Through the Interface: An interview with John Walker - Part 1

May 2015

Sun Mon Tue Wed Thu Fri Sat
          1 2
3 4 5 6 7 8 9
10 11 12 13 14 15 16
17 18 19 20 21 22 23
24 25 26 27 28 29 30


« Announcing two new Autodesk developer blogs | Main | More quiet command-calling: adding an inspection dimension inside AutoCAD using .NET »

September 12, 2008

An interview with John Walker - Part 1

On Thursday of last week (September 4th, 2008, for those people accessing this post sometime in the distant future :-), I had the great honour of meeting with John Walker, one of Autodesk's founders and its former CEO.

I've known for some time that John lives in the Neuchâtel area, and it turns out he lives just 10 minutes' drive from my home. I decided quite recently to get in touch with John and suggest I interview him for this blog, and I was delighted when he accepted. Not only was John very generous with his time - we talked for four hours - he also shared many fascinating historical tidbits and forward-looking insights into the CAD and software industries.

For those of you who are not aware, John authored/compiled a fascinating book on Autodesk's early history, The Autodesk File, which is just one of the interesting resources available on John's Fourmilab web-site. I strongly recommend anyone interested in Autodesk's business and origins read this book: it provided me, at least, with many valuable insights into the history and culture of the company at which I've chosen to spend so many years of my working life.

I've just spent three days transcribing the recording I made of the interview. To improve the flow I've edited the order of some of the questions and answers - to say I'm an inexperienced interviewer would be a gross understatement - but I've done my best to leave John's words untouched. I must stress that John's words are his own, and do not necessarily reflect the opinion of current Autodesk management and employees. Whether that's a good or a bad thing I'll leave for you to decide. :-)

At the end of the transcription effort I found myself with 18,000 words of text to publish. Needless to say I haven't put that in a single post, so I'm serializing over a number of weeks: it'll be a regular Friday feature until it's all published, which will also give me the chance to continue posting technical content over the coming weeks. Here are the planned episodes - I'll make sure these become links as the posts come out:

One last word on the conventions I've used in the below text: my own words - which you'll be thankful I've kept to a minimum - are in bold, John's are in normal text. I've enclosed editorial comments [as italic text in square brackets].

And now for the interview itself... enjoy! :-)

John Walker holding an M9900 and the prototype hardcover manual for AutoCAD 2.1

Arrival at Fourmilab

[John welcomes me by showing some AutoCAD artifacts...]

AutoCAD 1.3 manual from 1983

At that time there was no customization, except that you could make your own menu. And, surprisingly enough - as you could put multiple commands on one menu line - some people were able to make simple applications. That's what really gave us the... well I think we already knew from the outset that we wanted to make it programmable and extensible, but it was this tremendous cleverness of using something we didn't even think of as an application medium that put the priority on getting there.

AutoCAD 2.1 prototype hardcover manual [which John is holding in the above photo, along with the Marinchip 9900, AutoCAD's first hardware platform]

This is a prototype of the first hardcover manual. For AutoCAD 2.1. If you look at page 200, this is the Variables and Expressions feature, the pre-cursor to AutoLISP. In fact, all of LISP was there, it was just turned off. Primarily to save memory, but also because we didn't have the mechanism to reach in and retrieve entities or create entities and do anything like that. It just didn't make sense to put in a programming language that didn't have access to the drawing database. And so this was kind of half-baked. The people by this time that were doing advanced menu macros were just screaming for ways to turn modes on and off at that point without pumping in commands, so this was just a stop-gap.

I think it was the 2.13 [*] release - which was a disastrous release - but I believe that's where full AutoLISP came in. That was really the worst release of AutoCAD ever made - at least during my tenure - from a quality stand-point. There was a horrific bug - in x86 assembly language - and most of this stuff dates from the age of 16-bit machines... when you transfer from one program to another you have to load the segment register. It's two instructions and you have to do it in a particular order and in the AutoLISP code I got it wrong. It's a timing thing - one time in a million you'd get an interrupt between those two instructions and everything would go black. Of course on a computer, one in a million happens about every hour. So we finally figured that out, but that was really tough. We also had, as I recall, some driver problems, and we changed some stuff in the drivers, so that one - I think it was 2.15 we finally called it when we got it right.

[* A follow-up comment from John: It's a little more complicated than this. Variables and Expressions was included with 2.1 in May of 1985. This was the release in which the hardware lock was introduced on international versions of AutoCAD, but the domestic version remained unlocked. AutoLISP was introduced in version 2.18 in January 1986, but did not provide access to the entity database until version 2.5 in June 1986. It was in version 2.5 that the hardware lock was introduced in domestic versions.]

AutoShade manual

This was perhaps our second [**] product. It went through a long, bouncing-around history. We were going to acquire a rendering engine from another company and build an interface around it. That went around and around and around. Their rendering engine was much better than AutoShade in terms of what it could do, but it was just never getting done. So I went off, with one other guy, and built this thing - very, very simple. And actually, I don't know, some of the... at the point we put the Z-buffer shade command in AutoCAD, there really wasn't any need for this anymore. And at that point we were working with the Yost group doing 3D Studio. That was a real ray-tracer engine, and there really wasn't any need for AutoShade anymore.

[** A follow-up comment from John: Well, actually not. In fact, AE/CADD, CAD/Camera, and AutoSketch all preceded AutoShade.]

It was actually really embarrassing - we had AutoShade well before we had anything other than the pure extrusion 3D, and so we could make all these beautiful pictures, but you couldn't model them in AutoCAD! And so we jammed in this 3DFACE command, to try to plug that hole.

[Then starts the interview proper...]

Autodesk's early history

How did Autodesk became a CAD company?

The short answer is: COMDEX 1982. Up until that point we were really focused on the product Autodesk [an office automation system for small computers which only later became the name of the company] that was going to be the killer product. AutoCAD was kind-of this... it was interesting because we were pretty sure we could get it to market quickly. It and AutoScreen (a screen-based text editor), we were pretty sure we could get those done for COMDEX. It was a real push, but it could be done. Autodesk was a lot more of an ambitious thing. We could show a prototype - in fact we had a prototype - but getting the whole thing done would take a lot of work.

Secondly, and this came back to bite us a few years later, AutoCAD was licensed from Mike Riddle, who was also a founder of the company. Unlike everyone else who joined the company, he did not contribute his software unencumbered: he had a royalty deal associated with it. And that royalty deal was - given the price we envisioned for AutoCAD originally - so onerous that it was a very low profit product for us. Now, when you start out with $59,030, I think it was, and you're spending a thousand or two a month, anything that got money to come in, even at a low profit margin, was attractive. Because that would mean - since everyone was moonlighting - people could quit their jobs and start working full-time. Which would drastically increase the amount of work we could get done.

So it's not like we weren't interested in it. But we had the same opinion as the rest of the industry did. That it's a niche product, it's got nowhere near the market - I mean just compare the number of architects with the number of people that write documents - that's pretty obvious. It's inherently an expensive product - even if we gave it away it would be expensive, because at that epoch graphics boards with a reasonable resolution and digitizers were very high-end products, you just couldn't buy them cheaply. Most computers didn't have mice - there wasn't any graphic pointing device at all, so it was a battle even to be able to get a pointer on the screen. So that meant the hardware configuration was going to be expensive, and that again was going to reduce the size of your market.

Well, we got this AutoCAD 1.0 done, and in fact, I believe we actually sold a copy of the CP/M version - because we had it on the CP/M, the IBM PC and the Victor 9000 before COMDEX. But I think it was to one of our dealers: I'm not sure you can really call that a sale. But we took it to COMDEX, which at that point was on this exponential growth explosion. 1982 was the last year it all fitted into the Las Vegas convention center. After that they started expanding out to other venues and so forth. We were in the convention center, but were about as far from the action as you could get. Way up against the wall, near - as Barnum used to say - the Egress. Well if you were heading for the egress you would have to pass by our booth. A 10 by 10 foot booth, and I think we had four machines: a CP/M machine - and CP/M at that point had the nicest graphics displays available, so even though today you'd think of it as an archaic 8-bit machine, in fact they'd got the megahertz right up high enough, it was really just as fast as the IBM PC, and most people don't know that it ran in single precision, so the floating point was a lot faster. Really the two were quite comparable, but it had this gorgeous 512 by 512 SCION MicroAngelo display, and not only was that twice the resolution of the IBM PC in CGA mode, it had its own computer onboard and did the vectorization itself. So it could draw lines and circles and things - you just gave it the command, and so that meant you didn't have to set every pixel on the screen, which made it faster still. So that was the machine we did our demos on, basically. But we had it there on the IBM PC and the Victor 9000 and then we had another machine that was running our AutoScreen text editor.

From the day this show opened until the day it closed - which I think was Wednesday through Friday - the booth was absolutely jammed, you couldn't get in there. There were lines of people waiting to see it. Historically, at COMDEX '82 the IBM PC had come out earlier that year, and every computer manufacturer - or would be computer manufacturer - was furiously re-engineering their machines, because they were mostly CP/M machines before that, to make their response to the IBM PC. Better graphics, larger disk, more memory, apples flying out of the top, whatever. COMDEX '82 was when they all launched their products. They were all looking for a killer app. Something that could not only sell their machine but could distinguish their machine - and all the great stuff they've done to it - from the IBM PC. So we had Compaq, Texas Instruments, NEC and Wang, and a dozen others you've probably never heard of, coming by our booth saying "what would it take to get this application on our machine? This is the thing we want to demo, because we have better graphics, we can run faster, etc." And we, of course, being impecunious at the time, said "well, send us a machine and we'll make it run on your machine".

So not long after COMDEX all these machines started arriving and we realized "gee, this is going to be a lot of work". Especially when everybody that saw AutoCAD at COMDEX asked "can you add these twelve features?". Obviously the same people were doing these drivers and the features, so that was a bit of a trade-off.

Driving back - and this is how we got there, in a big station-wagon we rented in San Francisco - there were three of us that made the drive... we'd said we'll do five products and count on one being a hit. We now knew which one was the hit - there just wasn't any question. We maybe sold four or five copies of AutoScreen after that. When we got back, the phone was ringing off the hook with prospective dealers saying "how do I sign up to be a dealer?" or manufacturers saying "OK, how do we get you the machine? When can we have this ready?". From that point on there really wasn't any question that AutoCAD was going to be the product.

Well from the business standpoint - and I have to say I don't think we thought about the business standpoint for the next few months, because we were just trying to keep our heads above the water - we still had this royalty problem. It was then that it occurred to us that while the royalty affected the base product, it did not cover additional features that were sold at an extra price. That gave us a tremendous incentive to add features, and that was the birth of what was called the ADE [AutoCAD Dimensioning Extension, which was later renamed Advanced Drafting Extension] packages. Because the ADE packages were not subject to the royalty. One of the first things I did when I got back was to write this really idiot-stupid dimensioning package. It was done in a weekend - that's how simple it was. We sold it for $500 - and that was a $500 revenue that came straight to us and wasn't subject to royalties.

Suddenly the product was actually - with dimensioning, I don't know the numbers off the top of my head - it became maybe twice as profitable as it was before and that's why you had this progression of ADE-1, ADE-2, ADE-3 [ADE-1 was the original dimensioning feature]. We hardly ever sold anything but the highest level of the product, because the dealers obviously would demo it at that level, and the customer wasn't going to buy something that didn't have all the neat features that the dealer had shown them.

So that's really how it came to be AutoCAD, and from the weekend of COMDEX there really wasn't any doubt. We did continue to work on Autodesk for a while, but it was just clear now we had a product that people wanted, we had machines coming in that people could work on, we had dealers wanting to sign up. We even had, not long after that, writers contacting us that wanted to do review articles for magazines, and Autodesk was - at best - maybe 6 months to a year of being ready. We just didn't have time to work on it.

In any case, it was clear: we'd priced this thing at $1,000, and people were willing to send us $1,000. Now I don't think we could have got that for Autodesk. So that was another consideration. Autodesk would probably have been a product in the $100-300 niche that word-processors were going for, at the time. And now it would have been very cool to have what was essentially a Macintosh desktop running in text mode on the IBM PC. I still think that would have been a killer product if we'd gotten it done sometime in 1983, but after that point it was a question of, there was one guy who was working on it and he was really needed to work on AutoCAD.

So from that point on we really became the AutoCAD company. Now you asked, how did we become a CAD company? Well that's another interesting story. Because we never thought of ourselves as a CAD company, even then. We were a software company that had this CAD product. And if we ever had time to do anything else we might do other, completely unrelated software products.

Along comes 1985 and it's time to do our Initial Public Offering. Well, there had been a whole raft of PC company - not so much software company, but PC company - stock offerings in '82 & '83. When everybody was jumping in against the IBM PC and the products were being launched. And most of those stocks performed disastrously. It wasn't even so much the companies: there was a heck of a recession going on, at the time, and the stocks were way down. I used to say, at the epoch, if you wanted to clear a room of investors you only needed to say two things: one is "PC" and the other is "retail sales". Because the dealers were collapsing all over the place, too.

So there was no way - we probably could have made the stock offering, but at half the price - if we were perceived as a PC company or a PC software company. We positioned ourselves in the prospectus as a CAD company - "this is the low end of the CAD market" - because CAD was glamorous at the time, workstation CAD: Computervision, Intergraph, CADAM... and that was the hot thing, and that was what investors wanted to see. We were "the first CAD company with a mass market" - that was really the pitch. Interestingly enough, we did our secondary stock offering in 1987, and if you read that prospectus, we're a PC software company. Because Microsoft had gone public, and Microsoft had taken off in a big way, and PC software was hot. And so we didn't want to be seen as this niche CAD company. And that's what they call marketing. :-)

Autodesk - the company - was one of the most successful high-tech startups of all time, if you consider its lack of debt and external (Venture Capital) investment. Do you think that kind of success is possible to do, these days, without VC?

No! :-)

How did you manage it?

In one word, luck. We were at the right place, at the right time, at a unique juncture in the software business. When I say "no", I don't mean that it's not possible for a couple of guys with a great idea to make billions of dollars - look at Google - but not in the industry we were in.

The difference is, we arrived just at the time that the IBM PC had turned the whole PC business upside down. Suddenly you had this completely new machine, which could not run any of the existing software written for the Z-80 and CP/M, you had 15-20 manufacturers jumping in the market with both feet, Japanese companies bringing in boatloads of these PCs with absolutely no software to run on it, other than CP/M or MS-DOS. And quickly-ported and ghastly versions of WordStar and so forth, which were mostly hack-ported from CP/M and kinda worked, maybe. And there was this huge vacuum awaiting software. If you read the proposals from '82, that was the whole concept of the company: we are at a juncture here. There are millions of these machines coming in that need software. They are not going to sell unless they have software on them. And we - fifteen mainframe programmers - know how to build big software packages. And further, we're familiar with the applications that run on mainframes, and these machines in a couple of years are going to have the power of mainframes and will need mainframe-style software.

It was that game-changing moment. And in that kind of a moment, you don't have to have the production values that are assumed for an entry-level product today. Look at the investment that went, say, into Google Chrome, and they didn't even write all of it - they didn't write the rendering engine in it. And it's getting panned by a lot of people who are saying it's not ready yet.

Back then, at the point we launched AutoCAD at COMDEX, that was the work of three people over about six or seven months. The manual was written by one guy - the original manual, which I don't have - didn't have any illustrations in it. It's the manual of a CAD program, and it didn't have any pictures in it, because we didn't have any way to make them then. AutoCAD didn't work well enough, hardly, to get the pictures in the manual! It was printed out on a daisywheel printer; that's how the master was made. And so you didn't need Venture Capital when you had people - who, at that point, were pretty senior people in their regular jobs - moonlighting and able to throw something like that together. Well, today the investment that goes into the icons on the screen is greater than what went into AutoCAD. Everything is this monster thing.

And also you're on this outrageous and despicable treadmill of Microsoft changing the rules every 6 months, and just the investments it takes to keep up with their disastrous Operating System, if you can call it that, is something that requires probably a company staff larger than Autodesk had in '85 after we'd gone public.

So the game has just become so unbelievably complex and the standards so high.

Also, it was possible, back then, to make noise and be heard without a huge production and advertising budget. All we had to do was take it to a trade show: that cost us about $3,000 - which is a lot when you've got $60,000 in the bank - but it wasn't like the $50M roll-out they do with some of these things today.

And that provided an opportunity. I think you need only look at all the niche businesses that have started up by people putting their hobby on eBay and now they have fifty people on staff and are selling around the world. In Glenn Reynolds' book, an Army of Davids, which is entirely about the fact that, for example, what does it take to make Hollywood-quality film? Well, a camcorder and a copy of Adobe Premier! So independent film-making is really independent, because it's a question of "do you have the talent?" as opposed to "do you have the budget?". That's the kind of game-changing thing that happened to us, at that point for the PC software industry.

But I've said many times, what was responsible for our success? Luck. And what particular piece of luck? The IBM PC AT coming out when it did. Prior to the AT, it was just so painfully slow. And all of the stuff that we added would have just been laughable. I remember when we first got hatching working, and I did a demo of a complex shape for one of the venture capitalists - one of the probably hundred venture capitalists that visited us and decided not to invest. It took 45 minutes - this was not really impressive. "Er, gee - couldn't you actually draw it with a pencil faster than that?", people would remark!

But if the AT had come out, say, a year before we'd started, all of our competitors would have been on it, before we had anything to show, at all. And they were better, at that time. If it had come out later, we would have been out of business because we were trying to sell something that wasn't terribly practical. I mean it wasn't even a kind of "bleeding edge" early adopter, it was like a "bleeding stump" early adopter that you had to be to be an early adopter of AutoCAD on the original PC, at least without a nice graphics card and a bunch of stuff that ran the price way up. If you had the add-ons it was fine, but if you had a stock IBM PC, it was just bad.

If the AT had been late, we would have run out of money, probably. I don't know - maybe, maybe not. And that was something we didn't even anticipate, when we started. It just happened at the right time, for us. The AT, and then the explosion of the Compaq 386 was the thing: that changed the game again. I remember when we got one in from Compaq, just to verify that everything worked on it - it should, and it did, in fact - but it was, like, six times faster than anything I'd ever seen before.

That may have been what made applications feasible. Because of course, when you look at the origins of AutoLISP, we fell into exactly the same trap that every scripting language developer falls into. We said "well, it doesn't have to be fast, because it's pumping commands in! Executing the program is nothing compared with the time it's going to take for AutoCAD to do these commands." Of course not realizing that people are going to do things hundreds of times more complicated than you ever anticipated. And then it actually does become a bottleneck.

But again, suddenly, the fact that the machines were speeding up meant that we didn't have to address that problem as early as we probably would have had to, otherwise. And that was a dog-gone good thing, because until we got out of the MS-DOS 640K restriction, there wouldn't have been any way - it just wouldn't have fitted in the can.

Autodesk was very much a multi-platform software company at the time.

Until after I left. It was a multi-platform company when we showed AutoCAD - we had it running on three different machines with two different microprocessors: CP/M 80, Victor 9000, and the IBM PC. And, I might mention, in two different programming languages: the CP/M version was in PL/I and the x86 was in C. As a result of our glibly saying "oh, send us a machine and we'll get it to run", I think we were on something like sixteen or seventeen different machines by the end of 1983. In fact, there was a picture - at 150 Shoreline we'd rented this big, open space on the second floor, and that was our porting and QA lab. We actually once collected all the machines together, in a picture known as the Great Train Wreck - I don't know if there's a copy still in existence, but we put it out in one of the newsletters, and so forth, to show all the platforms it ran on. And by that time it was over twenty. Now, by that time they were all x86 microprocessor machines that had the same C code and different drivers.

The Sun was the first UNIX - and that was the 68000 architecture, at the time - and that was probably, I don't know, we probably started working on it in '86 and it wasn't done until '87 or '88, I'd have to go back and look at that. But that was our first leap. We actually had tried to do a UNIX port well before that, on a machine, a 3B2 from AT&T, or something like that, which was a little odd, as they didn't have graphics, at the time. But they were going to have graphics later, and we figured at least we can get over the hump of making the code run on UNIX, by then. And, as sometimes happens, it was given to this guy who never got anything finished and it just sat there for a long time. But it was really the Sun that made the difference.

Not long after Sun we had Apollo, which was another 68000-based machine, then I believe next we had MicroVAX. We never really sold very much on those platforms, but they had a tremendous influence on AutoCAD, in two ways: one, putting your code on a second machine finds hundreds of problems in the code, particularly when it's a machine with a different byte-order, different microprocessor, and all of those kind of things. Secondly, it gave us a preview of the future where we had unlimited - or what to us was unlimited - memory. At that point we may have been spending 50% of our time fighting with 640K problems. That was really just consuming a vast amount of our resources, and even though we knew we couldn't go away from the PC platform, we could see this was going to be fixed: people were talking about DOS Extenders, Microsoft was talking about a true 386 mode, OS/2 was coming on the horizon. The Microsoft-IBM alliance on OS/2 - that was going to solve our problems. And working on the UNIX machines showed us what that was going to be like. So we understood, both that things were going to get a lot easier from a technical standpoint, and the bar was going to be set a lot higher in terms of graphical user-interfaces, user expectations, and just the feature-level that people want on these platforms.

We did Silicon Graphics - that one came pretty late, in '91, I think - and, of course, Macintosh, which I did originally. We had this big project - there was always this Macintosh project of some kind - and they were trying to port it to the native Macintosh development tools. And, oh my golly, if you think 640K is bad, talk about 32K segments on the original Macintosh architecture! AutoCAD was just simply not written for that bizarre way that they set things up on the Macintosh. You basically have to re-write the whole program to do that. And so these Mac people were like "well, we'll just re-write the whole program! No problem!". "Erm, well, guys, you're talking about, we're up to maybe 10 man-years in this thing, now, and that's erm a bit of a job."

And so I suddenly realized, "wait a minute - I've got this Sun machine, here. It's got the same microprocessor as the Macintosh II. I can compile code on the Sun that will run in native 68020 mode, and it'll run on the Macintosh", I demonstrated that. So I had this thing called "the Titanic and the tugboat", where the Titanic was AutoCAD compiled on the Sun without any user-interface and the tugboat was this little thing on the Macintosh that had all the user-interface written in Mac code and just called the Titanic every time a command came in. And it worked! And we shipped it! It ran, and it was successful enough that we hired a bunch of Mac guys to clean it up. So what did they do? Well, this was awful. They started re-writing AutoCAD to fit it into the Macintosh development tools. So fortunately that didn't last too long.

Then it was only after Carol Bartz arrived that the winnowing of platforms started. I think we may even still have had the Macintosh version, then. Maybe, maybe not. But we definitely had Sun, HP Apollo, and SGI, still. MicroVAX had basically petered out, as DEC had abandoned the MicroVAX. Somebody was working on a NeXT - I don't think that ever got done. But the radical focusing on Windows/Windows NT was really something of the Bartz era. And let me say - that was something I was 100% vehemently for, at the epoch. If you read the remarks that I made at the introduction of the first AutoCAD for Windows, I was onboard with that.

You saw that as the future at the time.

I saw that as the future - one thing I didn't see was just how wretched it was going to be. I assumed that with the resources that Microsoft had, they could at least do as good a job as Sun and Apollo and Silicon Graphics. Which, admittedly, wasn't perfect, but it wasn't this junk that we've endured now for two decades. I guess the other thing is, we were already on OS/2 at that point. We'd done the OS/2 port, and OS/2 was done right. That was done by people that knew what the heck they were doing. So we assumed "OK, for business reasons there's been a falling out between IBM and Microsoft, but Microsoft knows how to do it, because they've been working on OS/2 for the last three to four years." So how can they mess it up? It's going to be better than OS/2, and that's what Microsoft were saying, "it'll be better than OS/2". And besides, even if they mess it up, there'll still be OS/2. People are not going to use an inferior product. Well, I was wrong! People did use an inferior product. And they still do. And I don't think anybody anticipated that, at the time. Because you could say "yes, sure, it's junk, but look where it was two years ago", if you compared Windows 1.0 to Windows 3.1, well it had come a long way. And obviously they had the processor power, they had the memory coming, the microprocessors that can handle large address spaces, and OS/2 had already done all those things. So it's just astonishing, and one of the most depressing things we've seen in the industry, that we are where we are today.

Now, did Autodesk make the right decision, to focus on the NT platform? Absolutely! Where would your market-share be, if you'd gone to the Macintosh or Sun, or something like that? Did the workstation vendors blow it, by not concentrating on marketing AutoCAD and taking out Windows, when they could have? Yep, they sure did!

Scott McNealy of Sun [Microsystems], always used to talk about the insurmountable opportunity, and the opportunity is only insurmountable if you don't want to try to do it. I sat in the Sun conference room and pitched to him and John Doerr. This was when they had the Sun 386, which didn't cost that much more to make than the PC. It was done by a rogue group in Boston, basically. This thing was an out-of-the-box solution that just runs AutoCAD fantastically better than any PC platform at the time. And that's the PC with the add-on graphics, with the add-on monitor, with the add-on memory extender, with the add-on DOS extender package loaded in, and all this junk you had to throw together just to get workstation - minimal workstation - capabilities. You unpacked the Sun workstation, you put in the AutoCAD install disk, you're in business - you're running AutoCAD! I said, "what you ought to do, you ought to get this thing in every one of our retail dealers. Because the dealers understand - this thing is cheaper than what they're selling now. It's not cheaper than the bare PC, but it's cheaper than the PCs they're selling with all the junk that they have to add-on to get this kind of capability." And McNealy's response was "well, we couldn't handle the support calls. We couldn't handle inexperienced users. We couldn't handle people who don't understand how to do UNIX system administration." To which I said, "Scott, get yourself an AutoCAD PC and look at how difficult it is to get that set up, and get everything working and configured and so forth. And secondly, I think the users are a lot smarter than you may think." But, of course, they had this dream of a next generation that was going to be really easy to use and then they could roll that out, but it wasn't ready yet. Well, you have to do it when you have to do it, sometimes. And if you have to hire a hundred people to read from support scripts, because there's nothing else you can do, well I guess we found out that works. It's not very nice, but it can be done. And I think they threw away a huge opportunity, there.

The other influence, though, to get back to your original question, the influence that Sun had on Autodesk, was that we moved our development from the PC platform to Sun. All the people that were doing the feature development of AutoCAD started to work on Suns. The quality of the tools was just so much higher that their productivity probably tripled. Now there was additional work, obviously - it's a risk: you develop on one machine, you target another, you're going to have problems - but the productivity gain from having developers working three times faster far more than compensated for the QA problems that we had, eventually making the PC version. Besides, it was just around that time that the native 640K version basically went away, and we now had essentially unlimited memory with the DOS extender package on the 386 machines.

All of these industry analysts - when I say the word "analyst", I'm thinking the word "idiot" - were saying that the 286 machines would be around for at least 10 years and that Microsoft and IBM are right in continuing to keep Windows on the 286 architecture. Well, all I can tell you is that in the AutoCAD market, they evaporated like the morning dew on Mercury. About a year after the 386 came out there wasn't anybody running AutoCAD on the 286, except somebody that had a legacy machine, but there certainly wasn't any new user - and there wasn't anybody interested in an upgrade of AutoCAD - that ran on the 286.

So that really let us get onto a platform that eliminated all the horrific memory problems. And then that made everything else possible, like the C-language ADS application system, and so forth.

[Tune in next week for Part 2 - AutoCAD's architecture & APIs.]


TrackBack URL for this entry:

Listed below are links to weblogs that reference An interview with John Walker - Part 1:

» An interview with John Walker from Between the Lines
My coworker a top notch developer, and also a fellow blogger (Through the Interface) Kean Walmsley just figured out that the founder of Autodesk lives not too far from the Autodesk office in Switzerland. Kean lives in Switzerland. So he had a visit and... [Read More]

blog comments powered by Disqus


10 Random Posts