The Broken Web

11 PM November 27, 2007

OSDC day 2 kicked off with Rasmus Lerdorf talking about security problems on the web. Highlights so far are:

  • Pointing out the that many browser plugins are URL handlers. If you can find a bug in a plugin that handles a URL, then you can compromise users clicking on that kind of link.
  • Live XSS hacking on http://www.liberal.org.au, the conference hotel’s website and a few others
  • The JS-Yamanner worm.
  • IE6 has some interesting problems, including:
    • in UTF-8, it interprets the character 0xE0 as the first byte of a 3 byte sequence. If you can inject an E0 into a data field and get it to redisplay on the browser, the browser will ignore the next two bytes, which may be an attribute close quote.
    • it will run Javascript source in an image tag: <img src=“http://evil.com/hack.js ”>.
  • Until everyone upgrades their Acrobat plugins, it’s possible to execute arbitrary Javascript in the context of a site serving a PDF, or even the user’s local machine. The only safe way to serve PDF files is from a domain other than your own.
  • Use crumbs to stop XSRF problems. A crumb is a unique piece of data put into each web form. When the user submits the form, the server should check that the crumb is the one that is expected.
By alang | # | Comments (0)
(Posted to Software Development)

In which I write a Letter To The Editor

8 PM November 27, 2007

Andrew Bolt presented very calmly on Insiders on Sunday, and I thought, “Hooray! The culture wars” are over. But no. Today Andrew Bolt is back denying the Stolen Generations.

In summary, his argument is that, because Robert Manne couldn’t name ten people out of the 100,000 in the Stolen Generations, the Stolen Generations must be a myth. And because of that myth, Aboriginal people don’t trust the government to take their children away, even when it’s obviously the right thing to do.

Great argument, except that the Stolen Generations are real. Many, many children were really, in actual life, taken from their families for no reason apart from a smug, racist belief that it would be better to raise them in White culture.

It got my blood boiling. I wrote a Letter To The Editor:

I was stunned to read Andrew Bolt deny the tragedy of the Stolen Generations. If Aboriginal communities have lost respect for their government, it is because of actual mistreatment suffered by older relatives and friends, not because a bunch of latte-sipping, academic, city dwellers told them a reconstructed history.

As Andrew Bolt points out, lack of respect for the government is hindering the intervention. Regaining that respect will be a long process, but it must start with acknowledging past wrongs and apologising for them. Saying “sorry” is the first step toward helping these little children.

The sooner people like Andrew Bolt start to ground their rants in reality, the better off we’ll all be.


And while I’m at it, I’d just like to point to an interview with Alexander Downer in which he describes the Northern Territory intervention in terms that the uncharitable might take to mean that it was just grubby vote-grabbing.

By alang | # | Comments (1)
(Posted to Rants)

The Consumer View of Technology

3 AM November 27, 2007

Well, it’s after lunch, and straight into The Consumer View of Technology by Steve Ellis and Cherie Carbines from OpenMedia, a New Zealand company that’s build a MythTV based PVR.

There seems to be problems around communicating with customers. Both explicit and implicit communication needs to be clear.

  • There’s many ways to mis-communicate with customers. Moral: be careful.
    • One of OpenMedia’s early case designs looked like an amplifier so customers were confused and thought it was an amplifier.
    • Another example: broadcasters in NZ advertised “High Resolution” services and customers thought it meant “High Definition”, but it wasn’t.
    • OpenMedia’s literature said the hardware was based on a standard PC, so some customers were planning to use the PVR as their PC.
    • Be careful what you promise to customers when they request features. Even if you can deliver, you might have a world of pain supporting it.
    • Don’t be afraid to say what your product doesn’t do. It can save a lot of disappointing confusion for your customers.
    • When selling a new piece of hardware, product lifetime is a customer’s consideration. We expect fridges to last 10 years. Phones we only expect to last a few years. Is your new product a fridge or a phone?
  • Eye candy is important for sales. Quote: “Blue lights sell products.”
  • Even the cheapest routers come with 24×7 support.
    • Do you want to put out a product that customers expect 24×7 support with?
    • Email and web support don’t work. People want to talk to People.
    • Poor support can ruin your reputation. (Cue clips from The IT Crowd)
  • Customers expect a printed manual, in the box. Pictures are helpful for the 20% of customers that read the docs. But too much documentation is intimidating.

BYAG means “Because You’re A Geek”. Consumers are your bread and butter, not the geeks.

Great talk. Valuable content that can be applied to online apps too.

By alang | # | Comments (0)
(Posted to Software Development)

Behaviour Driven Design

1 AM November 27, 2007

Tom Adams’ Better testing through Behaviour on the topic of Behaviour Driven Development. BDD is an offshoot of “Test Driven Development” and Domain Driven Development. The central idea is that you first specify the behaviour that you’d like your code to have, in the form of tests, and then implement that specification.

Tom presented a bunch of examples from his Instinct Java test framework. Instinct differentiates itself from JUnit by paying a lot of attention to producing readable test cases – readable in the sense that you can make sense of it in English. Because of this, the test cases end up more like specifications. In the example on the Instinct website, he has a class named “AnEmptyStack” that tests the behaviour of empty stacks. This class has a method named “mustBeEmpty” that tests that the stack’s isEmpty() method returns true. You would read this as, “An empty stack must be empty”.

In comparing to TDD and JUnit some of the points (I think I heard) Tom made were:

  • xUnit tends to work only at the code level, whereas Instinct and BDD tend to work both on high level stories and at the code level.
  • Instinct and BDD dosn’t guide you so much toward strict units of code, but into adding behaviour wherever it’s needed to support the specification. In cases where you do need very clearly defined APIs – such as when providing public APIs to distinct processes or libraries, you will need to take extra care.
  • Instinct does tend to drive more usable APIs.
  • Instinct makes for more readable tests than JUnit, and plays nicer with IDEs.
  • Because of the language, Instinct is a valuable resource for developer-developer and developer-analyst communications.

I think I’d like to try Instinct on one of my home projects, then perhaps use it in the office.

By alang | # | Comments (0)
(Posted to Software Development and Java)

Here at the Open Source Developer's Conference

12 AM November 27, 2007

OSDC just started. Great to catch up with Keith and Mark. It’s great that Google could support the conference by sponsoring the dinner, but it feels little weird having “Sponsor” on my name badge.

The conference opened with a nice little speech by Scott Penrose, chair of last year’s OSDC committee, passing the baton to Arjen Lentz and the new committee.

Then it was straight into a keynote, Rusty Russell explaining what it is that he likes about C, that it is close to the machine. By way of enticing the PHP, Perl and Python programmers in the audience, he also demonstrated that C can do shell scripts too, by way of the Tiny C Compiler, tcc. I found the talk an interesting insight into hard-core C development.

Right now I’m listening to Ian Clatworthy of Canonical explaining why I might want to use a distributed VCS. I can see that there might be advantages, and I can see how it better fits the open source development model. I’m not sure there would be overwhelming benefits in the environments where I do most of my work. He’s saying that he sees a market for only three big distributed VCS systems – Bazaar, Mercurial and git – but that they all have “maturity” issues at the moment, and don’t yet have the kind of tool support many developers expect. Oh, and Bazaar is just a few weeks away from a 1.0 launch.

By alang | # | Comments (1)
(Posted to Software Development and Python)
© 2003-2006 Alan Green