Jeff and .NET

MySpace killed MySpace. Duh.

(This is a repost from my personal blog, but I figure the audience here is also, if not more, appropriate.)

I don't usually read anything that Robert Scoble says, because he's about as relevant to technology as my cats are, but the High Scalability blog linked to his asinine dissertation about why MySpace failed. Scoble thinks it was because of the platform and the location, or perhaps because people starting trendy companies don't use .NET. Pick a ridiculous reason, he probably has it.

I'll make the usual disclaimer that as someone who works for Microsoft, I might be biased toward the products it makes. I would add that I'm not shy about being critical of the company either. Naturally, my first urge is to defend the platform. Microsoft.com is the 9th most visited domain on the Interwebs, according to Quantcast, and it's clearly running on our platform. But the more I think about it, the more I think about how the platform is irrelevant.

MySpace has been in a death spiral because it sucks. It always sucked. Before Facebook opened its doors beyond universities, I was lucky enough to see it via much younger friends who were still in school. I immediately saw how useful it was. The day it opened to everyone, I knew MySpace was history unless they could be more like Facebook.

And we know how that turned out. If Scoble would have made one pass at his own piece, he would've seen that he had the core engineering problem related to their failure: "...They can’t change their technology to really make new features work or make dramatically new experiences." That's a sign of a tightly coupled system, and that's poor design. You can fail in this way using any platform.

The other reason for their failure is likely a people problem. That problem is obviously platform agnostic as well. My first exposure to MySpace from an engineering perspective was at Mix06, where they passed out invites to a big party, and one of the senior people/execs got up on the bar and encouraged people to apply for the army of .Net developers they were hiring. It was a pretty diverse crowd at that conference, but I can't say that I encountered many people running to sign up. That was hardly a great recruiting strategy. I was working for a startup myself at that point, and there was nothing compelling about what they were doing.

I consider myself a technology enthusiast, and I love how we constantly see new ideas hit the Web. I'm also realistic about what I see in front of me. The very honest truth is that most tech Startups go nowhere, or at best, have short lifespans. Things like Google, Facebook and Amazon are the rarest of exceptions. The most exciting stories to me are about smaller companies that solve some problem, even if it is a niche problem, and they build a sustainable business around it. There is no end game around selling or going public, they just do what they do. The best part is that their barrier to entry was much lower than it would have been 20 years ago, because of the Internet.

People in the tech bubble just don't understand this. People like Scoble gauge interest starting with location (is it in Silicon Valley?) and how they're funded (what round are they on?). It annoys me to no end. For all of the toxicity hurled at the 37signals guys for sounding like arrogant dicks, perhaps some of it deserved, they're absolutely right for focusing on the practice of building sustainable businesses, not taking investors money and seeing how much cash they can burn. I respect that.

POP Forums v9 Beta 2 for ASP.NET MVC 3 posted to CodePlex!

Get the new bits here: http://popforums.codeplex.com/releases/view/60571

This is the second beta for the ASP.NET MVC3 version of POP Forums. Next step is a release candidate. It is considered feature complete, and ready for testing and feedback. For previous release notes, look here, here, here and here.

Check out the live preview: http://preview.popforums.com/Forums

Setup instructions are on the home page of this project.

What's new in beta two? Since the first beta:

• Avatars and images are resized on upload if they're too big
• Rich text box is turned off from profile, or if you view in a browser id'ing itself as mobile
• Edit posts, see that they've been edited
• Delete posts (soft)
• First post preview from topic lists
• Show edit/moderation history to mods
• Quoted replies
• User-to-user e-mail
• Admin: IP history
• Admin: Approve user photos
• Show IP's to mods
• Home page stats
• Decouple app start from global.asax
• Private messages (one-to-many)
• Admin: Mail all users (subscribed), with opt-out links
• Various bug fixes, style tweaks and refactoring

What's next?

The app is running in production on MouseZoom today, so it's being actively tested. With this testing, and your feedback, we'll put out a release candidate next. Between now and then:

• Refactor where it makes sense to reduce friction
• UI refinement where appropriate
• Explore potential for a setup thingy
• See if it's possible to really make it an atomic thing that can be NuGet-ed

Other notes

So far, so good! This release brings it to feature parity with the crufty old Webforms version, in about half of the code. It's nice to be beyond science project status. The only thing really keeping me up at night is a desire to be able to split the app off with its own dependency injection, so you can easily roll with whatever DI container you'd like. Right now, if you want to use it in your MVC app, you pretty much have to use Ninject. While I dig it, it doesn't mean everyone does. If you've got any feedback or ideas about how to do it, I'd love to hear it in a discussion topic here on CodePlex.

As for a time line, I suspect that the RC isn't that far off. Stay tuned!

POP Forums will be at Mix!

If you've never been to Mix, you're missing out on what is arguably one of the best conferences that Microsoft does. I'm not just saying that because I work here... I felt that way before, having been to most of them. The breadth of people and disciplines make it a really exciting event that pushes it well beyond the "Redmond bubble," as I like to call it. You should go.

In any case, there's an Open Source Fest happening the night before Mix starts, on Monday, from 6 to 9 p.m. There will be people there representing a ton of great projects, some as enormous as Umbraco, as well as people doing SDK's, controls and other neat stuff. Best of all, you get to vote for your favorites. Unless your favorite is Orchard, because Microsoft is sponsoring that directly. Or if it's POP Forums, not because Microsoft is sponsoring it, but because that's where I work in my day job. No prizes for me! Come by and say hello. I think the app will be nearly final by then, and it's already running on MouseZoom, one of my little side projects.

The quality and diversity of open source projects around the Microsoft stack just keeps getting better. Our platform is also pretty great at running stuff we don't make. This will be a pretty exciting Mix. Can't wait!

Lessons from rewriting POP Forums for MVC, open source-like

It has been a ton of work, interrupted over the last two years by unemployment, moving, a baby, failing to sell houses and other life events, but it's really exciting to see POP Forums v9 coming together. I'm not even sure when I decided to really commit to it as an open source project, but working on the same team as the CodePlex folks probably had something to do with it. Moving along the roadmap I set for myself, the app is now running on a quasi-production site... we launched MouseZoom last weekend. (That's a post-beta 1 build of the forum. There's also some nifty Silverlight DeepZoom goodness on that site.)

I have to make a point to illustrate just how important starting over was for me. I started this forum thing for my sites in old ASP more than ten years ago. What a mess that stuff was, including SQL injection vulnerabilities and all kinds of crap. It went to ASP.NET in 2002, but even then, it felt a little too much like script. More than a year later, in 2003, I did an honest to goodness rewrite. If you've been in this business of writing code for any amount of time, you know how much you hate what you wrote a month ago, so just imagine that with seven years in between. The subsequent versions still carried a fair amount of crap, and that's why I had to start over, to make a clean break. Mind you, much of that crap is still running on some of my production sites in a stable manner, but it's a pain in the ass to maintain.

So with that clean break, there is much that I have learned. These are a few of those lessons, in no particular order...

Avoid shiny object syndrome
Over the years, I've embraced new things without bothering to ask myself why. I remember spending the better part of a year trying to adapt this app to use the membership and profile API's in ASP.NET, just because they were there. They didn't solve any known problem. Early on in this version, I dabbled in exotic ORM's, even though I already had the fundamental SQL that I knew worked. I bloated up the client side code with all kinds of jQuery UI and plugins just because, and it got in the way. All the new shiny can be distracting, and I've come to realize that I've allowed it to be a distraction most of my professional life.

Just query what you need
I've spent a lot of time over-thinking how to query data. In the SQL world, this means exotic joins, special caches, the read-update-commit loop of ORM's, etc. There are times when you have to remind yourself that you aren't Facebook, you'll never be Facebook, and that databases are in fact intended to serve data. In a lot of projects, back in the day, I used to have these big, rich data objects and pass them all over the place, through various application tiers, when in reality, all I needed was some ID from the entity. I try to be mindful of how many queries hit the database on a given request, but I don't obsess over it. I just get what I need.

Don't spend too much time worrying about your unit tests
If you've looked at any of the tests for POP Forums, you might offer an audible WTF. That's OK. There's a whole lot of mocking going on. In some cases, it points out where you're doing too much, and that's good for improving your design. In other cases it shows where your design sucks. But the biggest trap of unit testing is that you worry it should be prettier. That's a waste of time. When you write a test, in many cases before the production code, the important part is that you're testing the right thing. If you have to mock up a bunch of stuff to test the outcome, so be it, but it's not wasted time. You're still doing up the typical arrange-action-assert deal, and you'll be able to read that later if you need to.

Get back to your HTTP roots
ASP.NET Webforms did a reasonably decent job at abstracting us away from the stateless nature of the Web. A lot of people criticize it, but I think it all worked pretty well. These days, with MVC, jQuery, REST services, and what not, we've gone back to thinking about the wire. The nuts and bolts passing between our Web browser and server matters. This doesn't make things harder, in my opinion, it makes them easier. There is something incredibly freeing about how we approach development of Web apps now. HTTP is a really simple protocol, and the stuff we push through it, in particular HTML and JSON, are pretty simple too. The debugging points are really easy to trap and trace.

Premature optimization is premature
I'll go back to the data thing for a moment. I've been known to look at a particular action or use case and stress about the number of calls that are made to the database. I'm not suggesting that it's a bad thing to keep these in mind, but if you worry about it outside of the context of the actual impact, you're wasting time. For example, I query the database for last read times in a forum separately of the user and the list of forums. The impact on performance barely exists. If I put it under load, exceeding the kind of load I expect, it still barely has an impact. Then consider it only counts for logged in users. The context of this "inefficient" action is that it doesn't matter. Did I mention I won't be Facebook?

Solve your own problems first
This is another trap I've fallen into. I've often thought about what other people might need for some feature or aspect of the app. In other words, I was willing to make design decisions based on non-existent data. How stupid is that? When I decided to truly open source this thing, building for myself first was a stated design goal. This app has to server the audiences of CoasterBuzz, MouseZoom and other sites first. In this development scenario, you don't have access to mountains of usability studies or user focus groups. You have to start with what you know.

I'm sure there are other points I could make too. It has been a lot of fun to work on, and I look forward to evolving the UI as time goes on. That's where I hope to see more magic in the future.

Installing Visual Studio 2010 SP1 or Windows Phone tools in your VM (danger!)

If you've read my blog for any amount of time, you probably know that I tend to develop stuff in a Parallels VM on a Mac. It's how I roll. I like VM's because I can trash them and do really stupid things with beta software. That said, there is a pain point that doesn't seem that well documented when it comes to installing stuff in this scenario.

The WP7 tools, and SP1 for Visual Studio 2010 (perhaps only if you already have the WP7 tools installed, I'm not sure), do something strange on install. As if it weren't already a long and slow installation, for reasons I don't understand, the installer fires up an instance of Windows Phone Emulator. As you may already know, the emulator doesn't run in a VM, because it is itself a VM, apparently. What it will do is fire up your CPU, make your comprooder hot and make the fans blow harder.

I found this out accidentally, as I started the (slow) phone tool installation once, and walked away. An hour and a half later, I came back to find it hadn't finished. But it was hot and the CPU was pegged, so I fired up the task manager to find XDE.exe, the phone emulator, cranking away. I had to kill it several times, and eventually the install finished. It fired up just once in the SP1 install, but it still had the same hanging effect.

I can't for the life of me figure out why it does this. In a VM, I can connect the phone to it and use that, so I don't need the emulator. But this install, firing up the emulator, will make it choke until you kill the XDE.exe process. Watch out!