Found one problem so far - I specified my SPF record as v=spf1 mx -all. I then sent a test email to google, and it was rejected with:

Received-SPF: fail (google.com: domain of mstevens@mstevens.org does not designate 2001:ba8:1f1:f1ef::2 as permitted sender) client-ip=2001:ba8:1f1:f1ef::2;
Authentication-Results: mx.google.com; spf=hardfail (google.com: domain of mstevens@mstevens.org does not designate 2001:ba8:1f1:f1ef::2 as permitted sender) smtp.mail=mstevens@mstevens.org; dkim=pass (test mode) header.i=@mstevens.org

The mx defined for the domain is on IPv4 and IPv6, and has A and AAAA records. I was expecting the SPF record above to mark both as valid, but google doesn't seem to interpret it as valid.

Looking at the discussion at a thread on the SPF mailing list I think it probably should be considered valid, although I'm not certain. Anyway, I've updated the SPF record with a ip6 entry for the specific IP address, hopefully that'll sort it. Possibly there's a bug in the google implementation, but I wouldn't know who to contact!

Waiting for the DNS to propagate and see if the change helps.

iptables -A OUTPUT -p tcp --dport 1234 -d localhost -m owner ! --uid-owner root -j REJECT

Will restrict port 1234 on localhost to only be accessible by root.

Found at Paranoid Penguin: Using iptables for Local Security.

Administrator}=>mobile device list
Name           : ZTE_MF627 (preconfigured)
	Storage id     : 19D2:2000
	Modem id       : 19D2:0031
	modeswitch     : 55534243123456782000000080000c85010101180101010101000000000000
	AT interface   : 1
	Data interface : 3
	Force USB 1.1  : yes

	Name           : ZTE_MF100 (preconfigured)
	Storage id     : 19D2:2000
	Modem id       : 19D2:0017
	modeswitch     : 55534243123456782000000080000c85010101180101010101000000000000
	AT interface   : 1
	Data interface : 2
	Force USB 1.1  : yes

	Name           : HUAWEI_E1690 (preconfigured)
	Storage id     : 12D1:1446
	Modem id       : 12D1:1001
	modeswitch     : 55534243000000000000000000000011060000000000000000000000000000
	AT interface   : 2
	Data interface : 0
	Force USB 1.1  : yes

	Name           : Zoom_4595 
	Storage id     : 1C9E:F000
	Modem id       : 1C9E:9603
	modeswitch     : 55534243123456788000000080000606f50402527000000000000000000000
	AT interface   : 1
	Data interface : 2
	Force USB 1.1  : yes

	Name           : H_E220 
	Storage id     : 0000:0000
	Modem id       : 12D1:1003
	modeswitch     : 
	AT interface   : 1
	Data interface : 0
	Force USB 1.1  : yes

	Name           : H_E1725Cu 
	Storage id     : 12D1:1446
	Modem id       : 12D1:1417
	modeswitch     : 55534243000000000000000000000011060000000000000000000000000000
	AT interface   : 3
	Data interface : 0
	Force USB 1.1  : yes

	Name           : H_E180 
	Storage id     : 0000:0000
	Modem id       : 12D1:1003
	modeswitch     : 
	AT interface   : 1
	Data interface : 0
	Force USB 1.1  : yes

	Name           : Alcatel 
	Storage id     : 1BBB:F000
	Modem id       : 1BBB:0000
	modeswitch     : 55534243123456788000000080000606f50402527000000000000000000000
	AT interface   : 1
	Data interface : 3
	Force USB 1.1  : yes

	Name           : Telsey 
	Storage id     : 1C9E:F000
	Modem id       : 1C9E:9603
	modeswitch     : 55534243123456788000000080000606f50402527000000000000000000000
	AT interface   : 1
	Data interface : 2
	Force USB 1.1  : yes

	Name           : ZTE_MF110 
	Storage id     : 19D2:2000
	Modem id       : 19D2:0016
	modeswitch     : 55534243123456782000000080000c85010101180101010101000000000000
	AT interface   : 1
	Data interface : 2
	Force USB 1.1  : yes

	Name           : H_E1762 
	Storage id     : 12D1:1446
	Modem id       : 12D1:140C
	modeswitch     : 55534243000000000000000000000011060000000000000000000000000000
	AT interface   : 3
	Data interface : 0
	Force USB 1.1  : yes

	Name           : H_E1752Cu 
	Storage id     : 12D1:1446
	Modem id       : 12D1:140C
	modeswitch     : 55534243000000000000000000000011060000000000000000000000000000
	AT interface   : 3
	Data interface : 0
	Force USB 1.1  : yes

	Name           : H_E1752Cu_OV 
	Storage id     : 12D1:1446
	Modem id       : 12D1:1417
	modeswitch     : 55534243000000000000000000000011060000000000000000000000000000
	AT interface   : 3
	Data interface : 0
	Force USB 1.1  : yes

	No device detected

One of the things I wanted to try was getting the "3G backup" support working. This is not very well documented on the internet at the moment.

The best documentation I could find (with the help of #A&A on irc) was the Be user group Mobile internet document.

I tried this out with an old Vodafone dongle - labelled as a "Vodafone Mobile Connect Model K3565 - Rev 2", but which I believe is really a badged Huawei E220.

I backed up the configuration as recommended.

I logged in via telnet and tried to run the Be documented commands. I found I had to add an initial mobile ifadd as the umts interface did not exist:

{Administrator}=>mobile ifadd intf=umts
{Administrator}=>mobile ifconfig intf=umts apn=CHANGEME
{Administrator}=>ppp ifadd intf=mobilebroadband
{Administrator}=>ppp ifconfig intf=mobilebroadband dest=umts
{Administrator}=>nat ifconfig translation=enabled intf=mobilebroadband
{Administrator}=>ppp rtadd intf=mobilebroadband dst=0.0.0.0
{Administrator}=>exit

I then went to the web interface http://192.168.1.254/_pppom_cfg.lp?be=0&l0=2&l1=2&name=mobilebroadband - replace 192.168.1.254 with the IP address of your router, and entered the username, password, and APN. For my vodafone SIM, the username was web, the password was web, and the APN was pp.internet.

You should then be able to fiddle around with telling the interface to connect, and unplugging the DSL. The new connection should show up under "Broadband Connection - Internet Services".

I'm not sure how to make failover work, or if anything extra is required. Since I was using a vodafone dongle, and my machines are statically configured to use A&A's DNS servers, I lost working DNS.

Hopefully someone else can fill in the rest of the details to make things behave sensibly, and even try an A&A SIM.

I also tried with a Huawei E367 - doesn't seem to be recognised.

Should make things a little bit more secure, although I really need to do dnssec as well.

• The 1gb memory upgrade arrived from crucial today. Will see how this affects performance.
• DVI output seems to be working in X now, but bizarrely the video quality is much worse than VGA. Colours are wrong, and large areas of colour appeear to be "moving".
• Perl builds nicely and passes tests :)
• X is faster in 16 bit colour mode, but I'm wondering if that's related to my DVI problems.
• Firefox is by far the slowest application I've tried so far.

I've been trying to install following the instructions at How to Install, although strictly speaking they are for the Yeeloong rather than the Fuloong I'm using.

Discoveries so far:

• The machine comes with a RTL-8169 network interface. The wheezy netinst daily build images do not provide this driver. You can make the install work by downloading the full loongson kernel deb for wheezy, extracting r8169.ko, and packing it into the netinst initrd. Have reported this as a debian-installer bug.
• The installer sets you up to boot grub from pmon, and linux from there. This does not seem to work. Booting linux straight from pmon is very slow but seems to actually work.
• Whilst the system is moderately noisy, it's much quieter than discussions on irc had lead me to expect.
• The source I bought the system from shipped a EU power cable. It appears to be a standard "C5" laptop power cable, and I was able to buy a cheap UK replacement on Amazon.
• The DVI output does not seem to work correctly in console mode - I was unable to see the Debian installer or any boot information. This was fixed when I switched to a VGA cable.

This post is pretty good, as in fact is the entire blog.

• theinfo.org: A site to discuss large data sets, although it seems quiet
• Comprehensive Knowledge Archive Network
• UK govt data site
• The London Datastore
• Datawrangling: some datasets available on the web
• Ordnance Survey OpenData
• Wikipedia database dumps
• Project Gutenberg mirroring HOWTO
• Amazon Public data-sets
• data.gov
• USPS Address Information Systems products (okay, this is commercial)

There's obviously some duplication here in terms of sites linking to other sites, I'm highlighting stuff I thought was interesting. I'll probably update this post as I find new data that seems interesting.

Rough Type: Nicholas Carr's Blog: More evidence of Net's effect on the brain

• Free, Public Data Sets
• Hacker news discussion on previous link

I bought one of their OTP C100 one-time-password tokens to play with. They seem to implement the OATH standards, which works nicely with a variety of software.

I decided to try to get this working on debian with PAM as an authentication method.

So far the software I've tried to do this is oath-toolkit, which provides a convenient pam_oath PAM module to use. I've got this working on a home debian machine for testing, requiring a one-time-password from the dongle every time I login.

Rough steps to setup:

1. Install oath-toolkit as usual. You probably want to install from source or download the supplied debs, the latest version seems to be the thing to get.
2. Run ldconfig - this is one thing that's not mentioned in the instructions but was needed before my machine would see the new PAM module.
3. Put this in /etc/pam.d/common-auth:

   # this bit restricts oath checking to the specified user                        
   # you could use a group or whatever if you had more than 1 token                
   auth [default=1 success=ignore] pam_succeed_if.so quiet user = useryouwanttoauth
   auth requisite pam_oath.so usersfile=/etc/users.oath window=20 digits=6
   

   You should have created the users.oath file in the normal course of installing oath-toolkit.
4. sshd_config should have:

   ChallengeResponseAuthentication yes                                             
   PasswordAuthentication yes                                    
   

After that, things seem to work with at least sshd and su, but I'm still testing. Annoyingly, if there's a ssh authorized_key, it seems to override password authentication totally. Ideally I'd like to combine ssh keys and OTP, but I haven't worked that out yet.

I've also got one of the gooze OTP C200 tokens, which looks very nice, and uses time based passwords, which I'd really rather use, but it's not supported by oath-toolkit yet.

Personally I think the wood one comes out much more nicely (and there's some fine detail you can't see because my camera isn't good at closeups).

An example of the fine Laser Engraving Services from Andrews & Arnold.

First, define some new notification commands more or less like this:

define command{
	        command_name    notify-host-by-sms
			command_line    send-aa-sms --file=/etc/nagios3/aasms.config -d $CONTACTEMAIL$ -m "***** Nagios *****Notification Type: $NOTIFICATIONTYPE$ Host: $HOSTNAME$ State: $HOSTSTATE$ Address: $HOSTADDRESS$ Info: $HOSTOUTPUT$ Date/Time: $LONGDATETIME$"
}

# 'notify-service-by-email' command definition
define command{
	        command_name    notify-service-by-sms
		    command_line    send-aa-sms --file=/etc/nagios3/aasms.config -d $CONTACTEMAIL$ -m "***** Nagios *****Notification Type: $NOTIFICATIONTYPE$ Service: $SERVICEDESC$ Host: $HOSTALIAS$ Address: $HOSTADDRESS$ State: $SERVICESTATE$ Date/Time: $LONGDATETIME$ Additional Info: $SERVICEOUTPUT$"
}

/etc/nagios3/aasms.config should be replaced with an appropriate config file for aasms readable only by nagios. Then define a new contact:

define contact{
		contact_name                    mstevens-sms
		alias                           Michael Stevens (SMS)
		service_notification_period     24x7
		host_notification_period        24x7
		service_notification_options    w,u,c,r
		host_notification_options       d,r
		host_notifications_enabled      1
		service_notifications_enabled   1
		service_notification_commands   notify-service-by-sms
		host_notification_commands      notify-host-by-sms
		email							your_mobile_here
}

Just add this contact to appropriate groups, and you should now get SMS notification of outages.

The key advantage of all this over the curl example on the A&A site is that you don't have to put your account details in the process name, and they can live privately in a nice config file instead.

While I was there, noticed some very old versions of some modules I had on CPAN, and deleted them.

(the cynic in me suggests the recent surge in articles on this topic is related to the forthcoming book from Nicholas Carr)

(Yes, I know I keep going on about this every 5 minutes to anyone who'll stand still long enough)

Doesn't really do much yet, but put up a github repo to play with at http://github.com/mstevens/aasms.

I like the idea of union membership to some degree, but I thought, amongst other things, the phone call telling me to vote for Gordon Brown was the last straw.