The Stack Overflow Podcast

This week's guest is Mark Russinovich, from SysInternals.com and now with Microsoft.

Chatrooms are chaotic! Jeff mentions that lots of spaces need editorial oversight. A lot of good information is available, but it's a hard to find it in the disorganizations. It's a chronic problem.

Mark and Joel talk about his command-line work. Mark had to reverse-engineer this stuff, almost from scratch. SoftICE was effectively a device driver that took control away from the OS, when it was active. Mark's become famous for being a Microsoft hacker (yes they exist) and for his work with rootkits, the problems with which are becoming an epidemic.

Mark started outside of Microsoft, but later his company was acquired by them. He's worked on Vista, Windows 7, and a bit of Windows 8, but is now on Windows Azure. For Azure, an OS for data centers, Mark works for the fabric controller team. Like the kernel in Windows, this defines processes and consumes application xml. Basically, he's all up and down the stack. One of their biggest concerns is upping consistency, to make Azure the best in the industry.

One of the project's other goals is to have a virtual machine deploy in less than 5 minutes, and update in 2 minutes or less. Right now, those times are 8-9 minutes at the 50th percentile. They're pursuing a variety of tactics to optimize the boot process. There are lots of moving parts to optimize. It's a fun project, and it's all new.

Not that many companies can deploy a cloud operating system at such a scale. Investment is expensive, although, as Jeff points out, machines today are more powerful than ever before. Still, although Stack Overflow is ranked #180, getting to #150 requires four times the traffic. Mark points out that yes, you can manage the servers yourself, make the investment, figure out all the parts, and so forth. Or, in nine minutes, you can upload your webapp to the cloud and pay only for what you use.

The cloud is best for companies who have traffic in bursts and periodic traffic. Companies where, say, there's a known holiday shopping rush or other specific types of workload patterns. By contrast, Stack Overflow's traffic is weirdly predictable. Mark notes that the other benefit to cloud computing is replication; if a disk fails (as 3-5% of them do annually) your data is cloned across the country.

Mark wrote a novel: Zero Day, which was published in March. It's a cyber thriller based around a cyber terrorism plot to bring down parts of the world using malware. It's readable and got lots of verisimillitude. The sequel, Trojan Horse is set to come out next fall.

Right now, while direct attacks are less common, spear-phishing (targeted phishing attacks) and good old exploitation of vulnerabilities in a system are still serious threats.

Jeff talks about the back-and-forth about putting anti-virus software on our servers. On the one hand, it's absolutely necessary, especially as Careers 2.0 has users uploading resumes and CVs onto the server. On the other hand, mention "anti-virus" in a Linux room and be prepared to get laughed out. There's also a serious performance question there.

Everyone should go implement 2-step verification on their email accounts (Gmail account!) right now. Well? Go! Do it now! We'll wait.

Mark says he would separate his password into tiers, with the top tier being ecommerce sites. Jeff says that this is part of why he's been pushing for third-party sign-ins, where the third party isn't a bunch of idiots. Mark believes we are converging towards this naturally, with the proliferation of Google and Facebook sign-ins.

Joel wonders if maybe there just aren't that many malevolent people in the world. Mark quickly counters with Facebook's admission that 600k logins are compromised daily.

He also points out that while our security is better (compare XP to Vista or 7's security hardening) the attacks are more sophisticated than ever. Just look at Stuxnet.

Be sure to ch...

Stack Exchange Podcast - Episode #24 w/ Eric Ries by The Stack Overflow Podcast

Jeff & Joel are joined this week by Eric Ries, author and expert on The Lean Startup.  Topics for the chat include:

Jeff Atwood is joining the podcast from his vacation. He has an announcement! He is having twins! In February! This will bring the total Atwood Child Count to 3, meaning they will outnumber the adults. Congratulations, Jeff!

Talk of children leads to talk of war which leads to talk of Battlefield 3. The core team spent some time playing today. It incentivizes working as a team!

ANYWAY. Eric Ries is our guest today! He's got a new book out called The Lean Startup. What is a Lean Startup? It's an analogy to lean manufacturing: a system of management about fast cycle time and building quality in from the beginning. Lean startups take those techniques and apply them to startups, where there are a lot more unknowns about the product and the customer.

Eric wants to convince Jeff and Joel not to batch deploy anymore. (We deploy multiple times per day. We have at least one per day, and other than that people can deploy as they see that they need to.) The discussion about the way the teams deploy changes leads to a discussion about unit testing.

Joel's criticism of lean startups: the combination of Lean Startups and the fact that any startup can get a huge amount of funding instantly leads to a lot of startups that seem to "pivot" an awful lot. Color is a classic example of this. Eric reminds us thatwinter is coming for entrepreneurship, and this might not be a problem much longer.

What is a pivot? A change in strategy without a change in vision. The key to the analogy is that in a pivot, one foot stays planted while you shift around to a new direction.

Innovation accounting is Eric's alternate accounting system that's designed to tell if you're getting close to product market fit. ROI, profitability, growth rates - all the traditional accounting metrics don't apply at the really early stages of a startup.

Joel's dream for the Stack Exchange Network is medical research. The problem is getting a critical mass of people together to make the site work. Currently, we branch into other verticals via "overlapping circles" - starting out with programmers who also have other hobbies.

Gaming is one of the biggest sites that has been created out of the "overlapping circles" theory. It's likely to be an excellent bridge between the existing community of programmers and civilians who also play games, so we are going to put time and effort into figuring out how exactly to make Gaming more awesome.

Eric is Mr. Pivot, so let's get back to that: Pivoting is not necessarily a mistake. It's the realization that a strategy that you used to be pursuing is working well for a specific customer base, and that you should pursue the parts of it that work. Gamers tend to pick a certain game to obsess over for a while, so it makes sense for the Gaming Stack Exchange to take a more specific approach to games as opposed to the Stack Overflow generalist approach.

If you could imagine having An Encyclopedia Of X, there could be a site about X. (There might not be an encyclopedia on Call of Duty, but there would be one about all games in general.) That leads to the generalist approach, which can get messy, but we allow users to participate in segmenting themselves.

Joel has decided to attack the Stack Overflow moderator flag queue. Some things he's noticed: There is a tendency to pile flags on people that don't speak English natively. The mental load on a new moderator can be very high as they learn the ropes of how to handle particular types of flags. Handling flags requires a lot of effort and decision from moderators. Joel has an idea on how to handle this! Discussion on moderation and flagging ensues.

Eric's book The Lean Startup, found on Amazon or right on Eric's website, peaked at #2 on the New York Times bestseller list!

Make sure to join us next week (at the usual time of course) when our...

Stack Exchange Podcast - Episode #23 w/ James Portnow by The Stack Overflow Podcast

Our guest today is James Portnow of Extra Credits. We are also joined in the studio by David Fullerton.

James Portnow is joining us! Extra Credits has been a thing for a few years. The idea struck back when James was working at Activision. He wanted to open up the conversation about game development and design to the consumer side, instead of continuing to speak in the industry-centric bubble.

At Stack Exchange, we're trying to make learning fun. All of the gamification that we do on the system is in service to the goal of making the internet a better place for learning.

Extra Credits did an episode about gaming addiction, which is related to the reason for the reputation cap on Stack Exchange sites.

Stack Exchange has sites for gamers and game developers! The Game Development site is distinct from Stack Overflow because developing a game is a bigger set of activities than just writing code.

Gamification is a way to get users to "read the manual", and get them to the point where they don't need the gamification aspects anymore at all.

Games like Simon and Dragon's Lair don't give you any choice or control. Games provide positive simulation in various ways - by feeling like you're acquiring a skill, by keeping things neat in Tetris, or on Stack Exchange, seeing somebody vote up something you wrote.

One Chance is a flash game with an interesting mechanic: it leaves a cookie that prevents you from playing the game again. It's an interesting concept on the bleeding edge of game design.

The dark side of gamification... is conditioned actions that make players continue to play FarmVille, slot machines, some MMOs, etc. Players become aware that they are not enjoying the experience, but they are compelled to continue nonetheless.

The danger in the Khan Academy is that for the American education system, this is the way to reduce our budget: have people record videos and have other people learn via these gamified websites. This is James's concern about the Khan Academy.

When gamifying education, everybody should start off at 1 and work up from 1 - not get docked points down from A+ or whatever. You also have to incentivize the class to help get each other's points up, not just each individual's own points. A high sense of agency is the sense of having control over your own existence and the world around you. When a student falls behind a little bit and does not feel like he or she can catch back up, they lose their sense of agency, and it becomes a monumental task to get the student back on track. Games teach us that outcome is directly related to our own actions, but with more instant results. (Programming is another way to demonstrate this direct impact.)

Joel peeled hard boiled eggs in the Israeli Army, so you can cross that off your Podcast Bingo card.

James is the hero in his own story. Games teach you that you can always win, and that nothing is unachievable. We will close on that hopeful note! James can be found @JamesPortnow or @ExtraCreditz on Twitter, or over at Extra Credits.

Oh, right, news from Stack Exchange: David, interim CTO while Jeff is on vacation, has no news. Except that we have a mascot now. (David had nothing to do with it.) Also, Jeff will be speaking at Oredev, which is November 7-11, and Punyon should probably go with him.

Oh, yeah! We have our own URL shortener! It's s.tk. Check out s.tk/joel and you'll pick up the gist.

Make sure to tune in next week when our guest is Eric Ries.

Stack Exchange Podcast - Episode #22 w/ Paul Biggar by The Stack Overflow Podcast

Joel (but no Jeff) is joined this week by Paul Biggar (who Joel originally met when he was a DevDays London 2009 speaker about scripting languages).  Paul currently works at Mozilla, having come off his own (not that successful) Y Combinator startup.

Paul's least favorite scripting language of all time is PHP. Paul works in static analysis, which is looking at a program that is not running, and making decisions about whether or not it will work, how to make it faster, what the security implications are. Paul has solved the Halting Problem... twice.

PHP stinks, so we talk about C and C++ for a while. Bjarne Stroustrup wrote a great book on the topic.

The people who love language design are not the people that are enthused by PHP, and they were scared off by the "poisonous community" (Paul's words!). The most popular programming languages that aren't very well designed: PHP, Perl, JavaScript, shell. Their creators "had no business designing languages". How did they become popular?

Haskell was a programming language that was well-designed but never gained any traction. Paul says there are two types of programming languages: those that start safe and try to build performance, and those that start performing well and try to build safety in. Haskell is the former. It "escaped" from academia... barely. F# comes from the same school of thought.

What about Dart)? Google released a spec. They've got a full implementation that's ready to go in Chrome.

The cool kids are using MongoDB, CoffeeScript, and tortoise shell glasses.

Enough about programing languages! Paul started a YC journalism startup called NewsTilt. It was the Future of Journalism, which is a terrible business to get into. Here's why it got shut down. In a nutshell: there were problems with the product, and problems with communication between Paul and his co-founder. Also, not being in Silicon Valley can be problematic... though Silicon Valley is not necessarily the be-all end-all of startup success. Perhaps most important was that it didn't solve a problem Paul really cared about.

Circle CI is a compiler-related startup that does capture Paul's interest. It's "continuous integration made easy"!

Paul didn't actually make the slides for his talk. But the message he wants to get out there is that working on compilers is actually very easy, and not something only wizards can do.

Paul can be found on Twitter @PaulBiggar, and at PaulBiggar.com.

Join us next week when our guest is James Portnow from Extra Credits - same place, same time.

Stack Exchange Podcast - Episode #21 w/ David Fullerton by The Stack Overflow Podcast

This week, Jeff & Joel are joined (in studio, no less) by David Fullerton, head of the NY Dev Department, and Jason Punyon, a developer here in the office.  Its a fast moving discussion covering all kinds of topics, like:

Stack Exchange 1.0 (which gave users wanting their own Q&A site the Stack Exchange software, without being official Stack sites) is touched on. Jeff discusses the clones that exist and their reason for existing.

Trello’s launch caused some kerfuffle on Web Apps.SE when general (and off-topic) help questions started being asked. In the larger sense, they discuss the necessity of applications and products to have their own unique help service.

Some recent changes made to Area 51 are discussed, including the restructured voting system for example questions. Joel discusses the problems that arose out of the previous method of judging example questions.

Fabian asks about overlapping proposals on Area 51, and David gives an overview of the process that goes along with the decision to merge proposals. Joel admits they aren’t too good at judging whether or not proposals are the right size.

Joel gives a call to arms for Area 51! They discuss the soon-to-launch Biblical Hermeneutics site and its relationship to the existing Christianity and Judaism sites.

Jeff brings up some other sites on Area 51, including LEGO and Firearms. The validity of a Healthcare IT site is discussed.

Alex just access to Stack Exchange’s real-time Google Analytics, so the traffic trends of the site are discussed.

Jeff plugs one of his pet proposals, and others discuss theirs (Krav Maga does indeed pop up).

Joel shifts the conversation over to Careers. You can visit Joel’s profile here! Careers and Stack Overflow aren’t integrated as well as they should be, and solutions to that are discussed. Jeff and David also talk about cool updates coming to Careers. On a sidenote, if you’ve ever wondered what it’s like to be an intern for Fog Creek, you’re in luck! David talks about that experience briefly.

Careers’ relationship with Linked In is mentioned and spurs on a wider discussion about the other career site.

The history of Careers’ filter is discussed, including how it ran originally and how it runs now.

Notifications on Stack Overflow have been modified recently. Jeff goes into the depth about how this was brought about. The term “yak-shaving” is involved.

A discussion about parenting questions on other Q&A sites reminds Jeff of a recent discussion on Parenting.SE, regarding the horrific-sounding “hot saucing”.

Join us next week at the usual time when we'll be joined by Paul Biggar and his wonderful Irish accent.

Stack Exchange Podcast - Episode #20 w/ John Siracusa by The Stack Overflow Podcast