A Conversation with Metrowerks’ Greg Galanos

General Magic is a tease. After releasing one of the most exciting software platforms in recent memory, the company was ill-prepared to meet the demand placed on it by the developer community. Because of this, becoming a General Magic developer required proving to the company that your application was, dare I say, insanely great.

All this changed with the May, 1995 availability of CodeWarrior Magic, a complete Magic Cap development environment packaged and supported by Macintosh tool maker Metrowerks. But who is Metrowerks and what is its vision for personal communicators? Certainly very well known in the Macintosh community, Metrowerks may require a bit of an introduction for others. And the man best positioned to explain this is Greg Galanos, president and CEO of the Austin, Texas-based toolmaker extraordinaire.

Compilers, Compilers, Compilers

Galanos started by describing the genesis of the company: “I started Metrowerks back in 1986, but essentially, actively at the start of 1988. We focused in the first year on educational software, most notably compilers for Modula 2. That was back in the early years, between 1988 and 1990.”

“But in 1991 when IBM, Apple, and Motorola announced the PowerPC alliance, I foresaw a point in the market where there would be a technical discontinuity between the general purpose 68K-based processors and the RISC-based processors, along with an absence of software tools to make the transition. So around 1991, I asked Jean Belanger to come on board to take care of the financing and marketing for the company, while I took care of the technology issues. We decided at that time to obsolete all the software we had, and go directly into the mainstream C and C++ business.”

Galanos recalled: “We acquired the compiler technology in Germany and, from that base, we hired senior architects who had left Symantec. We started working towards building a suite of tools consisting of C, C++, and Pascal, and we wanted to hit the market six months before the machine [PowerMac] shipped, which was our target.”

Galanos and team started seeding large companies with their new technology. He explained: “We did that with some of the largest companies, such as Adobe, because we really wanted some very large software programs to run through the compiler to be reassured of the industrial quality of the toolset. And in January, we shipped the first release.”

RISC Not So Risky

Galanos noted: “Today we have 23,000 registered users, and over 80% of all software on Power Macintosh is built using CodeWarrior. That number is probably closer to 90%, and we had to make some decisions back in ’94 as to how we wanted to diversify the business.”

He continued: “By 1994, we had built a 68000 back-end, a PowerPC back end, and we instituted in 1994, a project to build our x86 back-end. So we had, in the pipeline, three 32-bit processor technologies.”

Having heard Jean Belanger, chairman of Metrowerks, talk earlier about the importance of the embedded processor world—in which processors can now be found in devices ranging from CD players to microwave ovens—I was curious about Metrowerks’ involvement. Galanos stated: “There is a trend in the industry to move all four and eight-bit microprocessors and microcontrollers to 32-bit technology. And the reason for this is that manufacturers want to move off CISC [Complex Instruction Set Computing] and on to RISC [Reduced Instruction Set Computing] as fast as possible.”

He elaborated: “What a lot of people don’t realize is that there are economic reasons pushing the move to RISC, and the major factor is that when you are on a reduced instruction set computer, you are essentially moving the complexity of the processor technology of the silicon and into the compiler. By doing so, you are reducing the number of changes you need to make in the masks when a bug occurs on a chip.”

Galanos noted: “The problem with CISC technology is that every time you have to make a change on a mask, you have to stop the line and the cost of the change is a minimum of fifty million dollars. What happens also is that if you’re making changes to a mask, you are also affecting the yield, which drives up the price in the end product.”

“So there’s a huge push across the industry to move to RISC to simplify the processor technology so that you can drive down the price of the product. And so what will happen over time is that RISC processors will compete with the 8/16-bit general-purpose microcontrollers. The other advantage of RISC is that you have huge bandwidth with 32-bit compared with what you had with 4 and 8-bit.”

Room for Growth

Galanos outlined the advantages of this: “You can create a general-purpose RISC chip and put all sorts of cool things in the silicon next to the core. You can add TCP, you can have ISDN, and software modems. So a lot of the code is coming back on the silicon. What we’re seeing is that the embedded manufacturer will be forced to become much more software intensive than it is today.”

Galanos noted that this represents a new way of thinking for most manufacturers. “It’s very counter-culture because most of the manufacturers of toasters and other such devices have no software experience at all, having never delivered software in the marketplace. We look at that as an opportunity with the same type of technical discontinuity that the Macintosh went through. Essentially, over the next five years, most of the 8/16-bit solutions are going to move to 32-bit solutions.”

Comparing the market for embedded processors to the personal computing market, Galanos has reason to appear optimistic. He noted: “Worldwide, manufacturers ship about 2.5 billion microprocessors and microcontrollers, of which, 1.2 billion are 4/8-bit processors, and I think around 150 million are 32-bit processors. Out of 2.5 billion, you only have a run-rate of 150 million 32-bit processors. So the upside potential is going to be in the 32-bit, not in the 8/16-bit. There’s just going to be more and more movement there.”

“Worldwide, manufacturers ship about 2.5 billion microprocessors and microcontrollers, of which, 1.2 billion are 4/8-bit processors, and I think around 150 million are 32-bit processors. So the upside potential is going to be in the 32-bit, not in the 8/16-bit. There’s just going to be more and more movement there.”

He continued: “I think we’ve already started seeing it. All of the manufacturers are now looking very strongly at 32-bit designs for the controllers and for the parts. Because, if you’re going to invest in these plants, you have to invest in a RISC plant and not in a 4-bit plant that you can only use in one sector of your market. So there is a tendency to use 32-bit across the board, including embedded.”

From Down Under

Now that we’re familiar with Metrowerks, who is Greg Galanos? He began: “I have a Masters in Computer Science, but I’ve been active in business and technology since I was seventeen. I’m 37 now, so that’s twenty years.”

Galanos recalled: “I dropped out of high school in Australia and, when I was 17, took a one-way ticket to Singapore and travelled around the world for two years, settling down in Montreal. I started working with computers on Control Data mainframes using punch cards, and decided to go back to school and finish a Computer Science course, an accelerated course.”

Returning to work, Galanos spent six years at the utility giant Hydro Quebec. “Then I was accepted into a Masters program at the Universite de Montreal, without doing my Bachelor’s degree. I was still doing my Masters when I launched Metrowerks as a startup. It took me a long time to finish my Masters because I didn’t have time to do my thesis, but in the end I did get all of it done.”

In 1985, while working at Hydro Quebec in Montreal, Galanos implemented the largest Macintosh network in Canada. But while completing his Masters, he took an active interest in compilers. He remembered: “I had met with Nicholas Wirth when I was in Zurich and started working on Modula 2.”

Now a resident of Texas, Galanos stayed in Montreal from the Olympic year of 1976 to 1994.

Adopting the PDA

Galanos built on his explanation of the embedded processor market to outline his thinking about the emerging PDA market. “The embedded market is a very old and very entrenched market, but it will change with the new model of software, high-volume and low-cost software. That’s what we’re bringing to that market now, the high-volume and low-cost. And that market has never seen that before. They’ve seen specific tools for specific projects, that’s what they did in 4-bit and 8-bit.”

Putting this into the context of PDAs, Galanos continued his analysis: “The best way to have high-volume, low-cost is to go into these new PDA markets where the type of programmer is not the same type that you find in the embedded market. He’s the person that used to work at Apple or at HP or at SUN in the Valley, and who’s now doing cool stuff on his personal machine.”

He continued: “And that’s why we’ve taken this sort of approach of doing this first work with General Magic to support their platform. They decided that they were not going to have their own tools and that they were going to work with us as a partner to support the platform.”

The General Magic Connection

Galanos recounted how Metrowerks came to be selected: “A lot of the programmers at General Magic knew of us because they had Apple contacts. They chose us because our technology also runs under MPW [Apple’s Macintosh Programmers Workshop], because of our 68000-based compilers, and when they contacted us, we looked at the business and the business model and it made sense to us.”

For Galanos and Metrowerks, the deal was an attractive one. “It was an incremental investment for us to support the platform with respect to the technology we already had. And as part of the incremental investment, we were essentially starting off on the ground floor on a mobile technology. And so, that was the part that made sense to us. We didn’t have to go out and compete against the Newton Toolkit.”

“We view General Magic [Magic Cap-based] devices as a very interesting technology, and we’re in this for the long run. We view this a little bit like Sony or Panasonic. We tend to look at the software we build from a hardware manufacturer’s perspective and not from a software perspective. These are new technologies that will take time to mature and they will go through many rounds before they mature properly.”

Galanos concluded: “And so we intend to be there as part of those revs [revisions]. Already I think there are modifications in the PDA market. So it’s an emerging market, it’s definitely going to take time to mature, and that really means getting in on the ground floor.”

Being the only game in town makes it easy to measure success. As Galanos put it: “Absolutely everyone on Magic is using our products.” But what about General Magic and their internal development efforts? Galanos noted: “They’re still using an MPW configuration because so much of it is script based. We’re moving very quickly to the new plug-in architectures and so, over time, they will move all their tools to CodeWarrior.”

“We’re looking at this as a good forward business for us, supporting a new platform. There’s a lot of synergy between Metrowerks and General Magic. Just from a development point of view, we’ve been able to ship product quickly and easily.”

Galanos summarized: “We’re moving to the next level with respect to development on the machine. The machine itself is going through a transformation as they see new uses for the operating system and decide that this or that is not feasible. So we view it as an extremely interesting first step into this market.”

Interactive TV = RISC

In addition to personal communicators, Metrowerks has targeted the yet unestablished market for television set-top boxes, or controllers. Galanos elaborated: “We’re taking one strong partner in each of the markets for the time being. General Magic is a strong publicly traded partner in PDAs, with a lot of strong licensees. In the PowerTV deal, we’re using parts that are PowerPC-based, so we use our PowerTV compilers. PowerTV is an 80% owned subsidiary of Scientific Atlanta, which in fact, is the second largest cable-box manufacturer in the United States. Again, this is an incremental investment for us, and a smart move for our shareholders in the event that the set-top box market does take off.”

Galanos painted a picture of the near future: “We’re looking at a software intensive business. People are going to use PowerPC-based set-top boxes to do stuff other than just change channels. I mean you can use the high-speed cable—I don’t know what the transmission rates are, but they’re very high, very fast transmission rates—and you’ll be able to do email, you’ll be able to do content and interactive databases.”

Elaborating, he continued: “There will be content providers that are going to use video servers to contact the audience which are already hooked in through cable. So the distribution channel is already there, now they have to bring the content in. I actually think that over time you’ll see the set-top boxes become the server for PDAs.”

Set-top Boxes and PDAs

“One of the biggest issues is that the cable industry is there already, you don’t need to go in and put a new infrastructure in place. The fiber is being put in and it’s going all the way down to the neighborhood switch. From there, it’ll come down coax cable to the house, so it’s all there. Therefore it’s going to be quite easy for the cable companies to move people from let’s say, a dumb terminal-type box like the one that people have today, to one that’s a little more intelligent.”

Galanos noted: “The communication technologies are going to be in the set-top box and they’re also going to be present in the PDA. There is no reason why, for instance, over time you would not use your PDA for your personal communications, and to control your set-top box with your PDA. So you’ll have email coming down through cable . . . this wireless is all very well, but it’s pretty expensive. It’s expensive technology whereas the cable is already there.”

“We’re taking one strong partner in each of the markets for the time being. General Magic is a strong publicly traded partner in PDAs, with a lot of strong licensees. In the PowerTV deal . . . PowerTV is an 80% owned subsidiary of Scientific Atlanta, which in fact, is the second largest cable-box manufacturer in the United States.”

“The down link for email, I think, can run at 128 Kbps, which is the same as two ISDN channels. The problem with email on a cable system is that the uplink is only running at 9600 baud. But usually when you’re using a PDA, there’s more information being sucked down to you than being sent up. You’re not going to upload some incredible visual image. You’re just going to send off some email message, so 9600 is going to work.”

Galanos predicted: “What’s going to happen is that people are going to start selling software that’ll run either on the set-top box, or on the video server, or on both.” In this version of the future, Galanos feels that people will become even more aware of their communication needs and capabilities, independent of whether the system uses wired or wireless technology.

He explained: “I think that they [cable and wireless] are complementary technologies. I don’t know which one will dominate. I just want to use the one that’s available. And so, it will most likely be more cost effective that I use my cable medium while I’m in my home, and I use my wireless while I’m on the road.”

Sounds good, but will people be inclined to adopt yet another set of communication options? Galanos was upbeat but realistic: “I think, in my perspective, there’s a certain percentage of the work force that’s ready for this. The knowledge workers are ready for this.”

More About Magic

Moving back to the subject of PDAs, Galanos analytically described his interest in General Magic’s technology. “What captured my interest in Magic Cap was the extreme usage of C. I believe one of the factors to make or break a technology is the amount of work that it takes a programmer to understand what is going on.”

“And so every time you introduce a new language, you essentially are introducing one more barrier to success for your technology. In the case of both the Psion and the Newton, you have to learn new languages to program the device. In the case of Magic, it’s straight C with their own framework.”

Reminding him that Magic Cap’s object-oriented framework is not entirely trivial, Galanos admitted: “That’s true. Well, all new technologies that advance the state of the art have a learning curve. But, it’s less of a curve than going out and learning a new language.”

Portability is also a concern for Galanos, especially in light of General Magic’s stated intention of moving the second generation of its hardware reference to a RISC environment. He noted: “The operating system is more portable, which should give us more portability from one hardware architecture to another. Whereas the Newton, I don’t think, is that portable of an architecture. We believe we’ll be in the best position to offer the tools for whatever architecture they choose.”

While Metrowerks is an active partner with General Magic in providing tools for third-party development, Metrowerks has chosen not to invest directly in General Magic. Galanos explained the reasoning behind this: “First of all, we’re not a licensee partner, rather we’re a tools and technology provider. So there’s the difference—it’s not really a means for us to invest in Magic and frankly it didn’t come up.”

In terms of support, Metrowerks provides the same high degree of support to Magic Cap developers as it does to its army of Macintosh programmers. “We do all sorts of technical support but when it turns into a Magic question, it’s routed into a special email to General Magic and the response comes back through them.”

Personal PDA Preferences

During the interview, Galanos occasionally picked up and manipulated his Psion handheld computer. This begged the obvious question…

“I don’t use a Magic Cap device because of the form factor. I have a Psion, but I have no communication capabilities. You know, I’m stuck between a rock and a hard place. I wanted something that can fit in my pocket, but I can’t do email with it. So that’s essentially the reason.”

Of course, Galanos has a Magic Cap device, but not one that he carries with him on a day to day basis. He confessed: “I do have a [Sony] MagicLink at home, but I prefer this one [the Psion] because I can put it in my pocket. So that’s the straight up answer.”

What about other PDA platforms including, most notably, the Newton? Thinking strictly in terms of business considerations, Galanos answered: “I think the Newton is a great device, however, the business opportunities for us are not as great as on Magic. We don’t have an ARM [Advanced RISC Machine] background. So we have the expertise, but we’re talking about a major investment for a small market.”

He explained: “Just because it’s RISC doesn’t mean we’re going to do it. There are other RISC back-ends, like SPARC that we don’t support. We don’t support Alpha either. And the bottom line is that ARM is a low volume chip with respect to the PowerPC family or the 68K family or the x86 family.”

“It’s not just the up front investment of building the tools, but the engineering investment after that. And so you have to be very careful in the way you place your bets. Jumping onto another PDA OS at this time would be counterproductive for our business model.”

Since he is an active Psion user, how does he feel about the business opportunities for that platform, especially in light of the fact that over 500,000 Series 3a handhelds have been sold with over 25,000 more being sold each month. Again, Galanos showed his practical side: “The Psion uses a processor that we don’t support. They’re not using any of the RISC chips that we’re supporting at this time. So the Psion essentially is a great device, but it’s using a proprietary technology with respect to the chips.”

Galanos related: “It takes a lot of work to support a new processor architecture. It’s not just the processor architecture, but you have to also do the library work, and you have to do the debugging features that are hooked in the operating system. There’s a lot of little things that you have to do that cost a lot of money. And so that’s essentially the only reason why we’re not doing it.”

Metrowerks: Virtually Worldwide

How does a relatively small company keep on top of so many opportunities, while maintaining top quality for its products? Galanos explained: “Metrowerks is a virtual company. I have 32 people in the engineering group, of which five are in Austin, another ten are in Montreal, and then a lot of my people are off-site. They are in Phoenix, Berkeley, Cupertino, Richmond, Boston, and Hamburg [Germany].”

To manage such a group, Galanos described an extensive system incorporating email, weekly conference calls, status reports, and a range of other methods. He noted: “It works really really well.”

Managing projects involves breaking people into small teams. Galanos related: “The coordinators conduct a weekly meeting with me via conference call, and they circulate reports around the virtual corp. We have one or two contact points, one for source code integration in Montreal, and another one for CD production in Austin.”

In terms of head count, Galanos estimated that perhaps up to 70% of the company is now in the US. “We started in Canada, but we’re branching out and most of the high-end architects are in the United States.”

Notwithstanding that the engineering and marketing emphasis is located in the States, Metrowerks maintains its public listing on the Vancouver and Montreal stock exchanges, with a registration for trading in Toronto currently underway.

The company began by listing in Vancouver in March, 1994, and added Montreal in February of this year. Firmly rooted in reality, Galanos explained that: “We went to Vancouver for the valuation. We had a better valuation in Vancouver that we could have gotten with venture capital.”

“I think the Newton is a great device, however, the business opportunities for us are not as great as on [General] Magic. We don’t have an ARM [Advanced RISC Machine] background.”

In terms of listing on a US exchange, such as the NASDAQ over-the-counter market, Galanos replied positively: “It’s absolutely a goal, that’s where we’re headed. Our shareholders expect us to be in the United States.”

When it comes to hiring talent, Metrowerks has the advantage of seemingly being everywhere at once. Galanos described the process: “I leave most of the hiring up to my architects. People who contact me have to get through the engineering group’s review. If they cut the mustard as senior engineers, then they come on board.”

He noted: “We tend to hire highly experienced people, people who have a track record. Most of the juniors that come into the company come into tech support and then they move into engineering later on.”

Managing such a far flung team also enables Metrowerks to concentrate on the important issues related to engineering and quality. Galanos offered: “We don’t track when the engineers come into the office, we track what they deliver. So the focus is 100% engineering, and it works well.” Being spread apart also offers another advantage. “The fact that we’re a virtual corporation limits the amount of politics that people get into.”

“When we shipped the product in January 1994, I think we were ten people, whereas today we’re sixty. And the engineering group has gone from probably five or six at that time to 32. We have a certain number of the engineers working on specific projects such as Magic or PowerPC, and the others working on the Mac OS. Will it change over time? It’ll probably just get bigger.

So how do you manage quality over these long distances? Galanos responded: “I use email and I carry a big stick.”

Greg Galanos
President and CEO
Metrowerks
3925 West Braker Lane, Suite 310
Austin, TX 78759-5321
(512) 305-0400
(512) 305-0440 (fax)