In-Depth

Mr. Windows

Jim Allchin, considered by many as the father of NT and the brains behind Longhorn, talks about the future of rich clients and why you should trust Microsoft to power your next PC.

• By Doug Barney
• 08/01/2005

Jim Allchin, Microsoft's group vice president of platforms, may be running a multi-billion dollar business, but at heart he freely admits, "I'm just a geek."

Reporting directly to CEO Steve Ballmer, Allchin is responsible for Windows client and server; the next-generation client and server, code-named Longhorn; and the entire Windows Server System. No wonder 7:30 a.m. to 7:30 p.m. is a typical Allchin office day.

Allchin wasn't always a geek. He was once a professional musician, though the professional part is debatable. Allchin says when his food stamps ran out, he'd eat Cheerios with no milk.

But software was in his blood. He earned a Ph.D. in computer science and turned his attention to networking, serving as the principal architect for Banyan VINES.

In his 15 years at Microsoft, Allchin has earned a reputation as a straight shooter who admits when Microsoft makes mistakes. But he's also a fierce competitor who pushed the integration of IE with Windows 98 to thwart Netscape inroads, a move that earned Allchin a center seat at the U.S. government's anti-trust case. Later, Allchin implied that open source was "un-American," and was skewered repeatedly by Linux fans across the world.

Today, Allchin is focused on quality. He avoids travel and prefers to spend "think days" in his lab looking at software. Ironically, that obsession with quality almost kept Allchin away from Redmond. Unimpressed with MS-DOS, early Windows and OS/2, it took him a year to agree to come to Microsoft.

A Work in Progress
Redmond: How will the new visualization and organization capabilities in Longhorn change the way end users interact with their computers? Is Microsoft copying the Mac in this regard?

Allchin: Longhorn will make it easier for people to organize their information and data in ways that are much more visual and natural than today. Today people are just flooded with information and it's so hard to sort through to find what's really important and relevant to them. It's not just the Web. It's pictures, songs, docs, e-mails, spreadsheets and so forth. When you search for something today, you get a list of files, and you're not sure if what you get back is really useful.

One approach we are using is to visualize what is in a file or folder using the actual content within the files, instead of more static icons which just relate to the file type. It takes icons to a whole new level. For example, file icons become "snapshots" of the document and folders can visually show these snapshots even when a folder is not open, to help a user differentiate the content. Instead of seeing the icon for Microsoft Word, you see the first page of the document. So the user gets instant visual cues about the information contained in the files.

Another issue deals with organization. Today you have one choice: folders. And unless you make multiple copies of the files (which creates a whole new set of issues), you can only have that file in one folder. Our solution is to create virtual folders and allow users to query for information based on content or attributes.

So, in Longhorn we're attempting to provide users help in addressing the information overload problem. When we add capabilities like this, we look to our customers for what they want and need and to our research group for innovative technologies that can address those needs. It's not about looking at competitors.

It's important to note that Longhorn is still a work in progress. The first beta is due out this summer and will influence many decisions about Beta 2, as well as the final product.

Redmond: I have a Windows XP laptop. There are a lot of ways that it's infinitely better than the Windows 98 desktop I ran six years ago. But sometimes it hangs or drags in an inexplicable way that means it's just as slow as the old machine. How much potential is there for faster starts and better performance on the Longhorn client?

Allchin: I'm obsessed with quality, which of course performance is a big part of. I think we did a lot to improve quality in Windows XP, and even more with Windows XP SP2 and Windows Server 2003 SP1. I'm proud of the work we've done so far. But yes, we can do even better. With Longhorn we're focusing obsessively on the fundamentals.

Problems on XP could be us, but often it's third-party code that prevents normal operation of the system in some way. My view is that we need to prevent this from happening to the maximum degree possible. In Longhorn, this is a key focus for us. We are creating isolation layers so that applications cannot negatively affect the system the same way as before.

One thing I've wanted for a long time is 'instant on.' Longhorn will support S4 hardware, so the PC can go into a very low power state so it seems off, and the resume time is very fast—in most cases, one to two seconds. In terms of other performance issues, we're doing something called 'SuperFetch,' which means that Longhorn will keep track of the files and data most likely to be accessed and store it in RAM rather than on slower disks—even under heavy system loads. In addition, we're ensuring that disks are automatically optimized without a user having to do anything.

"Is there more code [in Longhorn]? Yes. But it is better code, it is better-organized code and it is better-tested code."

Safe, Simple, Sexy
Redmond: What's going to be the 'I've gotta have that!' technology in Longhorn?

Allchin: I asked this question of some IT managers when I was in New York last fall. And the answer had nothing to do with fancy features. They want security, reliability and easier deployment. That was it. They said that would be enough for them. They wanted to save money. Our No. 1 priority with Longhorn is nailing the fundamentals—security, reliability, deployment, migration and so on.

Consumers will of course benefit from that in addition to businesses. I think what consumers will love most about it is that it's cool. I mean it's really cool. It looks cool, it's easier to use, more powerful and we think there will be some cool new applications taking advantage of the new platform capabilities. We're even working with hardware vendors on making hardware not only work better but look sleeker and fit better into the environment, whether that be the living room, the office or in a mobile scenario. We talked about what we're doing with data visualization, but in the end I think people will want Longhorn just for the quality of the fundamentals. Safe, simple and sexy. Just remember that.

Faith in Engineering
Redmond: In a survey done by our magazine, a large majority (about 70 percent) of respondents believe that the complexity of Microsoft's operating systems makes them more vulnerable. Is Longhorn going to add to or reduce the complexity of the OS?

Allchin: Complexity doesn't have to mean more vulnerabilities. In fact, I think the record is on our side even today that we have fewer issued updates. And complexity doesn't have to mean harder to use, either. In fact, a car today is much more complex than the cars I drove when I grew up—yet they are much simpler to drive and safer! I believe the same is true with software.

I do believe you have to continually improve your quality bar. Certainly, we are in the middle of doing a major re-engineering of our engineering 'factory' today. Engineering Excellence is like a religion now. Get it right the first time. We asked a team from our research group to join Windows that was focused, among other things, on automated testing tools. People are good at testing, but computers are much faster and better at it. These tools don't catch issues after the fact; they catch problems in the code via source analysis, such as the buffer overruns that cause security vulnerabilities, before they get checked into the source control system. All developers have to run a suite of automated tests and pass them before they can add code. We call these quality gates. The results have been fantastic—we're reducing the amount of time it takes to produce a test version of Windows by orders of magnitude, which speaks to the higher quality of the code and how we've reduced interdependencies between different parts of the system.

We have a group of architects, some of the best engineers in the world, working across the development team to strengthen the underlying architecture of the system. They've used tools from our research group to trace the layers in the code. We can then move code around to different modules, in effect componentizing the system, which makes development simpler. This type of isolation will make a big difference in terms of allowing us to focus our testing and find bugs like security vulnerabilities before they ever get into the main build. I truly believe we're innovating the way software is built like no one else. Is there more code [in Longhorn]? Yes. But it is better code, it is better-organized code and it is better-tested code. We're more disciplined than we've ever been.

Redmond: What's the status of the WinFS project?

Allchin: Work continues and we're on track to deliver the first beta of WinFS when we launch Longhorn.

Redmond: Microsoft's long-held vision has been a PC on every desk, which in the industrialized world has largely been realized. How do you bring that vision to lesser-developed countries with far less income?

Allchin: I could joke and say that you're right … we clearly thought too small in the past. How about a PC for every person? Seriously, that is a new dream of ours. PCs will continue to morph into amazing devices and I think this dream is something that we should aspire to.

Last year we introduced Windows XP Starter Edition, which is a localized Windows offering targeted at first-time computer users in developing technology markets. We worked in partnership with international governments to create these. These versions of Windows have customized feature sets and a lower price-point. We're also making it easier to have shared computers as this is often how people get their first experience with a PC; for example, at school or in an Internet café. We know that in rural villages this may be the way we can first help people who are worried about weather affecting their crops or when they need medical advice or help.

"We clearly thought too small in the past. How about a PC for every person?"

The 64-Bit Question
Redmond: How can 64-bit change the computing landscape?

Allchin: The shift to 64-bit is just huge. The shift from 16-bit to 32-bit was huge. We had 10 years of 16-bit, 10-years of 32-bit, and the next decade is going to be all about 64-bit. With the 16- to 32-bit shift we had some hiccups around compatibility, moving from segmented memory spaces to linear address spaces, plus we had one UI for 16-bit Windows and another for NT. The transition to 64-bit will be much smoother, which may be why it's not getting as much attention as the shift to 32-bits did. The 64-bit machines run 32-bit apps and generally they run them faster. At some point, all new machines will be 64-bit, first on the server and then on the client. It'll just happen. It's inevitable. In fact, we believe the transition to 64-bit will happen faster than most people expect. We think x64 will see the fastest adoption of any new processor architecture. Once the machines are present I think you'll see over the next decade lots of interesting uses of the additional address space—not just on servers, but on the client as well.

Redmond: Windows Server 2003 is stable, secure and scalable. What new frontiers are there to explore?

Allchin: Windows Server 2003 is a product I'm very proud of. We're getting a great response to it from our enterprise customers. One of my favorite features is role-based deployment, where we configure settings automatically based on whether it's going to be a file/print server, an Exchange Server and so on. This really was just a beginning though. We're going to go farther with this scenario-based approach. For example, we have the Small Business Server now, and we're working on a server for medium-sized businesses. We'll tailor administration to those companies and the types of problems they need to solve. Another big focus area is automation—super wicked intelligence to make servers run themselves, self-diagnostic, self-healing and so on. Then there will be massive parallelization with multi-core.

One of the most exciting areas is the interconnection of servers with services, or federation, so that servers at different companies can work together. The world is becoming more interdependent; it's a global economy with lots of suppliers located everywhere you can imagine. Companies have already put some of their business interactions online, and it's all happened at an incredibly fast pace. But still, inter-enterprise work is quite hard to do today. For example, say you want to share a document with a partner. Today you e-mail it back and forth. But there's no way to maintain control of it, to make sure they don't forward it, etc. Plus, what if you're both editing it; who has the latest version? There's lots of room for improvement and I'm excited about the work we're doing here.

It's the People, Stupid
Redmond: What are the biggest challenges for the Windows platform from both a technical and competitive perspective?

Allchin: The world keeps changing and Windows keeps changing. It is a journey. Today, our next step is to nail security and safety. And even though I know there will always be 'bad' people in the world, we need to help our customers protect themselves. We also need to continue on our re-engineering process and the work we're doing on the architecture of the code. This will let us be more agile, ship more innovation and do it with even higher quality.

"We've put technology first for too long. We have to start with people."

From a competitive perspective, I believe that the focus of the industry needs to change. It's interesting that you mentioned 'technical challenges.' I don't think our biggest challenges are technical. If you look at the progress of the industry over the past 25 years, the advancements in technology have just been staggering. But there comes a point when more technology isn't enough for people. I've been doing some work with Joe Pine, one of the authors of The Experience Economy, and I think he has it right—the next wave of economic growth is about experiences. What am I trying to do or accomplish with technology and how much am I enjoying it? How well is technology helping me connect to my passions? We've put technology first for too long. We have to start with people. That's how companies will be able to differentiate themselves. If you sell PCs based on technical specs, they all start to look the same. But if you combine hardware, software and services into unique offerings that people want and can only get from you, then you've got a leg up on the competition. We are following this philosophy with Longhorn, and I've been encouraging our partners to do the same. Our biggest challenge is how we connect with people on a personal level. How do we make them see the PC and devices as more than just tools but essential to a productive and enjoyable life? Our ecosystem is a large one, with a lot of players. I wouldn't have it any other way, because I want our customers to have choice, but we have some work to do to align companies around a holistic approach to creating compelling and immersive experiences for businesses and consumers alike.

Redmond: How will virtualization technologies change the way computers are deployed and licensed over the next few years?

Allchin: Virtualization technology comes in many types.

Any OS that supports virtual memory is doing 'virtualization' of memory. Today, generally it allows customers to run multiple server or desktop operating systems on one computer, meaning they can be much more cost-efficient in their hardware deployment. I hear it discussed primarily for server consolidation. For example, Virtual Server 2005 enables customers to use one piece of server hardware to run one copy of Windows Server 2003 to host business applications, other instances of Windows Server for infrastructure capabilities such as file serving, and still others to run test or development environments. The resulting benefit is hardware cost savings through consolidation and more complete utilization.

Patch Strategy
Redmond: The monthly patch cycle has been in force now for about 18 months. What's working and not working with that process?

Allchin: In the 18 months since we moved to a predictable monthly release schedule, customer response has been very positive. We made the move to a monthly update cycle because a lot of customers told us that updating needed to be more reliable and predictable. Since we made this change, we've seen a 400 percent increase in Windows Update usage and an increase of over 320 percent in Automatic Updates usage. So I think the monthly update cycle has been quite successful. I think our transparency and consistency has been quite appreciated.

In November of last year, we started an Advance Notification Program for enterprise customers. Three business days before we release a monthly security bulletin we publicize the maximum number of bulletins, maximum severity rating and products affected. This helps IT administrators plan their resources accordingly for the following week. The information we provide is general in nature so that it doesn't disclose vulnerability details that hackers can get a hold of and put customers at risk. Overall, the process is going really well, but we're continuing in our efforts to make updating as painless as possible. For example, Longhorn will require fewer reboots to apply patches so we believe IT and consumers will be able to update their machines in a more silent and unobtrusive way.

I should be clear, though, that our goal is to reduce the number of updates while at the same time making any updates reliable, silent and automated to deploy.