Well, I finally got sick of being the top google result in NZ for "Application error (rails)". So welcome to my new, shiny, ultra-simple blog. Written in perl. By me. It even sort of works.

Now all I need to do is blog about all the events I meant to blog about since mid-November.

The subject says it all, really. I submitted my MSc thesis on Wednesday afternoon and expect to start an exciting new job at Innaworks on Monday. So it's all go.

So I apt-get upgraded to Ubuntu Edgy Eft today, seeing as it was released and all.

Problems I've had:

• Cryptdisks has difficulty asking for a passphrase on boot
• X was broken (error "cannot find default font 'fixed'")

I was able to manually bring up my encrypted disk stuff, so that's not a killer, but X breaking was less fun.

It turns out that between Dapper and Edgy fonts have moved from /usr/share/X11/fonts to /usr/share/fonts/X11. As I had a customised xorg.conf pointing at the old locations, the X server had trouble finding the "fixed" font. A simple search-and-replace solved that -- working out what caused the problem was much more of a hassle.

So everyone seems to be releasing their unconstrained DSL pricing today.

• Telecom have had theirs out fir a while
• TelstraClear have too, although they require you to spend $10/mo on tolls
• Woosh have stuff up, courtesy of buying Quicksilver
• Orcon have some nicely priced new plans
• ihug have new plans confusingly tied to their new "buy your telecom landline through us" scheme
• Actrix also have new plans

And I'm sure I've missed some. Not that any of these exciting new plans seem that exciting. You're still forced to spend ~$40/mo on a phoneline.

My left arm is now cast-free, although still a bit sore.

Yay!

There don't seem to be any decent ones. There's the MT format, which seems to be a vague de facto standard, but it's a pile of text-based yuckiness. And there's RSS, which doesn't include comments. And there's, as far as I can tell... nothing else.

I care about this because the typo website seems to have died, and typo itself is enormously bloated -- it makes this machine rather slow -- so I'm thinking of switching. I'll already be stuck writing my own typo database -> some other format converter. Does it make sense to define an exciting new XML format which no one else uses, convert to that, and then write something that allows you to transform between it and MT import format? Or should I just give up and transform directly from icky typo DB to MT format and get on with $work?

It's not as if defining an XML format similar to the MT format but with the most egregious problems removed would be terribly difficult -- but would anyone else use it if I did?

Left forearm in plaster. Bike accident. Front brakes locked up. Probable scaphoid fracture. Typing speed somewhat reduced.

Because I keep on telling people about this and it works so well, I thought I'd drive up blog readership by mentioning it here.

Those of you who read my previous post on spam may well wonder how I cope with the false positives of RBLs, especially as I count "no reverse DNS" as an RBL.

The answer is surprisingly simple: I ignore the RBLs when the MX or A record domain in the envelope from header is near to the host trying to send me mail. This neatly solves the "small-business/geek on cable/DSL" problem -- they almost always send mail from their single IP, which is usually their mail or web server. It even solves the "incompetent admins at small organisations who don't send reverse DNS" problem, while still allowing me to assume that most mail from hosts without reverse DNS is spam.

I have a qpsmtpd plugin to do this that I really will get around to releasing at some point.

We had the Interface AGM. New exec! Which I am not on! But...

...as a joke, ages ago, people proposed that I be made the King of Interface. At the AGM, people actually voted on this. And it passed.

Then, it was noticed after the AGM that we probably needed a constitutional amendment for this. And that people had been told this would happen. And that the constitution allowed us to call an SGM, with no warning.

Therefore, subject to approval of the amendment passed at the SGM by VUWSA, I am now the King Of Interface.

So I've found out why mail to lists.dis.org.nz was not appearing.

My overcomplicated spam filtering scheme resides on a machine in the US. Why?

• Cheap bandwidth (I don't want to pay lots to recieve spam)
• Available IP addresses (need a port 25 for incoming mail)

Now, of course, our flat network contains several machines of mine that need to recieve mail. Usually this involves one host getting forward TCP port 25 traffic from the single public IP our major-telco ISP will give us, and forwarding it on to the relevant internal host.

But this is a pain to maintain. So, I thought, given that the mail all goes through the US anyway, and that it has a nice reliable tunnel to OCCAID via SixXS, giving it decent IPv6 connectivity: Why not send the mail for the new host back via IPv6?

This seemed simple. The VUW computer club, whose network I sporadically maintain, has a layer 2 tunneled network with a ::/64. I could easily extend that over a tunnel to the host at home. Assign myself a static address, add a static mail route, and there's one less thing to go wrong.

Or so I thought.

Of course, all I was doing was shifting points of failure. The computer club gets its IPv6 from the Computer Science people at the university, where I suspect I am the only person who actually uses IPv6. They get their IPv6 from a faceless major-telco, as an experimental service set up by an employee who has since left. IIRC, they get their IPv6 from another telco as an experimental service. You can see where this is going...

There's a wierd connectivity problem. Mail-filter in the US can't see host at home. They can both see the dancing turtle.

Now, if this was an IPv4 problem, I'd just complain loudly at the ISP at each end until it got fixed. People expect IPv4 to work — it's what they pay for. The only network service they expect more reliability from is the PSTN. But all of the IPv6 services this scheme is relying on are free or experimental. The lowest-bidder US colo company doesn't appear to be interested in implementing IPv6. The major-telco our flat connects has the experimental serivce mentioned earlier, but there doesn't seem to be any interest in moving it into production.

The problem is further compounded by us not having a Telecom phone line, and thus not being able to sign up to one of the many minnow ISPs who would provide us internet service over it, one of which seems to be doing at least something with IPv6.

We have a TelstraClear cable connection, but they're even worse at wholesaling connections over their CATV network than Telecom is at providing usable UBS connections. Only a few Wellington-based minnows provide service over it, and all on outdated plans (2Mbps/512Kbps, vs the current 4Mbps/2Mbps retail plans TelstraClear sells), at rates higher than TelstraClear's retail prices.

So: We can't get a connection from an ISP who seems even slightly serious about IPv6 without paying Telecom $30/mo for a phone line we don't want, and even then we'd only get service over their overloaded DSL network. Suck.

We may be in range of Araneo's network, but the only ISPs listing their pricing for delivering over Araneo at 2Mbps want at least $199/mo, four times what our evil major-telco residential connection costs.

It was suggested at the session on IPv6 at NZNOG that v6 wasn't going to be deployed until customers asked for it. So who are we residential customers, held hostage by the major telcos, supposed to ask? And would anyone listen?