After some digging I realized that this exception was happening because I was calling execute on a non-typesafe reference to an AsyncTask ie the reference had no generic parameters. I don’t think this should raise an exception but it looks like the Dalvik VM doesn’t like it. Alas. If you run into the same problem just make sure you’re calling execute on a type-safe reference.

(This should be obvious from the documentation but the documentation is very poorly organized.)

Stop making social networks, Facebook won
Frankly I don’t think Facebook has done much of anything beyond building a very good email client. In the big picture this is just not terribly interesting.

Digg v4 Troubles are Symptom of a Bigger Problem
Some surprisingly conservative, old-media thinking comes out of the folks at Digg.

Micro Instances for Amazon EC2 announced!
$15/mo is pretty impressive for that much ram and I suspect this price will only come down. I guess cheap personal servers are coming, just not in the form that anybody imagined.

A Taste of What’s New in the Updated App Store License Agreement and New Review Guidelines
It’s a start though I very much doubt Apple can give up its controlling, bullying ways. At least not this easily.

Checking in with Facebook
It’s difficult to see checking in as a positive thing but, I guess, in another light, perhaps it’s just another kind of publishing.

It’s somewhat difficult to believe. Frankly, most software systems out there simply aren’t that complex. Taking stuff out of the database and then putting it back is hardly brain surgery. Let’s face it: kids could do it.

And even the most complicated systems aren’t all that complex. There’s an important distinction between complex and complicated and while large software systems may be complicated as they involve many moving parts, integration points, and failure scenarios they are hardly complex like the weather or women and other phenomena that are truly unknowable and uncontrollable.

I’ve seen systems that are horribly difficult to understand and maintain but ultimately all of their behavior was fully captured in a precise set of computer instructions and this made it only a matter of time (sometimes months) before they could be handled.

And if complexity really was such a big problem you’d have to ask — well who’s fault at that? If software engineers are continually taking on more complexity than they can handle well, then, maybe they should stop doing that. Where’s the backlash, the drive for simplicity?

Perhaps “complexity” is just a code word, a vague way of admitting ignorance. And one does wonder what the programmers of the future who, undoubtedly, will be working on vastly more complicated systems, will think of such modern cries against complexity.

Simplicity Is Highly Overrated
Questionable anecdotal “evidence” aside I don’t buy into the simplicity vs flexibility distinction. In the most successful systems simplicity and flexibility drive one another.

Developing for Android
It’s somewhat charming to see people harp on user experience so much… but really, would the iPhone have gone anywhere if it hadn’t had a great browser? I don’t think user experience is nearly that important.

Some random thoughts on programming
Perhaps complexity isn’t the problem. If complexity was the problem then one would expect that software systems developed today would be less successful than those developed in the past. Is that really the case?

Is Transactional Programming Actually Easier?
It makes a lot of sense. It’s not so much that transactions are easier to understand than locks — they’re likely not — it’s that they’re simply more composable.

Re-thinking JDK 7
It’s not even really a debate. Waiting another 2 years would be disastrous.

As Digg Struggles, VP Of Engineering Is Shown The Door
I find it very strange that a company like Digg couldn’t do sufficient load testing to get some idea of their scaling requirements. Heck, what else are these web mega-sites good for?

Secure your codebase: OpenVPN in the (Rackspace) Cloud
Good to know. Will need to play around with OpenVPN one of these days.

State of HTML5 canvas in QtWebKit
It’s great to see the Qt guys continuing to innovate so relentlessly. If only somebody had done this with desktop Java…

I’m assuming that you have an Ubuntu server and a Windows client. The first step is to get git installed on the server. But before this can be done you’ll need to generate a SSH keypair. See the Ubuntu SSH keys tutorial for how to do this and once you’ve done so use WinSCP to grab a copy of your private key (the one that doesn’t end in .pub) and stash it in the C:\Documents and Settings\$USER\.ssh directory.

Now you can go ahead and install git and gitosis on your Ubuntu box. Be sure to initialize gitosis using the correct public key. Also note that gitosis may store your public key in a file named $USER@$HOST.pub in its keydir directory. This will only cause problems later so rename it to simply $USER.pub. You’ll also need to modify, commit and push gitosis.conf to refer simply to your username.

Now create a new repository in gitosis. Make sure you run the mysterious git remote add origin gitosis@$HOST:$PROJECT.git command and make sure you can then push the new origin to the remote master. You should now be able to modify, commit and push on the server. It’s time to setup the Windows client.

Install msysgit on your Windows box. The net install can be very flaky. If it breaks on you just delete the C:\msysgit directory and run the full install.

Now install Tortoise Git. When you do, as noted here, be sure to tell Tortoise Git to use the OpenSSH, Git default SSH Client and not TortoisePLink.

Now you can go ahead and clone the repo you created on the server using Tortoise Git. The repo URL will be of the form gitosis@$HOST:$PROJECT. Note that unlike the Linux git client, Tortoise Git doesn’t seem to like it when you add a .git to the end of the repo name. Also note that the username is gitosis. Technically you could use other usernames here, like your own, but then you won’t be able to share the repo.

Make some changes in the repo and commit. But if you try to push your changes back to the master you’ll get a very strange error message telling you you’ve already checked out the ‘master’ branch. In fact you have: on the server, when you initially created the repo. To fix this you’ll need to convert the repo on the server to a bare repository. Once this is done you can now happily commit and push away!

Enjoy!

It’s difficult to take the latest “Google made me stupid!” nonsense from Carr seriously. We’re supposed to believe that links are a “distraction”. A distraction from what? From the author’s message? But unless you’re reading Viagra spam the links were added by the author for the very purpose of increasing your understanding of her message. They contribute to the “meaning” of the text just as much, perhaps even more, as any other literary artifacts like punctuation and composition. Carr plays a transparent, naturalist game by trying to draw some deep, inherent distinction between the text and the links when no such distinction is meaningful. This leads to such silliness as putting links at the end of his blog posts so as to not distract his dear readers. (No, seriously, he does this — see the first link.) This ripping of the links out of the text is as reasonable as ripping out punctuation, capital letters, or paragraph breaks. So why stop at links? Perhaps Carr’s quest for pure, focused text will lead him to haiku. What’s next on the man’s hitlist? Does he even have a reliable process for identifying those artifacts that contribute to “cognitive load”?

Links have obvious benefits. Anybody who’s ever researched a topic on Wikipedia understands that links contribute significantly to a deeper understanding of the matter at hand by allowing the reader to place the idea in a context, to see how it’s connected with other ideas and events, to see not only the static, serialized form of the idea but grasp where it came from, how it moves and lives in its environment. Anybody who’s read a blog (like this one) quickly grasps how links can expose readers to contrarian viewpoints as well as introduce evidence that makes the argument stronger or weaker. And anybody who’s ever, say, browsed the web understands that links can be a powerful mechanism for discovery and exploration. Links are an extraordinary force for creating understanding, spreading knowledge, and learning new things. (See: the world wide web). And yet Carr suggests that links via the mechanism of cognitive load “impede comprehension and learning.”

What’s really going on here?

The cardinal sin of links is they make the reader think. Their crime is that they indicate a universe larger than the text. For people like Carr this is a real problem; reading is no longer a one-way transmission of wisdom, no longer a matter of a dumb receiver picking up a broadcast. With links the authority of the text is inherently compromised and the text is no longer, literally and figuratively, an end in and of itself. This is a pretty dangerous turn of events, particularly if you’re in the broadcasting business.

In Defense of Links, Part One: Nick Carr, hypertext and delinkification
This is such a dumb argument. Once you move past Carr’s trite, science-y “cognitive load” it should be clear to everybody that links, more often than not, only deepen one’s understanding of material.

This is how I feel about buying apps
This makes difficult sense. Most apps are worthless, not even use-once-throw-away valuable.

The problems with ACID, and how to fix them without going NoSQL
That’s actually very nifty. Such a scheme makes perfect sense — when you’re not using disks at all. But this doesn’t describe all the problems with ACID, notably the problem of having to split different types of data across many different systems.

Goodbye Blog. Hello Letter.
Really? This seems like a huge step backwards to me. But at least it’s good to see the “amateurs” gravitating towards paid content.

Rely your web startup on Rackspace Cloud? Think again.
I expect this kind of horror story to become increasingly common going forwards…

The Future of Silverlight
I find it interesting that developing serious functionality in Javascript is so universally loathed and yet also still so popular. It explains a lot about C, Windows 95, and the universe.