Alf Watt at WWDC: The Journey from iStumbler to Apple and Beyond

Each year at WWDC, TMO interviews a few Apple developers who want to tell their story. The result is usually a number of serious insights into the state of mind of the developer community. In our eighth interview, Dave Hamilton chats with Alf Watt famous for iStumbler on the Mac.


Dave Hamilton: You’ve had a varied and interesting career. Most of TMO’s readers are going to know you from iStumbler. But what many TMO readers may not realize is that while you were developing iStumbler and giving it away for free, you were working for a company that was making a lot of the wireless technology that TMO’s readers were using.

Alf Watt: This is true. I spent five years working for the Fruit company. I was working in the Mac hardware group – which is responsible for the Wi-Fi client for OS X. So that includes the menu extra, the system preferences, the utilities and a whole bunch of stuff in the background that no one ever sees.

Actually, I started iStumbler 10 years ago.

TMO: I was wondering which came first.

AW: Well, one thing leads to another. I distinctly remember a Macworld Expo in New York [1999] where Steve Jobs got up and demoed the iBook for the first time. Between the slightly hucksterish presentation, the reality of what he showed off there, here’s this base station and here’s this laptop and everything’s integrated, and it’s all wireless, was something we’d never seen before.

The antenna systems were built in, the cards were built in, there’s nothing to attach, no dongles, no weird equipment to buy and configure. It all worked seamlessly.

TMO: And it did work!

AW: It was amazing. I ran out and got one and started playing with it. And it just really captivated me. The technology, the ease of use, the ... magic ... if you will of being able to pick the thing up [iBook] and walk around with it and still be surfing the Internet.

But being a geek at heart, I really wanted to know how it worked. And so, like a lot of really good products, iStumbler is written for me. I wrote it so that I could understand what was going on. I really wanted to wrap my head around this stuff, and back in Mac OS 9, there was actually an AppleScript API. That was actually the first iStumbler that I wrote. It just an AppleScript that sat in a loop and read the network names, and read them out loud, using a synthesized voice.

I was living in Manhattan, very briefly. I would put on a pair of headphones and walk up and down the streets, to the West Village, listening to network names. This was just a few years into the Wi-Fi revolution, not even two years. And I realized that if you could stitch these things together, my gosh, you could have just about continuous coverage all the way around the block. Of course that hasn’t happened. We went in a very, very different direction on the industry.

But there a lot of little companies that looked at that and tried to do it -- I was involved beiefly with Sputnik in the early days

Fon is the company that has come closest to this notion that, hey, we’ve all got these backhauls, let’s just link them altogether and share. Wouldn’t that be nice? So, in order for me to move some of these concepts forward and understand them better, I needed good tools. And so I think my whole career up to this point, after 10 years, has been focused on taking wireless, which is this invisible, mysterious, indistinguishable from magic technology and making it visible. And debuggable is maybe the biggest thing.

TMO: Your tools, at the least the ones I’ve used, some that I didn’t know you had written [inside OS X], allow us to troubleshoot.

An early iStumbler version

AW: Working on the Mac OS client, one of the first things we did, of the three things I did at Apple that I can talk about, is that there is a public API for Wi-Fi in the Mac. So, coming in -- I had been working on iStumbler for five years -- when you sort of sign your soul away to the devil, if you will, when you go to work for Apple, one of the things they’re very clear about is, thou shalt not use any private APIs for anything. And iStumbler relied on private APIs.

So, it took awhile to get it done, [but there was] an escalation all the way up to a V.P. and took some effort to get it published, but we do have a public API in the Mac OS client for Wi-Fi access now, so you can legitimately look at it. And that allowed me to, after a few years of quiet, publish a version of iStumbler that worked on recent Macs. Which is huge.

TMO: Very interesting.

AW: The thing about that, you know, I did it for maybe selfish reasons, if you go look in the Mac App Store now, there are close to a dozen of these apps that rely on this API, some of which are commercial, some really good ones: inSSIDer from the Metageek guys, gives you a really great, visual display of where the networks are and how much power is in the signal.

Of all the things I did, that is the one that I like to think back to.

TMO: You lived the geek’s dream! You had an itch, you scratched it a little bit, then you went to the source and fixed it so that everybody could scratch this itch.

AW: And the goal has always been the same for me. The wireless user experience, which is what I was responsible for a little while, really needs to be ... nothing. We haven’t gotten there as an industry, but there are efforts in place, we’re working in it really hard, the Hotspot 2.0 guys are pushing it closer.

But when it comes down to it, networking user experience is connectivity, performance and reliability. Those are the three things that people really care about more than anything else, from any communications technology. And we’re moving the industry so heavily towards everything wireless all the time, that we have to have really good tools in place to achieve that.

One of the things that really stood out in the [Monday] Keynote, was the 802.11ac product introduction. I’m going directly from here to the Apple store to buy a unit and tear it down. I haven’t actually seen the antenna arrays in the new AirPort base station. I have to get my grubby little hands on one.

But the tools that Apple produced to show energy usage per application are critical to bring visibility into this. If you’re unconscious about how much bandwidth your application is using, you’re just going to keep using it until a customer says, “Hey!” But if you have the tools available as you’re developing it – developers will naturally take it upon themselves to self-correct.

Being able to supply those tools, enabling the market place for those tools, I hope in some small way, has improved the lot of people trying to use their laptops to get their work done.

The absolute last thing that anybody wants to do is fiddle with configuration settings [as we used to do with VPN], or adjust parameters or any of that.

TMO: Well, some of us like to do that! But there’s times when when don’t want to have to.

AW: The times when I don’t want to have to is when people call me. Another one of my goals is: how to I reduce my own personal support burden to none. It never gets there of course. But, overall, people just get a better user experience.

TMO: Now I have a technical question for you. I figure, if anyone has the definite answer, it's you! If you're running a dual band access point, using 2.4 and 5 [GHz], no “ac” involved yet, should the SSID for 2.4 and 5 be the same or different?

AW: They should be the same. You want to have an Extended Service Set (ESS). And the reason being, because there’s a couple of things going on there, 2.4 and 5 have different propagation characteristics which is fancy way of saying that 2.4 [GHz] goes further. They also have different bandwidth limitations. In 2.4, the size of the channels we’re allowed to use us 20 MHz, call it two lanes if you will, in 5 GHz, we have 4 lanes, 40 MHz.

802.11ac changes that equation significantly in that we’re going to have 80 and 160 MHz channels eventually. The upshot of that is that you get incredible amounts of bandwidth. The balance point of that is that you’re consuming more spectrum. In 2.4 GHz we have three [discrete] channels, but in 5 GHz — it varies by region — we have many, many more. In short, if you start looking at the number of available channels as you start doubling those channel widths, you’re pushing 5 GHz closer to where 2.4 is, which has been a bit of a concern for me, is that we’re chasing the bandwidth numbers so fast that we’re willing to sacrifice spectrum.

Spectrum is the precious resource of our times. In fact spectrum management and spectrum policy is is going to start becoming a bigger and bigger political issue. We’re already seeing how Wi-Fi has changed the way the FCC does spectrum allocation. We’re getting more spectrum for Wi-Fi, we’re filling in blocks that we didn’t have before, gaps in the 5 GHz spectrum. It’s because we, in the industry have been able to take the so-called junk spectrum and turn it into something genuinely valuable.

Figuring out how to cooperatively manage that spectrum is going to be a big deal for the rest of human history. The goal is figuring out how to move as much data as we can with as little power as possible. That’s the next thing. Besides spectrum usage.

The disadvantage of those wider bandwidths is that they consume significant amounts of computing power. So, it used to be in radio systems that you paid to transmit, but receiving was free. What’s changed since 802.11g came into Wi-Fi is that we now pay more to receive because the power used to decode the signal has become more significant than the power usage to amplify the signal going out.

TMO: Interesting!

AW: That means significant amounts of power on the Wi-Fi chip just to pull the data stream out of the constellation. That change is significant. And that scales more or less linearly with channel width. Because of the way Orthogonal Frequency Division Multiplexing (OFDM) works is instead of having a single carrier that covers, say 20 MHz, which would have been typical for 802.11b, they break it down with subcarriers. So instead of one big curve, you’ve got a bunch of little curves packed in. On a spectrum chart, you’ll see this nice clean curve. The beauty of OFDM is is they they’re able to get into those bandwidth corners by packing a bunch of small waveforms right next to each other. It’s called a Bart Simpson curve because it looks like Bart Simpson’s head. All those little peaks across the top.

So this is spectral efficiency at work, but it costs it terms of compute power. each one of those subcarriers has to go through an FFT block which are relatively expensive computationally as things go. Recall, 802.11a, when it first came out, which was the first OFDM Wi-Fi standard, had a reputation for being a power hog for this reason. So we’ve gotten a lot better at that. So that’s one of the things with 802.11ac, even if your access point point is willing to set up a 160 MHz channel for you, we’re not going to go past 40 or 80 MHz on mobile devices.

TMO: We’d just burn our batteries up.

AW: Right, and the CPUs on the devices aren’t able to handle that much bandwidth. Period. Just think about the data rate that that represents. Sure, they’re dual core, one GHz machines, but they’re still CPU bound, memory-bandwidth bound. So we’re at an interesting point where we have the ability to push a lot of bandwidth, but we don’t really have a low power [receiving] solution just yet.

TMO: So this is the answer to my question, and I’m glad we went there. This is interesting stuff.

AW: Some of this is going to end up on the cutting room floor, [not yet!] but the key thing there ... I am a really strong believer in flat networks.

TMO: In a home, that’s easy to do.

AW: Now, we’re getting AirDrop on iPhones. We worked really hard on AirDrop on the Mac, I wasn’t directly involved with it, but some people on my team just outdid themselves with some amazing work. That technology is interesting for two reasons. One because it’s the first, commercial, peer-to-peer Wi-Fi connection protocol. It grew out of Wi-Fi Direct which was a semi, not yet totally failed effort to make Wi-Fi a lot more like Bluetooth. At least in terms of its ability to pair.

The thing that I always come back to with AirDrop is, working for a Wi-Fi equipment provider is, the degree to which the industry has gone down the road of ... your network ... my network ... never shall they meet. Except maybe up in the cloud. Think of all the times you’ve been standing right next to someone, and Dropbox was the only reliable way to get data from you to that person. It’s really an absurdity.

This has all been done in the name of security, but now at the same time, we’re working around that with AirDrop and Wi-Fi Direct and Miracast, and all these similar protocols.

But as these peer-to-peer services start to kick in, a lot of networks are going to see this as rogue access points and their intrusion detection systems will start to kick in. But the other more critical problem is, going back to spectrum, that the infrastructure doesn’t have the opportunity to give you hints about what channels to use. So spectrum and capacity management is going to become more and more critical.

TMO: It almost seems like these AirDrop clients need to advertise. Hey, I’m doing this peer-to-peer thing, don’t worry about the channel I’ve chosen, it won’t be in use for long.

AW: Actually, there is a Wi-Fi standard called TLDS that supports the infrastructure to give you a hint. Recommend another channel that’s relatively clear.

TMO: I wanted to ask you about your current employer.

AW: Sure. You know, what was really interesting about the Keynote was the introduction of the 802.11ac product line. And the new form factor in the new base station is particularly intriguing because they have a “beam forming antenna array” in there.

Apple's new 802.11ac AirPort (Image Credit: Apple)

The company that I work for now, Ruckus Wireless, easily has one of the most advanced antenna systems on the market. They’ve been doing beam forming since the beginning and have easily the best performing products on the market because of this technology. Wi-Fi depends really heavily on good antennas, good hardware design, and some of our systems have as many as eight antenna elements. User selectable and steerable. We have switchable ground planes and all sorts of amazing little things, so we can actually take the same RF energy that you would use to cover, say, one room and focus it in one direction very narrowly. And dynamically, packet by packet, per client.

For example, we don’t know, spatially, where each client is, but we do know which antennas are serving them best. And so we just keep track and we say, “Oh, you’re doing really well on this set of antennas, so we’re going to send data that way.” For the most part, it’s really a question of figuring out what combination of antenna sub-elements and ground plane power works best for each client. Some heavy duty math there.

It all allows our products to have much better range. So when customers do a deployment, they’re buying fewer access points. They do cost a little more, but we like to think the quality is worth it.

TMO: Are these products anyone would use in their homes? Or maybe a small office? Or is it targeted a little beyond that.

AW: It’s all very enterprise driven. The company is moving towards service provider and carrier networks. So there’s incredible interest from the AT&Ts and Verizons of the world, in terms of deploying Wi-Fi networks as adjuncts to their 4G network.

[Some side discussion went on the cutting room floor.]

Getting back to Ruckus, we make, I think, the highest performance access points in the industry. We’re pushing into the service provider market so that no longer will you have to curse your carrier because their 4G network is overloaded because they’re going to hand you off seamlessly into a local Wi-Fi network. Which is going to have a lot more bandwidth than that 4G tower.

So the industry started with big radio towers, then [analog] cellular, then we went to digital cellular and then Wi-Fi. The trend is higher and higher bandwidth with smaller and smaller cell sizes. With more and more radios. I think the obvious thing to do there is go find a company that sells radios. Follow the physics.

TMO: Who’s making the actual boxes?

AW: My job for Ruckus is that I manage a group that creates mobile applications. So we have a few developers and some partners that we work with to produce apps that help people plan and set up their wireless deployments, apps that help you monitor the deployment once it’s up and running, apps for performance and tuning, some of my favorites. We’re supporting the hardware products by making them as easy as possible to set up. We want to have enterprise grade deployments but with the same ease of use that you expect from the AirPort base station.

Most importantly, we want to give you visibility into how the network is performing and how the users are experiencing it. A real-time view of what’s happening with the network.

And coming from the world of the client, and I did with Mac OS devices, it’s an interesting view of the world because the infrastructure sees everything. I’m used to having just one radio ...

So, you think, aw, I gotta take it off channel to scan, and that’s going to impact my backhaul performance. In the most recent version of iStumbler, I took away the slider that allows you control how often it scans. Because, I get people saying: “Hey, you’re consuming this much of my total bandwidth, while I’m trying [to do something else]!”. So my response, just let me take care of that for you. So the current beta tries to be as unobtrusive as possible. while the app is in the foreground, it’s scanning relatively frequently, but when you push into the background, it goes into a kind of, “Yeah, we’ll get to that” mode. It’s adaptive, knows whether it has focus.

TMO: So no longer is it overly costly to leave iStumbler running all the time.

AW: That’s the goal. That tuning will probably go on for over a decade. [laughter.]

TMO: I’m just glad to hear that iStumbler development will continue unabated.

AW: The problem with having a toy project on your own is that there’s no product manager around to tell you “Cut and ship!” So I got really close to the current betas to shipping, but then I thought, “You know, why don’t I just overhaul the entire user interface?”

TMO: Yeah! Why not?

AW: So I really have some great stuff queued up in iStumbler, but it has to take a back seat to the day job. But it’s been ten years now of making the invisible wonders of Wi-Fi visible. Ruckus is an amazing place to work.

TMO: You get to do the stuff you love to do.

AW: That’s what it’s all about. If you do the job you love, you’ll never hate your job.

TMO: It’s been an absolute pleasure.

AW: How much over our limit did we go?

TMO: You don’t want to know. John’s going to kill me.

John Martellaro:  It was actually quite terrific!