std-dev = 73.475/73.699/73.894/0.164 ms

Awesome.

So, this evening wasn't too exciting.  Then in #xicada...

23:02 < prox> I still wanna run BGP, wahh.
23:04 < cyberconte> prox: lets

Now Xicada is split into two private autonomous systems, 65011 [cyberconte] and 65003 [me, and everyone else who's not running BGP].  Quagga handled the eBGP and iBGP connections decently, and the configuration wasn't too complex.  Although, for some reason Quagga reports the eBGP session as iBGP.  Gotta be a bug in bgpd, I'd think.

Now to convince the other two members to move to BGP, and get off the [pretty diverse] OSPF area 0.0.0.0 we've been using for awhile.  It'll be nice to be able to redistribute directly connected public netblocks via OSPF and not worry about nullrouting someone else :)

Also, this past weekend I redid the network again.  Converted my nat/router box here from a Celeron 333 w/Debian running iptables, to a Celeron 2.0 [the old starfire!] s/FreeBSD running pf.  Also threw four NICs in it, grabbed a third RR IP, and am now having pf load balance outgoing internet connectivity between my PIX 506 and NetScreen 5GT.  Yeah, I'm running OSPF on both of them, too.  I'll update my network diagram soon ... I promise.

Fun!

I'm getting clumsy in my old age.

Less than a month after getting my E815, it breaks.  Ok, well, at a hockey game today, I was juggling too much (nachos, drink, tickets, etc.) and tried to take a photo.  Ended up throwing the phone to the ground instead.  Completely toast, I can't seem to persuade it to power up.  Ironic, since I just enabled Bluetooth OBEX stuff last night and voided my warranty.

The Charlotte Checkers hockey game, on the other hand, was pretty fun.

http://news.bbc.co.uk/2/hi/technology/4504256.stm

Steve Jobs should be person of the year, no doubt.

So.. it doesn't look like FreeBSD does multiple routing tables.  Linux seems to do it, but I don't think Quagga is aware of it, and will probably mess it up.

Basically, I'd like the following:

bfe0: link to local lan
tun0: vpn tunnel to.. somewhere

xl0 is a member of the "external4" table
xl1 and bfe0 are members of the "internal4" table

table external4
------------------------------------
destination   type  gw          int
0/0           S     <via dhcp>  xl0
<some isp>    C     n/a         xl0

table internal4
------------------------------------
destination   type  gw          int
0/0           S     1.1.1.2     xl1
1/24          C     n/a         bfe0
2/24          O     1.255.255.2 tun0
3/24          O     1.255.255.2 tun0
[etc...]

Assume there's an Internet6 table in there somewhere, but I only need one of those so far.

Anyway, does anyone know an easy way of doing that in *BSD/Linux?

I'm really looking for some hierarchy with regard to routing tables and interfaces.  Interfaces being a member of only one routing table, etc.  Yeah, something like what NetScreen does, but with reverse path filtering.  Maybe even a /30 and two virtual interfaces that virtually connect vr's.  Oooh, fully-meshed virtual routers.. now I'm thinking..  Also babbling.

Picked up a Christmas tree today.  Turned out nice, now I need ornaments.

I picked up a Cisco PIX 506 today, too.  I hope I can get software for it that supports tab completion, otherwise it's gonna be frustrating ...

Interesting, they've got live lingerine models in Maine, now.

< toster> I can't really see that sort of stunt bringing in women to shop

Nope.

So I figured it was time for the nat box at my parents place to get upgraded from lilo/devfs/2.4 -> grub/udev/2.6.  This is all on Debian 3.1.  I've done it before successfully ... with two reboots, and a total of 20 minutes spent.  Since this box has 3 NICs (and a 2.4 -> 2.6 upgrade typically renumbers interfaces), I figured I'd setup an /etc/mactab and add the appropriate lines to /etc/init.d/networking to invoke /sbin/nameif.

I managed to first screw up my MBR with Grub (I've done it a dozen times before, you'd think I'd learn by now, right?) and burn like 30 minutes waiting for an Ubuntu LiveCD to boot Gnome, before I got anything done.  This machine has 128MiB of RAM, so loading up an xterm took 5 minutes and thrashed the disk.  I would have used a Gentoo LiveCD.. but I didn't have any handy, and no nat box to provide connectivity to obtain it.

Additionally, the nameif stuff didn't work at all.  All the network interfaces were detected, but after nameif ran (eg., after networking was started), most of the services complained and gave some weird errors.  OpenVPN just failed with a binding error of some sort, snmpd spewed errors about interface counters being invalid, and there was, erm, a lack of connectivity.

Ok, back to square one.  I added the modules (3c59x, ne2k-pci, b44) in the correct order to /etc/modules, rebooted, and hoped it might read that file first.  Wrong.

Just for giggles, I removed everything from /etc/modules just to see what order the system would bring up the interfaces, figuring I'd just bite the bullet and reconfigure daemons and firewall scripts.  System booted up - so I jotted down what ethX name went with what card, and started reconfiguring stuff.  Ended up rebooting again, and to my dismay... the names changed AGAIN.  It turned out that every time the box rebooted, the interface names would be different.  I was really hating hotplug/udev at this point.

So, it was suggested that I write my own udev rules, in the hopes that it would rename the interfaces the right way.  This resulted in some "cannot rename interface; File exists" error that I didn't bother looking into.

I finally broke down and added a few lines to /etc/init.d/udev and called it a day.

Yeah, this was a good portion of my Saturday afternoon.  At least stuff is running 2.6 here, now, not that I'm going to see any immediate benefit.

Back to work tomorrow, horray.