Oral History of Charles (Chuck) Thacker

History of Charles (Chuck) Thacker"/>

Oral History of Charles (Chuck) Thacker

【出处】

.05.01.acc.pdf

 Interviewed by: Al Kossow

Recorded: August 29, 2007

Mountain View, California

 

Al Kossow: It’s August 29th,2007. This is tape #1. We’re talking to Chuck Thacker today. Thanks for comingby.

Charles (Chuck) Thacker: My pleasure.

 

Kossow: And I had prepared a list of things I’d like to talk to you about kind ofstarting with kind of the bio things, just where you’re from, when you wereborn, all the usual stuff like that.

Thacker: Sure, well I was born in Pasadena, California in February of 1943 and Igrew up mostly in the Los Angeles area until I wound up going to Berkeley whenI moved north, and I’ve never moved south again.

 

 

Kossow: You have any siblings?

Thacker: I have one brother.

 

Kossow: Okay, is he still down there?

Thacker: No, he lives in Reno. He’s not really a technologist at all.

 

Kossow: Oh, okay. I read the talk that you had with Mihai Budiu in October of2006. So there are a couple of things that came up from that, that you wantedto be a physicist early on. So, was your father in technology?

Thacker: My father was an engineer and I wanted to be a physicist not because ofthe physics but because of the engineering part of it. I wanted to be anexperimentalist and in those days that meant particle accelerators.

 

Kossow: Where’d your dad work?

Thacker: He and my mother were divorced when I was quite young so I never reallyknew him very well, but he worked on power stations and things like that.

 

Kossow: He was an electrical engineer.

Thacker: Yeah, but I was attracted by, you know, the large number of technologicalproblems in a particle accelerator, spanning the gamut from, you know, logicdesign of the data acquisition stuff to designing enormous power supplies forthings like the magnets and the klystrons [ph?] that run them, and all that sortof thing. And so it seemed like a good way to really do some very interestingand cutting edge engineering.

 

Kossow: Where did you go to high school?

Thacker: I went to high school in Highland Park, California, a high school calledFranklin, which is now not really a very academically oriented place. But, youknow, the demographics of Los Angeles has

changed a lot since I lived there.

 

Kossow: And then from there you went to Cal-Tech?

Thacker: I went to Cal-Tech for a while. Then I went to UCLA for a while and thenfinally I moved north and went to Berkeley and I finally graduated fromBerkeley.

 

Kossow: So, when did you go to-- ?

Thacker: 1963.

 

Kossow: Okay, and then your Bachelor’s was in?

Thacker: Physics.

 

Kossow: In?

Thacker: 1967.

 

Kossow: Okay, and then you started onGenie in’68?

Thacker: Yeah, I had planned to go to graduate school but I was actually married atthat time. My wife and I were married in ’64.

 

Kossow: Okay, you met her up here?

Thacker: Yeah, I met her in Los Angeles, actually, and I had planned to go tograduate school but since I was fully self supporting, I had to work for awhile before doing it. So, I took a job, actually, working forthe only inventor in the Berkeley phonebook, a guy named Jack Hawley whom you may know peripherally. Yeah, we’veactually put Jack in the mouse business.

 

Kossow: So, what was he doing back then?

Thacker: Well, he had run and I met him at a company that I had worked for earliercalled Berkeley Instruments, which made digital automated weather stations（自动数字气象站）. And they were very interesting to people who wanted to put weather stations ontop of mountains and things like that, because they were very low power. Theyused transistor circuitry but with a lot of fairly clever mechanical stuff inthe sensors. Jack was a very interesting guy because heknew a lot about mechanics but not much about electronics. So, when I graduatedand was looking for a place to work, he had set up as a sole proprietor of an inventorship in Berkeley. Actually, itwas in Albany and I went to work for him, I told him, for a fixed period oftime and I made him a deal that he couldn’t refuse. I said, “ I’ll design yourelectronics for you, if you teach me how to run all yourmachine tools.” So, he did and he had, you know, he had a good lathe, and agood milling machine, and a good drill, and all the surrounding tools that youneed to actually build mechanical stuff. And so I learned howto use all that, became a fairly good machinist.

 

Kossow: Have you done much of that?

Thacker: I have not and I would be absolutely doomed today because it’s alldigitally controlled now, and of course at that time you used micrometers andcarefully turned the dials on the machines in order to fabricate things. But in any case, oneday a friend of his dropped by the shop named Paul Persons, I believe, who wasa rep for an electronics distributor. And he had been calling around in theBerkeley area and he called at the Genie project, and he mentioned that theywere looking for an engineer and he thought I might like to interview there.And I said, “Well, you know, I’m pretty happy with this.” Jack said, “Well, youknow, the project is coming to an end,” and so on and so forth. So, I went andtalked to the Genie folks and I guess they liked what they saw because theyhired me, and that was in some ways the best decision and in some ways theworst decision in my life, because I never finished graduate school. I got sointerested. I had worked with computers before, but—

 

Kossow: What did you work on?

Thacker: Only peripherally. I mean, I started out by trying to program anLGP30 but discovering that, you know, the challenges of actually spacing outthe instructions on the right place on the drum were just daunting. And so I didn’t really enjoyit, so I stopped doing that, and then as a physics student, I actually had towrite programs because FORTRAN programs are needed for analyzing experimentalresults. And so I learned how to program in FORTRANand programmed a 1620, in fact.

 

Kossow: Where did you pick up your electronics background?

Thacker: At Berkeley. I took both physics courses and engineering courses, but alot of it was just building things.

 

Kossow: Okay, so you weren’t ever a Ham or anything?

Thacker: I was a Ham. I was a Ham radio operator. I was in a science club. I was inall the things one did in high school if one was a young nerd, and so where wasI? Right, so Genie held together for another two years, roughly, and theyhad built one of the machines that actually captivated me, which was a 940, andwrote a timeshare operating system for it.

 

Kossow: So, I’m really curious about that because that’s where everybody startscoming together. You meet Butler [Lampson].

Thacker: Right, I met the whole bunch.

 

Kossow: And Peter Deutsch.

Thacker: Right.

 

Kossow: So, what was it like? What was the environment like? WasKen Thompsonthere then?

Thacker: No, he was not, or he may have been somewhere but he wasn’t associatedwith our project.

 

Kossow: Because he kind of drifts through it.

Thacker: Yeah, he drifts in and out, but we were on the fourth floor of Corey Hall,which was at that time the top floor, and in a little thing on the northeastside, which was the suite of offices and a lab for the project. And the 940 itself was in amachine room down on the first floor.

 

Kossow: Was the 6400 in by that time?

Thacker: No, the 6400 was in the computer center, which was down the hill, more onthe central campus. Because at that time, there were two computer sciencedepartments. There was a computer science department that was part of the collegeof letters and sciences, and there was the computer science department that waspart of the EE department. And now they’ve merged, they no longer have that distinction,but the one down the hill was actually trying to build a timesharing system forthe 6400, which never apparently worked out very well.

 

Kossow: That was Paul McJones and Butler.

Thacker: Paul McJones and Butler was there, and Charles Simonyi, and so the 940 wasactually a marvelous machine and I enjoyed it a lot because for the very firsttime you could sit down and program interactively, which was a new thing.You know, I had programmed by punching punch cards and getting a printout, aprint deck back.

 

Kossow: So, had Dave Evans left by that time?

Thacker: He had left, yeah. He had left. That’s right.

 

Kossow: So, it was Wayne [Lichtenberger] , and Mel [Pirtle]?

Thacker: People often believe, and in fact, Dave Patterson believes that I actuallytook a PhD from Berkeley. In fact, they made me an honorary distinguishedalumnus a few years ago, and I asked the chairman of the department at the time,I said, “Wait a minute, but I didn’t graduate from the engineering department.I was physicist.” And then he fessed up. He guiltily said, “Well, the problemis, Chuck, we’ve been going two of these year andwe’re now out of distinguished graduates. So, we’re broadening out net.” And Isaid, “Ah, in a few more years, it will be anybody in the U.C. system.” But, inany case, yeah, Dave had left.

 

Kossow: So, I spent a few days with Wayne. Wayne is a really nice guy, but I neverhad the chance to meet Mel.

Thacker: Yeah, he’s great.

 

Kossow: So, what was Mel like?

Thacker: Well, Mel was an interesting guy. He was the project leader, the principalinvestigator, and very smart, good teacher, not a great businessman. But Inever really got to know Mel very well. I did get to know Butler and Peter, because Peterand I shared an office. But, Mel traveled a lot, for instance. So, in any casewe had planned to build a follow onto the 940, a much larger machine and wewere planning that and figuring out how to do it, andwe finally realized that it would just be very difficult within the constraintsof the university to actually do this. Because we were talking about a lot of money,and although ARPA was a fairly freespending agency at that time, particularly for innovative computer stuff, thiswas even kind of beyond their capabilities. We knew that it would take, youknow, several million dollars and so we looked around for alternativesand we decided to do what was then a startup, although that term had not reallybeen coined yet. And we looked for investors, and Mel and I actually went to New York one time and talked to ArthurRock, and put together the necessary money, although we didn’t really get asmuch as the business plan said we would need, and that’s bad because that waswhat resulted in the demise of the company.We just couldn’t get it funded and there was finally a recession in 1970, whichmade it even harder.

 

Kossow: Right, did the [Univ. of California] Regents end up investing?

Thacker: Yes, they did and I think they are a little bit irritated about that, but,you know, they’re no longer regents anymore. So, it’s water over the dam now,but in any case we built the machine and it worked more or less to spec, and it was quitean astonishing piece of engineering because it had two large drum memories, alarge Bryant disc file. It had core memory, a central core memory.

 

Kossow: Sort of flaky Ampex memory.

Thacker: It was kind of flaky Ampex memory, that’s right, and it was fronted by athing, which today we would call a cache. We called it the fast memory becauseit was very small, of course. And then it had five independent other processors,there was one that handled scheduling for the whole system, one that handledI/O to the terminals, one that controlled the rotating memories, and two whichwere slightly different. They were specialized to beCPUs. So, this was a multi-processor in 1968.

 

Kossow: All microcoded.

Thacker: All microcoded. Microcoded, the machine was very interesting because wewrote, actually I wrote this, and micro assembler which complied for themicroinstructions of the machine, or assembled for them, the output of which was papertape. And that paper tape was fed to a numerically controlled drill, whichdrilled holes in a fixed pattern circuit board, into which you inserted diodes.So, this was a diode matrix memory. We actuallyexperimented with several different kinds of memory because IBM at that timehad been building all kinds of strange microcoded things, transformer, readonly memory, capacitive. We actually had one oftheir capacitive memory machines upstairs.

 

Kossow: Controlling the tape drives.

Thacker: Controlling the tape drives, and the cardpunches, and the line printer,because IBM actually used to build the most reliable peripherals on the planet,and we’d had a lot of experience at the Genie Project with buying flaky peripheralsbecause they were inexpensive. One of my first jobs there was to make a printerthat they had bought, an old chain printer, actually work and so we decided tobite the bullet and rented an IBM Model 30.The IBM Model 30 was one of the very few computers that I have ever seen, infact probably the only one that part of its equipment makeup was an aircompressor. And the reason is, is that the capacitivestorage had to be held carefully between the plates of this capacitor. Thestorage was actually IBM cards made out of Mylar, and the lower capacitancewhen you had made the whole, made the storage work. Andso they wanted to gently push it, so there was a little air pillow on one sideof the sandwich that pressed the card gently and evenly into the arrangement.It was quite crazy but we used diodes, but yes, Iwrote a lot of microcode for those machines.

 

Kossow: So, when did Ed Fiala come out from the East Coast?

Thacker: Fiala came out from, I guess, BBN.

 

Kossow: So, he was at BCC, right?

Thacker: Was he at BCC?

 

Kossow: Or did he come out for MAXC ?

Thacker: It’s possible, oh that’s right. No, I think he came out for MAXC. That’sright.

 

Kossow: Because there’s some conflicting information on him. In some cases, he waswith the BCC folks?

Thacker: Well, he was certainly with the BCC folks to the extent that they all wentto PARC, but I don’t think he was at BCC. So, the machine was working and wewere beginning to commission it, and we were looking around for money to buildmore and we just couldn’t get it. So, in those days—

 

Kossow: So, what’s the connection between Shell Development and BCC?

Thacker: In the early days, we rented time on the Shell Development 940 machine forsoftware development.

 

Kossow: Okay, so there was this disconnect by that time between Cal and BCC?

Thacker: Yeah, pretty much, and that’s reasonable. I don’t think there was anyacrimony there. It was just that, you know, they still don’t like universityfacilities being used by private companies. So, no, Shell Development was just a source ofcompute cycles, because we certainly didn’t want to buy a 940. It was tooexpensive and we didn’t want to program a 360.

 

Kossow: And there’s just this whole body of code that gets dragged forward fromGenie onto the BCC machine.

Thacker: Not really. The machines were actually quite different.

 

Kossow: But you could emulate a 940.

Thacker: That’s true. That’s true.

 

Kossow: But by that point--

Thacker: That turned out to never be a very big part of the whole consist. Therewas new software for almost everything and you actually have a lot of that softwareon your website. I’ve looked at some of those things, including memos that Iwrote 40 years ago that I don’t remember writing.

 

Kossow: Those memos are really fascinating because of the structure. Every startupI’ve been involved with, you’d never write memos like that.

Thacker: They look like RFCS don’t they?

 

Kossow: Exactly, they look like reports that would come out of-- amazing history.

Thacker: That’s just the way we did engineering in those days because you had to doit that way, because you had to keep track of everything by the documents. Imean, if you read Fred Brooks’ book, he talks about what’s needed for alarge software project, and one thing is a librarian to control the documents.

 

 

Kossow: Did that follow through in the early days of PARC then? Because I’ve neverseen the early PARC documents.

Thacker: Yes, it did, although probably to a lesser extent. But if you think aboutit, there were a lot of manuals written for a lot of things and usually themanuals were written by and large before the thing was built and served as specifications. So,right, so we tried to figure out how to move BCC forward and just reallycouldn’t do it.

 

Kossow: And then the recession.

Thacker: And we cut a deal with ARPA to buy it and move it to the University ofHawaii, because they needed one. They needed a computing system for theircomputer center, excuse me, could I get a little bit of water here? So, in 1970 we,having talked to ARPA about getting rid of the machine, it turned out thatBobTaylor who used to be the IPTO Director, but by that time had taken aposition at the University of Utah, but didn’t really like itthere—

 

Kossow: He was only there for, like, a year.

Thacker: He was only there for a year, yeah, that’s right, and we started talkingto him because he had been hired by George Pake to set up a lab for the XeroxCorporation. And he was looking around for people, and of course he turned to hisold ARPA buddies, and so I was actually only in that crew peripherally, in thesense that I was never an ARPA principal investigator, for instance, as Melhad. But in any case, when he went to PARCfinally and relocated, he came up to BCC with George Pake and interviewed manyof the engineering staff. I’m not sure exactly how many, but certainly I thinkall the major players.

 

Kossow: Inside BCC?

Thacker: Yeah, at BCC, because it was pretty clear by that time that the companywas winding down, and some of us actually decided to take the leap. I was alittle bit nervous about moving to the South Bay, because I’d never been therevery much. And I’d actually just had my first child, my wife had our child, andso uprooting her would have been a problem, but I decided that the opportunitywas probably pretty good. Best decision I ever made. So, I actually went beforemost of the others because the software was not quite done, which is the usualsituation in these matters.

 

Kossow: Okay, and they had to deliver something to Hawaii?

Thacker: And so the people who were actually going to do with the machine to Hawaiiwere the ones who packed it up and moved it, and I moved to PARC. So, I wasactually one of the very first people to arrive in the lab.

 

Kossow: So, is 3180 Porter still the same building?

Thacker: No, it’s been destroyed.

 

Kossow: How big a building-- ?

Thacker: 3180 Porter used to be two buildings, which were about, one of them wasabout 100 feet on a side and had sort of an open garden in the middle, sort of.So, it was offices in a ring around the central court, and then there was anothernon-descript building which was not used for much of anything, but it came withthe deal.

 

Kossow: So, you were one of the first people to go over and work for Bob?

Thacker: I think I was the first. I’m not positive of that, but very shortlythereafter, Jim Curry and Bob Flegal came from Utah. Curry was subsequentlyinfluential in the development of the BCPL complier and Flegal went off and worked with DickShoup on graphics. And Dick Shoup came over from BCC. He was originally fromCarnegie and then the rest of the people began to arrive, Butler and Peter.Charles Simonyi had elected to go off and workfor Mel at Ames, which was setting up a center for the newly hatched ILLIAC IV.And so Charles worked there for a while as he was going to graduate school at Stanford, and then he came to PARClater.

 

Kossow: So this will be early 1971?

Thacker: This is early 1971 now, and we looked around for what to do aboutcomputing and we thought about buying a 940, and that would have been verypolitically correct, because Xerox had just bought Scientific Date Systems. But SDS didn’treally like the 940. They were trying to replace it with another family calledthe Sigma Family, and we looked at the Sigma Family with the idea of what itwould take to build a timesharing system on it, anddecided that it was just not the right system. I mean, the Sigma was originallyarchitected as an attempt to compete effectively with IBM in IBM’s markets, notrealizing that if you’re going to compete with IBM’smarkets, you have to have all the IBM software, because the software is whatmatters. And so it was never a very successful machine, and we finally decidedthat it wasn’t really what we wanted because it wasnot what the university research community, and particularly the ARPA communityused for doing research. They used the PDP-10 suitably modified by a set ofhardware that had been built at BBN called theBBN Pager.

 

Kossow: So, you had never thought about doing another BCC?

Thacker: No. Well, I think we were all pretty tired and we were investing a lot insort of setting up the new lab, you know, sort of building the necessaryinfrastructure to make it actually tick. And so after quite a bit of thought, we concluded that theright thing to do was,since it would be unseemlyto buy a PDP-10 from an SDS competitor, that we would build one. And so wedesigned the MAXC system.

 

Kossow: What does MAXC stand for?

Thacker: Multiple Access Xerox Computer. It is also a play on the name Max, as inMax Palevsky, who was the President of SDS. And that was actually planned anddesigned very rapidly. I mean, it was an interesting machine because what wedecided to do was rely on commercially available packaging technology, primarily.And that meant wire-wrap cards that you could by from Augat, and backplanes, and so on and so forth. So, everythingwas wire-wrap.

 

Kossow: Was Dave Poole or Pettit, or any of the guys from SAIL involved in thedesign?

Thacker: No, they were not. They were up on the hill doing other things at thattime.

 

Kossow: So, it was you and--

Thacker: Pettit actually joined SRC later for a while, until he went off toPittsburgh to be a mystery writer, right after they did the Foonly.

 

Kossow: So, it looks like from the specs that, I think the first MAXC came up inthe summer of ’72.

Thacker: Something like that, yeah.

 

Kossow: Because there’s an 8.1 version of the spec that says it’s going to be andthen by July it says it is. So, it’s probably, you know--

Thacker: I mean, that was a long time ago.

 

Kossow: One of the bits of folklore is the story about doing your own 10 and thenhow it comes up so fast.

Thacker: Right, well we were very careful because we knew about the second systemeffect, right. BCC was the second system. It took a long time to get right.

 

Kossow: So, yeah, this is the third. You did Genie, you did BCC--

Thacker: And by that time—

 

Kossow: You knew how to build timesharing systems.

Thacker: We knew how to build systems and so the only thing that was kind ofaggressive that we did was we said, we will not use core memory. We will usethis newly invented semiconductor RAM called the 1103.

 

Kossow: Which was a total nightmare.

Thacker: Which has an interesting story. So, we designed the memory board for themachine. I remember it well. It had 96 1103 chips, and they were semiconductormemory chips with external sense amplifiers. And these were packed in arack with four back planes, each of which contained, had to get it right, 12 ofthe cards and controller, or maybe, no, it was more than that. And then therewere four cabinets, and they were strung togetherby cables. We designed our own cables and connectors for them. Our associationwith Jack Hawley began at that time, because we had him make the connectors. Very nice engineering job for lowprice, but they were all twisted pair and transmission line terminated at theend. So, electrically the thing was quite conservative and it worked very well,except for the 1103. But we knew that there was a potentialfor problems with the 1103, so we put single error connection and dual errordetection in the memory. And when the system came up and the CPU started beingable to run memory exercises, the memory workedflawlessly. And so later, when we got ready to do the Alto, we said, “Well,let’s just use the MAXC boards as the memory for the Alto, because we know howthey work. They work very well, and by theway, we don’t really need error correction. We can just put parity on thememory.” Mistake, because the thing that I didn’t realize is that although Ihad put extensive logic into the memory system for correctingthe errors and reporting the existence of single errors to the system, thesoftware guys never read. And so the thing was actually sitting there making alot of errors, but it was fixing them and so we nevernoticed. And when we put it into the Alto, the same result happened. It lookedvery reliable, very rarely got a parity error until we started running morecomplicated software that exercised the memory morethoroughly and then suddenly we would see various failures. And so this turnedinto an arms race between me and Intel. I would write a diagnostic to trackdown some particular perverse flaw in the1103, and then we’d run it. By that time, we had e-mail and we ran it overnighton the Altos, and the machine would automatically detect errors and send e-mailto the technicians. So they could come in, inthe morning, and change out the bad boards, and then run them through anothertester that we built, that could by that time do very comprehensive testing ofthe board. After that, we always used errorcorrection and I’m amazed that present day PCs don’t. Servers do but desktopmachines do not, and I actually suspect that many of the bad events that happenon the PC are actually memory errors, becausememory’s not as reliable anymore and it’s getting less reliable. So, in anycase the MAXC came up and we used it. We subsequently built, I think, two more,one or two more, which were slight improvements, butnothing serious, because by that time we finally understood that we ought tohave built the Alto and we did it.

 

 

Kossow: So, there’s that run up to that in ’72 when there is the TFS, all thetimesharing stuff that was going on. VTS.

Thacker: Yeah, there was another system being built in the laboratory, not in ourlab. There was also a thing called the System Sciences Lab, which was directedby Bill Gunning of GTL fame, and they had built a system based on interconnectedNova mini computers.

 

Kossow: So, is that really where the affinity for Novas starts?

Thacker: POLOS, no, actually it started before that. They chose the Nova to someextent because we did it, and we did it because we actually did a bakeoffbetween the PDP-11 and the Nova, and for our purposes, which was programming inhigher-level languages, the Nova actually worked out to be better because itwas faster. And it got through its instructions faster, and the 11 was betterif you wanted to write assembly language program, do assembly languageprogramming because that’s whatGordon Bell designed it todo. And so there were always several different ways to do the same thing on the11, and so the compilers always had to figure out, well, which one do I use inthis circumstance? And it made the compilers more complex, and we didn’t likethat. So, we liked the Nova and of course we patterned the Alto on the Novawith a few improvements.

 

Kossow: Pretty much all of your tools for doing MAXC and everything were runningon Novas?

Thacker: Correct. All the debugging stuff for the Alto itself ran on a Nova, but wenever really thought about using them as part of a distributed system, and thePOLOS, PARC Online Office System Group, which built the system out of amultiplicity of Novas had that idea in the beginning. The Nova had kind of aclunky network for interconnecting something like 16 machines.

 

Kossow: So, that was like Chris Jeffers and Bill English?

Thacker: It was Bill English, and was Chris there or was he with Alan at that time?No, I think he was with Alan Kay in the learning research group. He had a thirdgroup. So, we had been thinking about smaller machines. We had a couple ofIMLACs and really liked the idea of interacting with this screen in a way thatyou couldn’t really do with glass teletypes. But the IMLAC was kind of badbecause it used a magnetic calligraphic display, where the beam is drivenaround to various positions and then turned on and off, so you can draw lines,and you can draw characters that way, but they’re kind of ugly. And you have todraw them pretty rapidly because you have to repeat 30 times a second. So, amuch better solution is the raster scan display, and in fact the POLOS peoplehad designed one, which used a custom character generator to generate more orless arbitrary fonts, and it looked really nice. And we kind of liked that, butit was expensive. We said, “Well, if we’re going to build a lot of thesethings, it’s got to be cheap.”

 

Kossow: And that’s where the industrial design for the display came from?

Thacker: The industrial design for the display was done for the POLOS project.That’s right. We just took it, and their display actually had a little bithigher resolution than ours, but we thought about it a bit and decided that 606 x 808 was probablygood enough. And it was good enough that if we could produce an arbitrarybitmap, then we could really do a lot of pretty interesting things with thatdisplay, and we did. The display controller was extremely simple, because theAlto was arranged in such a way, and this was something that had actually beeninvented once before but we didn’t know it, on the TX-2. When we heard about it, Ed McCreight and Iwere talking to Wes Clark and he told us that he had done something similar onthe TX-2, and we said, “Oh.”

 

Kossow: The task multiplexing?

Thacker: The task multiplexing, that’s right, and Wes said, “Yeah, the paperappeared in FJCC in 1965 or something,” and Ed said, “Well, Wes, the onlyexcuse I can make is that at that time I was five years old, or nine years old, or somethinglike that.” But in any case, the task multiplexing was very fast. You couldswitch tasks in one cycle, and the result of that was that we kind of turnedaround the way computers had been architected until then.

 

Kossow: But you had done similar things with MAXC already, with the disccontroller.

Thacker: Not as thoroughly. I think that was kind of an exploratory probe, but whatwe wound up doing was making all the hardware for the device controllers very,very simple and doing the hard work in microcode.

 

Kossow: And by that time, you had bipolar PROMS, so you could actually havereasonable sized micro stores.

Thacker: Reasonable sized, but still not great speed. So, where reasonable, Ithink, was 4K instructions rather than 1K. And so in a normal minicomputer, thememory is multiplexed between the IO devices plus the processor. Now, that’s not theway it works in the Alto. In the Alto, the underlying microcoded machine ismultiplexed between a lot of tasks and it owns the memory. So, we multiplexedthe computing rather than multiplexing theaccess to memory and that leads to an enormous simplification in the devicecontrollers, which makes the whole system cheap, particularly if you likecomplicated devices, which we did.

 

Kossow: Well, maybe this is a good time to break then. 

Kossow: It’s August 29th,2007. This is tape number two, oral history of Chuck Thacker. So,we were talking about the early days of the development of the Alto.

Thacker: So, we designed it. It was actually based on a memo that I had written andthen a follow-up memo that Butler had written to the laboratory about why weought to do this. And finally decided to do it and built about a half dozen of themand see how it worked. And so we used all the same technology that we used atMAXC, and the memory boards, and the back planes, and the so on and so forth,but with a much smaller design. And I hadworked out the design based on this idea about multitasking. And so we dividedthe machine up between Ed McCreight who did the disk controller and diskcontrol microcode. I did the CPU, and the CPUmicrocode, and the display controller, and that was really all initially. Alittle bit later, like six month, Dave Boggs and Bob Metcalfe started, when werealized we needed a network, Dave Boggs and BobMetcalfe started doing the Ethernet. But the original machine was just thedisplay controller, the memory, the CPU, and the disk controller. I thinkthat’s right and later there came a few other things added.The Ethernet controller was one and more complex controllers for things likelaser printers were also built a little bit later. But the early machine wasjust that, it was very simple and the microcode was all inROM, Read Only Memory, not programmable and so that posed a bit of a problemfor doing things like debugging. We built a large gizmo that you strap on theback of the Alto to give it read-write control store andalso run a debugger interface so you could do a credible job of debugging themicro program as it ran on an attached Nova. And we brought the machine up thatway. The whole thing took about, literally,three months.

 

Kossow: So, fall of ’72?

Thacker: We started in the winter of ’72 and finished all six machines by themiddle of ’73, I think. The timeline sometimes fades into oblivion, but in anycase that first phase was very, very quick and the machine was ready, of course, as isusually the case, ready before the software. So, we had a fair amount of timeto sort of hone things in preparation for having to run the real software.

 

Kossow: Gene McDaniel was working on the initial-- ?

Thacker: Gene McDaniel worked on, yeah, a lot of the initialization stuff and someof the microcode, and debugging it. We worked very closely together debuggingit. Yeah, that was when I first realized that we really had something with bitmapdisplay because we got, in a micro program, no CPU program, we got the SesameStreet Cookie Monster to eat a cookie. And that was one of the few aha momentsI’ve ever had in graphics, the others being what,when I first saw a PC running color graphics, when I first saw Doom running ona PC, amazing piece of work. And what recently? The other stuff is allevolutionary. In any case, pretty soon we were beginningto use them for computing in the laboratory and then, of course, all of thegood software that was built for the machine was actually written.

 

Kossow: So, how did that end up happening? The people just needed a tool and theyjust started writing it?

Thacker: Well, we were lucky because we actually had a goal for the laboratory,which was essentially to create the paperless office. We didn’t realize thatthat was not what we were going to do at all, and in fact we were going tocontribute to the death of a lot of trees, but fortunately trees are relativelyrenewable if you’re using the kind you use for paper. But in any case, so weneeded things like text editors for program editing. We needed, you know,compilers to compile programs. We needed a little operating system just to runthe IO system. We needed a disk file system. Most of that stuff was all stuff thatwe needed or wanted to use to satisfy our own needs for various things.

 

Kossow: What was the size of the group? Was it a few dozen?

Thacker: Oh yeah, a couple dozen. And so the high quality text editing and thelaser printer grew out of our desire to produce camera-ready copy for papers. Idon’t remember if you remember how that used to work but you started out with alarge piece of poster board, essentially, that the journal sent you, with thedisplay area outlined in light blue and then you used rubber cement and a pairof scissors to format your text. A very time consuming and error prone, andactually ugly process, and so we stopped all that. We could producecamera-ready copy immediately. And the same thing with laser printing, I meanthat grew up because we wanted it. So, we were lucky because what we could dois we could do work that was actually new things in the world because we neededthem and the motivation of being able to do things that you actually want isextremely high. And it’s hard to do today because you have to sort of cast yourselfinto the representative of a much larger community. I mean, and if you’re inMicrosoft you’re thinking about how 100 million people or half a billion peopleare going to use what you do.

 

Kossow: That’s the problem that all operating system architects have, that youhave to project yourself out. How is the system going to be used?

Thacker: Right, and we knew because we were the users, so we were in a fortunateposition, which is very hard to do today. But, let’s see, where was I,wandering around. The machines worked. We basically started. We built afollow-on. We built an Alto II. I didn’t have too much to do with that,actually. I was moving on to other things by that time.

 

Kossow: And so it was John Ellenby.

Thacker: That was John Ellenby. Yeah, John Ellenby actually did the engineering andhe brought on board a team in Los Angeles called the special products groupthat had done engineering within SDS, and so they designed the Alto II.Fairly nice implementation. By that time, we thought the Alto was getting alittle old and, you know, it was now maybe three years old from start.

 

Kossow: So, you were sort of sucked into the office system, and the D0 then?

Thacker: Not immediately, no. The thing that came first was the Dorado [ph?].

 

Kossow: Oh, really?

Thacker: Yeah, because we thought, I had actually been doing some work on cachingand I realized that caches were really going to make a big difference incomputing performance. And so I thought, and also, you know, ECL Technology, emittercoupled logic, had become relatively inexpensive, and so I thought maybe itwould be possible using ECL technology to make a Nova-like machine, or an Altolike machine that could run with a 40-nanosecondclock rate, which would be unprecedented. I mean, that’s a really fast machine,particularly in a personal machine. And it would have a cache, and a little bitlater we decided it should have a virtual memory system as well because we werebeginning to push up against the limits of the standard problem that architectsalways have, not providing enough address bits.

 

Kossow: Then there was the whole Mesa--

Thacker: And the Mesa Development, Mesa needed more space and, you know, we werejust getting more sophisticated. And so we started working on the machine andit evolved over a fairly long period. This was not a fast track project, ittook several years to actually get it working, and it had its up and downs,because at that time some of the members of CSL had gone off and SSL primarily,had gone off and formed actually a business divisionwithin Xerox called the system development division. This was run by DaveLiddle and Dave said, “How would you like to come over to SDD and help us bydesigning a machine that we can use for an officesystem that we can sell?” And I said, “Well, the only thing I have at hand isthe Dorado.” And he said, “Well, that might be okay.” So, I went over to SDDand I was there a for a couple of years, but it turnedout that the Dorado was way too expensive for general consumption at the time,so we investigated a little bit using it as a controller for very high speedlaser printers, because those things were actuallybeginning to be made by Xerox commercially at that time. They were just gettinginto the business and as usual, they did a real good job at the high end firstwith a thing called the 9200, which was based on the upperechelon of Xerox copy machines. And that would have required an extremely fastcontroller but they decided that the 9700 should be operated with a controllerthat was more specialized, more custom. And sothe Dorado was out of that picture, and so that was when the D0 started. Thefirst idea with the D0 was to build a machine that was much more Alto like, useTTL technology, rather than ECL, whichcould be built relatively inexpensively with the technology of the time. And soBrian Rosen and I, Brian had come to work for us from CMU, I believe, designedthat over a period of probably six, eight months.

 

Kossow: So, did you split the design between Brian and yourself?

Thacker: Yeah, I don’t remember exactly how we split it up because I remember I didthe memory system and I probably did part of the processor, and Brian did mostof the IO.

 

Kossow: D0 is weird because it supported multiple displays.

Thacker: Yeah, it did a lot of interesting things.

 

Kossow: And the boat anchor, the three-wire interface.

Thacker: Well, we wanted a machine that would run the CAD systems. We werebeginning to get into LSI in part of the Mead & Conway thing. I mean, LynnConway was down the hall in SSL working on the book with Carver. And so one of thethings we wanted was the ability to run a graphics editor that you could usefor chip layout, and that requires color. So, the D0 actually had the firstcolor display that we built which ran a lower resolutioncolor monitor, but it was okay for that purpose. But the D0 was never really acommercial success, although there were a few of them sold as LISPworkstations.

 

Kossow: The 1100?

Thacker: The 1100, just as a few Dorados were sold as LISP workstations.

 

Kossow: 1132.

Thacker: 1132, right, but the problem was those machines were really too expensiveand they were beginning to be, the performances that you could achieve with athing like that were beginning to be achievable with micro electronics.

 

Kossow: The D0 was odd. It doesn’t have any removable storage either. It just hadthe Winchester, so you had to have it on a network.

Thacker: Well, we had networks so that wasn’t actually a problem. And so we built afew D0s and, you know, we used them within CSL, but the Dorado came along soquickly in such a short time after, because when we had stopped working onthe Dorado and SDD, the project went back into CSL. At which point Butler andseveral other people kept working on it, and in fact morphed it into somethingquite different than we had originallydesigned, and wound up with something that was actually quite nice in terms ofits characteristics. I mean, you could read the reports and—

 

Kossow: Yeah, I was just looking down the list of names. Ed is on there.

Thacker: Yeah, Gene McDaniel, Ed Fiala, Severo Ornstein worked on it. Quite a fewpeople worked on it.

 

Kossow: It was a big machine.

Thacker: It was a big machine.

 

Kossow: Loud too.

Thacker: It is not the loudest computer that I have ever worked on, but it’s close.So, let’s see, right, so Imoved back into CSL along with the D0 and CSL thenbuilt a number of other small machines sort of branching out from that. They built theDandelion, which was used finally as the basis for the Star Workstation.

 

Kossow: So, that initially was a CSL design?

Thacker: Yeah, Butler designed it. It used a somewhat different form of taskingthan the Alto did, but it was pretty simple. It also used a bit slice processoravailable from AMD. So, it was much more economic. So, that was the Dandelion.There was a machine built in SSL, or no, not in SSL. It was built by Dave Boggscalled the Dicentra, the purpose of which was to provide a D-class machine. Bythat time, we actually had a name for theclass, the D machines, which had a standard IO bus, so you could run standardperipherals.

 

Kossow: Do you remember if the three-megabit Ethernet at the time, came out ofStanford? Or did that come out of-- ?

Thacker: The three megabit Ethernet? No, it came out of—

 

Kossow: For the Dicentra, the multibus?

Thacker: No, I don’t remember.

 

Kossow: It shows up at the Sun workstation, so I wasn’t sure who did it.

Thacker: Well, a lot of ideas that came out of PARC showed up in the Sunworkstations, it turns out. I remember visiting with Andy Bechtolsheim in thetop floor of Margaret Jacks Hall, because they actually wanted to buy displaysfrom us, so they didn’t have to make a display of their own for the Sun-1. The problemwas, we couldn’t figure out how to sell it. 

 

Kossow: In that time frame, you’re into kind of the world starts hearing about it.There’s the university grant program.

Thacker: Well, we had done the university grants, which were extremely successful.

 

Kossow: That happened in ’79?

Thacker: Something like ’79.

 

Kossow: So, it MIT, the University of Rochester, CMU, and Stanford?

Thacker: And Stanford, right. And they got a half dozen, or eight, or something Altosand a Dover printer. And of course that’s what they wanted. They all wanted theprinter.

 

Kossow: Because at that point in the 1980s, all those institutions startdistributing their papers.

Thacker: Right, and in fact there were Altos as a probe by a little organizationthat Jerry Elkind had started. I can’t remember what he called it, but Charleswas there, Charles Simonyi [ph?]. They put Altos in the Carter White House and they werefairly successful too, you know, for electronic mail anddocument preparation.Of course, the White House produced a lot of documents.

 

Kossow: And then there’s the GIN publishing.

Thacker: There was a thing done at Ginn called Gypsy and again, a lot of that wasleveraged by the existence of laser printers, which were primarily, you know,designed by Starkweather and Sproul, and Butler to some extent, and I helped.So, yeah, by now the 70s are winding down and we’re moving into the 80s, and webegan to look around for something really new to do. And Taylor had gotten intoa tiff with the Xerox management because hedidn’t think we were being well enough supported, and wound up in a lot of usdeparting. And quite a few people did stay on, and of course PARC kept goinguntil Xerox recently spun it off, I think.Excuse me, I’m going to grab some water. So, in ’83 a few of us left and lookedaround for something to do, talked to several companies about setting up a labfor them and finally wound up setting up the lab forDEC. We went back East and talked to Sam Fuller, and Bill Strecker, who wassort of the CEO of DEC at the time, and I think Gordon Bell, and they finallydecided that it would be a good idea to have aresearch lab. They hadn’t done much of that in the past and then they hadgotten the Western Research Lab, which had been started up by Forrest Basketwell before, about a year before. But we wanted tostart another lab and it was a little bit confusing because WRL [ph?] was justdown the road, and why weren’t they the same, and well it because of theresearch management styles of Forrest and Bobwere quite different.

 

Kossow: So, had DEC started up with Cutler and those people by then, or is thatlater?

Thacker: That’s later. That’s quite a bit later. That’s ’89.

 

Kossow: So, at this point, Forrest and Smokey are doing Titan?

Thacker: Neil Wilhelm.

 

Kossow: Neil Wilhelm?

Thacker: Yeah.

 

Kossow: Okay, so Titan is going on.

Thacker: Titan is being built now and we had, in the last couple years, beenlooking very closely at multiprocessors, because that seemed like a good way tosort of get a step in performance. Which, if you did it right, it was a sustainablething. You know, Moore’s laws carrying you along this curve, and you step up toa difference curve.

 

Kossow: So, McCreight stays behind and does Dragon?

Thacker: That’s right, he stayed behind and elaborated, and finally implemented,along with Sun in some way that I could never quite fathom.

 

Kossow: Robert Garner is over there.

Thacker: Right, Bob Gardner, the Dragon, which was a machine that I had started theproject and named it, but the main thing was the idea of exploring the effectof multiprocessors. But then we left, and so some of them stayed and continued it,and by that time I was real busy starting up SRC, and so I don’t know muchabout what happened to it. You’d have to talk to those guys. So, at SRC weactually decided before we started it what our first sort of softwareenvironment would be like, and hardware. And on that basis, Sam Fuller andButler, and Bob and I sat around in a meeting room at the Concord Inn, in Concord,and sort of planned out the first year of SRC. And then we went back to theWest Coast and did it, got the building on the corner of Lytton and Alma andbegan to hire people, not just people from PARC, although we would, but wenever approached them. I mean, we didn’t really recruit our colleagues fromPARC, because Xerox told us that if we did, dire results would—

 

Kossow: Ed Satterthwaite comes over then, or was he already-- ?

Thacker: So, Ed came over and Butler came over. Butler actually came over a littlelater. But by that time, Butler was living in Philadelphia because his wife hada position there at UO P, just as he now livesin Boston because she has aposition there. So, he was a commuting member of the staff. Let’s see, Charleshad gone to Microsoft by that time, and who else moved here immediately? Well,Phil Pettit joined us right about that time. I have a photograph of the firstfew people, Roy Levin, Jim Horning, Mike Schroeder were all part of the earlycrew. Boy, I wish I remembered exactly who’s on that photograph, because it’son my sort of kitchen wall. I should. So, we began to do the Firefly, which wasan up to sixway multiprocessor, initially using 68000. But before we builtthem, we switched over to the MicroVAX because it seemed like it would be abetter thing. That was a new chip processor from DEC that was a single chipfax.

 

Kossow: The MicroVAX II, the one that DEC worked on?

Thacker: Well, it was MicroVAX actually. The MicroVAX II was a little bit later.But, in fact, I think the MicroVAX II was the one that actually went into themachines, you’re right.

 

Kossow: But you ended up doing a lot work on snooping protocols than on the-- ?

Thacker: That’s right, we did. Actually, it’s interesting. We have some of theearliest patents on snooping protocols, but nobody knows it because we didn’twrite a whole lot of papers about it. But in any case, that came up and alongwith it came a programming environment that allowed you to program themultiprocessor, and do a good job at writing programs with threads, which issomething that programmers today still don’t do. But in 1985, roughly, we had afull-bore operating system that used threading very extensively and ran quitewell on six processors, thank you.

 

Kossow: The environment is interesting, because Niklaus Wirth does his sabbatical,goes off, builds Lilith, does Modula 2, and then you end up using Modula 2.

Thacker: Well, we, of course, knew Niklaus very well and still do. We vacationedtogether in France last year.

 

Kossow: These interesting things, like Rosen does the PERQ.

Thacker: Right, yeah. Brian actually went off to do the PERQ deliberately on ThreeRivers , and basing it a lot on the D0, although he had the much nicer displaybecause he was working with a guy whose name I do not remember. I never met anyof them who had designed this marvelous 600-line resolution, no 300 lines perinch resolution display, beautiful. Put up a drawing of it and it wasmarvelous. But in any case, we got that machine working and then we used itquite successfully. I don’t think we built very many extensions to it, and thenDEC began to think seriously about the Alpha. And actually—

 

Kossow: The PRISM/Alpha Wars?

Thacker: Yeah, the PRISM/Alpha Wars happened at that time, but a little beforethat, we essentially changed our focus. We were no longer solely interested incomputers and we started getting interested in networks, because it was prettyclear-- it was not clear then that the Ethernet, which was then over ten yearsold, was going to survive the way it has, and it has incredible longevity. Infact, it’s really like George Washington’s axe because it’snothing like the original, it’s just the name and the idea of a 1500-bytepacket. But, we explored some networks that were pretty interesting for theirtimes. We built a network that we first calledAutonet until the company that ran somenetwork for selling used cars threatened to sue us. And that had some prettyinteresting properties as a local area network, because the network switches were actuallyactive. They had programs in them and the thing could organize itself to beself-healing when something broke. So, all the workstations were dual homed totwo switch ports, and if a switch broke, the entirenetwork would recognize that and reorganize itself into a potentially new spanningtree that allowed you to route messages from everywhere to everywhere else. Andit was a switched network, so it was actuallyquite fast. The signaling was actually done in a slightly weird way. It wasdone bidirectionally on a 75-ohm coaxial cable, which turned out to be kind ofa bad idea, but I was always trying to do interestingelectrical engineering things. But, that network worked out very well, and sowe, at that time there was a lot of discussion about, you know, what’s longhaul networking going to look like? Is it going to be fasterEthernet? Is it going to be ATM? Is it going to be a network called FDDI, whichDEC was pushing, for fiber distributed data interconnect. And we decided weliked ATM networking because it was prettysimple, and we decided to build an ATM switch. And so we built a switch thatwould switch 16 622 megabit per second ports, and this was a big crossbarconnected with my very competent line cards that hadfiber optic interfaces on them. Fairly substantial piece of engineering, allcontrolled by a small control processor for the switch just to set it up, andthe fabric was actually scheduled by microcode. It ranin a Xilinx Chip and set up the crossbar at one ATM cell intervals, and it didthe necessary calculation to discover how to configure the 16 x 16 crossbar tomost efficiently route the messages between the inputports and the output ports. And that was kind of nice, but I still rememberinventing the algorithm because I invented an algorithm to do that while lyingon the beach in Hawaii. But the problem was, I couldn’tquite figure out how rapidly the process would converge. So, I went back and Italked to one of the theorists, and he said, “Well, I don’t know, you know,this is a matter of bipartite graph-matching and that’sNP-complete.” And I said, “Well, I don’t care about getting a perfect solution,I just would like to get a good solution. So, does this do that and how fastdoes this converge?” And he came back a couple ofdays later and he said, “Oh, the proof was trivial. It converges in fourcycles.” So, you can run this algorithm for four clock ticks and then you haveyour switch schedule, and this was subsequently elaborated by a guy named NickMcKeown who’s now a Stanford professor, into an algorithm that was used tobuild a terabit router. So, I was quite pleased at that. So, we built thisthing and it eventually worked, and we used it, replacing by and large the AN1 network,and we participated in a test bed that was being run by Sprint around the BayArea. They actually put fiber into our buildings and we got a lot of trafficstatistics on that network, and one of the things we did, that we were actuallypushing quite heavily was ATM does not use flow control. And as a result, youhave to be very careful about arranging connections through the network,because otherwise you can lose packets. And I have always believed that loosingpackets is a sin. The user, after all, the customer paid you to deliver thosepackets and dropping them on the floor is sort of a violation of the trust betweenyou and your customer. And so, we actually worked out a scheme for doing creditbased flow control along a link that meant that the network never droppedpackets, I mean, unless there was an error. We actually tried to sell this tothe internet community and they weren’t buying. They believed in TCP, you know.Congestion control is signaled by loss, and of course that’s not true at all.Congestion control should be because ofcongestion, not because of loss. Loss is because of errors. So, in any casethat thing actually became a DEC product. It was called the Gigaswitch/ATM, butat that time there was a lot of interest in networking andwhich protocols, and, you know, what underlying technology would win out. A lotof different contenders, and of course, Ethernet just came in and swept thefield. And so, it all looks kind of quaint from the 21st Century to look back at that time, which wasby now the late 80s and see the sort of discussions that went on in the variouscommunities that were interested in these issues. Because, of course, the Internet wasgoing by that time, haltingly, but going, and so, you know, this now because aquestion, it had a lot of potential economic value. And it was no longer just alot of people sitting around in IETF working group.So, in any case, that was our little foray into networking and then we builtanother computer, because Bob Supnik, who was at that time the project managerfor the overall Alpha program, including the productionof the chips and the generation of the necessary software, came to me and said,“You know, Chuck, I’ve got a problem. The programmers aren’t programmingbecause they don’t have a machine to programand they never write code for simulators. And so we are going to wind up in theenviable position of having our first products and not being able to sell themfor a year. How can you help mitigate this?” And Isaid, “Well, you know, maybe we can do this,” because they had been building achip that was never actually sold, called the EV3, which was a test chip forthe Alpha, which included a fully functioningprocessor, but very, very small caches and it had the desirable property thatits IO pins could be either ECL or TTL, or CMOS interface levels. So, we tookthat thing and we built a machine around it, which was actuallyanother one of those—

 

 

Kossow: Larry Stewart

Thacker: This was Larry Stewart, Dave Conroy, and myself.

 

Kossow: So, Dave [Conroy] is out here by that point?

Thacker: Dave came to work for us. He was thinking of leaving DEC and we talked himinto coming to the West Coast instead Similarly with Dick Sites who was inroughly the same boat. I’m glad we did. So, we looked at building this machine andwe built a machine called the ADU, which was the nosiest machine I ever made.This machine was built on printed circuit [ph?] boards that were about this bigand a board was either a memory board, or a processor board, or an IO board.And we decided very early on not to build a lot of IO devices, but instead tobuild a high-speed channel off to a VAX that could run the IO devices for us, just the way we ranthe tapes on the BCC-500. So, we designed this 50 megabyte per second channel,and Larry Stewart designed that board, Conroy designed the CPU board, which included the Alphas and its caches, andI designed the memory system, and sort of architected the overall thing. And itwas a stellar success. It really worked. We built about 50 of them, distributedthem around the company and the programmersstarted programming them because they could, and Supnik has told mesubsequently that he believes that the existence of that machine cut one yearoff the ship time of the Alpha. Which sort of in thelong overview of history is probably not very important, since it’s dead now,but there you are.

 

Kossow: Let’s take a break for a while.

 

 

  

Kossow: It’s August 29, 2007, this is tape number 3 of the oral history of ChuckThacker. So we’re just to the point where DEC is about to come out with theAlpha so you’re doing prototyping machines at SRC.

Thacker: Right. So we built these ADU machines. We actually wrote some nice papersabout that, which appeared in ACM, communications of the ACM in ’93. But wenever built very many of them for our own use because they were very expensive.ECL is an expensive technology. They actually used, they had a lot of memory bythe metrics of their time, which was very expensive in those days, and we reallycouldn’t justify the use of those things for research.

 

 

Kossow: So most of the research machines at that point at SRC are UNIX machines?

Thacker: They’re uniformly UNIX machines, yeah. DEC, they’re standard DEC productsrunning UNIX with, you know, somewhat larger DEC machines in the basement in themachine room.

 

Kossow: So sort of the last gasp of the Fireflies was early 90s or?

Thacker: Yeah, ’93, ’94. Some people still kept them running, but not for any goodreason.

 

Kossow: We have a couple of them in the collection.

Thacker: Right.

 

Kossow: So let’s see, so when the ADU wound down—So did SRC end up gettinginvolved in the PRISM alpha wars?

Thacker: A little bit. A little bit. We were involved in another war—

 

Kossow: Do you want to talk about the background at all?

Thacker: Well, I can in a second but yeah, the earlier war was a thing called theBird Wars, where we were trying to decide whether to build product machinesthat were representative of the machines that were being designed in Marlboro, whichwas the VAX 9000 or the machines that were being designed in Littleton, whichthe VAX 8200 is an example, a much kind of sleeker machine. And DEC was justabout to take an enormous beating on the VAX9000 because they didn’t work fundamentally. And they were dominated by thelower end VAXs in CMOS silicon. Dan Dobberpuhl had finally you know, worked out how to build an extremely fast VAX in achip. Which seemed fairly magical at the time but of course it’s ho-hum today.

 

Kossow: Right, and then Dobberpuhl goes on to do lots of other stuff.

Thacker: Oh, he’s done lots of other things. He did the StrongARM, he went andfounded the company to do PowerPCs called CyByte, which was an insult toBroadcomm. He started a company called Palo Alto Semiconductor which he’s at now,producing, I beg your pardon, the Broadcomm thing was a MIPS, maybe it was aMIPS because PowerPC is being done by the PA Semi .

 

Kossow: Right, so it was Where Satterthwaite goes?

Thacker: And that’s where Ed went, it’s where a guy named Mark Hayter that wasinteresting guy who joined us out of Cambridge, joined SRC. Those are the twoprimary people, although several people from the Hudson Enterprise on the Eastcoast joined up with Dan. But,

 

Kossow: So you have the same us verses—

Thacker: Right, so the Bird Wars, I was the judge in the Bird Wars, and I finallysaid, you know, these mainframes don’t make much sense for the same reason thatthey had made no sense for SDS, you know, IBM is just, if you want tocompete with IBM you’ve got to have the IBM software. We didn’t, and so thatcoupled with the fact that the machines were vastly expensive and fundamentallydidn’t work because we used this weird packaging technology.

 

Kossow: So you said the Bird Wars?

Thacker: Bird Wars.

 

Kossow: So which were the Bird Wars?

Thacker: Raven and Eagle.

 

Kossow: Ah.

Thacker: So in any case, later when we were doing the Alpha, there was a lot ofdiscussion about the operating system that Dave (Cutler) was working on for itand the machine that ought to run it. And Dave wanted the operating system PILLAR tointo in a machine whose architecture was called PRISM.

 

Kossow: This is Dave Cutler?

Thacker: Dave Cutler, right. And Dave was the primary software architect of the VAXfamily and very highly respected guy in Microsoft.

 

Kossow: Still at Microsoft?

Thacker: Now within Microsoft, still at Microsoft, still highly respected. He’s asenior technical fellow, one of the few people above me. In any case, he hadset up this place in Washington called DEC West in Seattle in fact and Redmond, andwhen the company finally decided that it was not going to pursue that, hedecided to leave and go to Microsoft. And that worked out very well for him.But, so we moved forward on the Alpha. At that time, Idon’t remember, I guess we were, actually we were building the AN2 network atthat time. And then that wound down and I was looking around for some otherthings to do and by this time it’s ‘96 and Microsofthad been talking to me about joining them for awhile, since about ‘91. I’dalways believe that there was probably no place for me because they’re not ahardware company and I’m primarily a hardware person.Although, I have done a lot of software. I know how to do it and all of that.

 

Kossow: When did Butler end up going over?

Thacker: Butler went about a year and a half, two years before I left. And so hewas highly instrumental in recruiting me, as was Jim Gray, who had been therefor awhile. And, of course, Rick Rashid I knew very well because one of Rashid’s earlyclaims to fame was Alto Trek .

 

Kosssow: Right.

Thacker: Which has become sort of a cult classic in the field. And so I had talkedto Rick and to Nathan Myhrvold and to Butler and to Jim and ultimately to Billand, what I was doing was talking, trying to talk them into setting up a lab in the BayArea. And they decided they didn’t want to do that because they were worried,Microsoft was at that time, and today even still is a bit, very,Redmond-centric All the development is, most of the developmentis done in that area.

 

Kossow: So this is 1996?

Thacker: Late 1996.

 

Kossow: Okay, so there’s the Mac products and had Web TV started? So there’s a fewthings going on.

Thacker: I don’t think Microsoft had bought them by that time but—

 

Kossow: Yeah.

Thacker: But they had started. And so, finally, this must have been in the latesummer of 1996 because, yeah because Roger Needham, who had been visiting usfor a month a year for 20 years roughly, and I were both talking to Microsoft at thesame time. And what would transpire is some Microsoft people would fly down fromthe north and rent a hotel suite in San Francisco and in the morning I’d go upto talk to them about a lab in the Bay area and in the afternoon Roger wouldtalk to them about a lab in Cambridge.

 

Kossow: This is Cambridge, England?

Thacker: Yes, and we would, you know, cross paths through the door and then, andthen Roger would come back to SRC where he was based and we would comparenotes. But finally the company decided that it didn’t want to set up a lab inthe Bay Area so I began to look around for other things to do because I knewthat, I knew that DEC, the writing was on the wall for DEC and I was justlooking around but, you know, by that time—

 

Kossow: Ed had gone off to General Magic and--

Thacker: Yes. By that time my kids were grown and gone and my wife had retired andso we had very little holding us and we said let’s just take a year off. Let’sdo a sabbatical somewhere, because we’ve never done that, you know, let’s eithergo to either to ETH and do a sabbatical with Klaus Wirth (ph?) or go to theCambridge Computer Laboratory and do a sabbatical there. And she, thought asthat might be an interesting idea, was a littlenervous about Cambridge but—

 

Kossow: Had she spent much, or had you spent much time in England?

Thacker: No as a matter of fact, I had but she hadn’t. And so one day in the earlyWinter I think of ‘96, I heard a rumor that Roger’s lab was actually going tocome to pass, they’d continued to talk after I dropped out of that discussion and oneday I got a call on the phone, I know, what happened was I heard about this andI sent Roger an email at the computer lab congratulating him and asking himwhat was going on? Roger is always very crypticin his email responses and he sent back something that said, ”Well, yes, itlooks fairly positive and you’ll hear more about this soon.” Thirty minuteslater the telephone rang and it was Nathan Myhrvold and hesaid, and he said, “You know we’ve been talking about this and one of thethings that you know a lot about how to do is to set up a research lab, howwould you like to move to Cambridge for two years andhelp Roger set up the lab, because he’s never actually done that and you’vedone two?” So I said “Well, that sounds pretty good, let me talk to my wife andI’ll call you back in a few minutes.” And I called mywife and I said, “Love, how would you like to spend a couple of years and livein Cambridge and I will start, I will help Roger start a lab for Microsoft?”No, I didn’t say that, and help start a lab there, andshe said “No, I don’t like Boston. It’s too cold.” And I said, “No, the otherone.” And she said, “Oh, well, that might be interesting.” And so I called backNathan, I called Nathan back and I said “Sure, I’ll doit.” And that was how the whole thing, my participation in the whole thing gotstarted and it was a two year assignment, so I went over there and helped setit up, help do a lot of the initial hiring and set up theyou know, policies you need. Industrial labs are a little different than academicplaces and so you know, you have to work with things like the Human ResourcesDepartment and all of that. And Lord, with thePublic Relations people and Press people because of course, Microsoft going toCambridge was a big deal in the British mind, they thought. They actuallythought that we were going to drive the house pricesin Cambridge through the roof. Now they have subsequently gone through the roofbut it’s not our doing. It’s just that Cambridge has flourished, as a center of technology and innovation. But Iremember talking to a reporter from Cambridge Evening News and he said, “Wellyou guys are going to drive up the prices because all you Americans,” and Isaid “Wait a minute, this is not an American lab,it’s a University, this is an European lab, I’m the only American here and I’mgone in two years.” So, right, we were also dunned when McDonalds set up aseparate, a second facility in Cambridge, thoughtit was for us. And I said “No, we eat at the pubs just like you do.” But in anycase, that was a very pleasant and I think quite productive time. I didn’tactually do a whole lot of technical work during that time. ButI did some. I worked a bit on electronic books, how to organize an electronicbook and how to build one. And that was where some of the early thinking forour Tablet PC was done. So when my wife and I movedback at the end of the two years, we had decided that we really didn’t want tolive in Redmond because some people are affected by the disparity between thedays and the nights and the various seasons andit bothered my wife quite badly. So, I said to Rick Rashid, “Well, I’m notgoing to move to Redmond, am I still working for Microsoft?” And he said “Oh,sure, you know, we’ll find something to do.”

 

Kossow: But by that time Gordon [Bell] is working down here isn’t he?

Thacker: By that time, Gordon may have been here but he was not, I don’t know whatthe deal was. Jim had his thing in San Francisco at that time, but Jim reallydidn’t want to grow his, his organization that was just an office for him, because helived in San Francisco. When Gordon moved to San Francisco he started workingfor Jim. So, yeah that is about the same time. But instead what I did was Italk to the electronic book people and I had beenworking with them a bit on some of their prototyping work and when the guy whowas running the group got the charter to build a Tablet PC, I went to him and Isaid “Why don’t you hire me because I knowhow to do this, because we built a Tablet PC at SRC in 1993.” And so, MarkHayter had done that. Mark Hayter and Satterthwaite. And, yeah, and then, andAndrew Birrell as well. Andrew Birrell had bidit, he was in that photograph that I mentioned. So, so I got embroiled in theTablet PC for about two years.

 

Kossow: Okay, and that’s the time when you and Butler gave the talk on the Alto for the museum?

Thacker: That is correct. Yeah.

 

Kossow: Yeah, I remember you were working on that at the time.

Thacker: Right, and that became sort of a main line thing in the industry actually.I mean Tablets are still, the Tablet PC has not been the kind of commercialsuccess that the laptop has but it’s gratifying that it’s the only such machine in that ilkthat has been in the market for six years now. And it still survived, I mean,all the earlier ones died. So more recently let’s see, what have I been doing?So I went back to Microsoft research when I finished allthat up, I worked on a couple of other little projects in an incubation grouprun by Craig Mundie, who’s now the CTO, CTO then for that matter. But whenthere began to be a discussion within MSR about actuallydoing computer architecture research, I decided to rejoin the organizationbecause of course, that’s what I do. Or what I did for a long time. And I’vebeen there now for almost, a little more than a yearand a half, yeah. And so we’re setting up a group, we’re doing some things;we’re doing the thing that I mentioned to you.  

Kossow: Right, with Patterson.

Thacker: With Patterson.

 

Kossow: So, are you going to retire next year?

Thacker: Probably not, I mean, I find myself still enjoying going to work and Ifigure if you can that why retire?

 

Kossow: Are you working over here? Or do you have a—

Thacker: Yes, I’m two blocks away from your Museum. And in fact I’ll be a littlecloser because shortly we will move into this building on Pear Avenue becausethe Mountain View campus is filling up and we’re growing.

 

Kossow: Were you involved in the Xbox at all or any of the other--

Thacker: Yeah, I actually was a bit, I helped.

 

Kossow: Dave Conroy had worked on that a little bit.

Thacker: Dave actually worked on hardware for the group that became the Xbox group.But by the time the Xbox started, he was no longer in that group, he wasworking with me. And then he decided to go off to Apple, so he’s doing that andapparently happy as a clam.

 

Kossow: Yeah, my office was two offices down from him.

Thacker: Well, Dave does love to build things and of course he had--

 

Kossow: Dave is very good. He’s extremely good.

Thacker: Dave is one of the few people I know that’s actually a better engineerthan I am. Because he’s careful, he’s also extremely smart, we havecomplimentary skills in many ways, but there are things he can do that I, you know, can’t. And I’msure the same thing is true about the other direction.

 

Kossow: He’s very smart, he’s very fast.

Thacker: That’s right. Now I enjoyed working with Dave. As I said, I hired him atSRC, we worked together for almost 15 years and, so in any case, I’ve just beendoing this and I probably will not retire, I mean, I might. I threaten to from timeto time but if I did I would go off and find something else sort of to do inthe technology area, I don’t know what it would be but.

 

Kossow: Yeah, it’s hard to do systems things now.

Thacker: Well, that’s one of the reasons that I’m having fun these days because Ibelieve this RAMP project actually has a hope of revitalizing systemarchitecture research because if you look at the papers that’s been published over the last 15or 20 years, they’re all very incremental in the architecturearea and theredon’t seem to be many breakthroughs and the reason is, is it’s just very hardto build systems that are very different from what’s in themain line because you can’t build chips any more. And so the, one of the keyideas behind the Ramp program is that FPGAs are now big enough that you can’tbuild something that is both as good, is asgood as something you can implement in silicon. But you can actually build acredible emulation of a big system that runs at a sufficiently high speed thatit will run real software. And therefore, you can do aclass of experimentation that you just simply couldn’t do without machines likethat. And, so I’m quite enthusiastic about it as I mentioned earlier, I willbe, I’m working with them on getting their next machinegoing and—

 

Kossow: So how many software people do you have working with you?

Thacker: Oh, the lab is mostly software people except for me, I mean, we’re tryingto hire more architecture people, we have hired one very good person from Stanfordand another sort of coconspirator who came to Microsoft after getting aPhD, after being essentially the CTO of SGI. And so the three of us are tryingto attract more architecture people but that’s a slow process because our standardsare very high and, so what else can I tell you?

 

Kossow: Oh well, there are a few questions that they wanted me to ask.

Thacker: Okay, sure.

 

Kossow: While we grind through those. What were your most important life lessons?

Thacker: Life lessons? I would say it’s always a good idea to follow Einstein’sdictum and make thing as simple as possible. But no simpler.

 

Kossow: And what was your proudest moment?

Thacker: Oh, I don’t know. I could think of a number of things, you know, birth ofmy first child, although I didn’t have nearly as much to do with that as mywife did. You know, any of the nice awards, you know, but—

 

Kossow: So what might you have done?

Thacker: What might I have done? If I hadn’t chosen the path that I did? I wouldhave probably have been a retired fairly mediocre Physicist.

 

Kossow: So, lookingback, what do you see as you’re most important contribution?

Thacker: I have tried to figure that out and I can’t really do it because therehave been several different things that I’ve actually been quite proud of. I’vebeen extremely proud of the Alto of course. Because it basically launched a whole line of,line of stuff that we are still developing today. And I’m quite proud of someof the larger machines just because they demonstrate that if you do careful engineering,and keep it simple, you can build things thatare very high performance. I’m proud of the Tablet PC because I like eventuallyit will be, it will take over and all PC’s will be tablet-like. That might bevery slow, it will require a lot better power consumption, and alot of sort of attention to the low end in power which of course is now one ofthe problems that is facing us as an industry. We’re not going to be able to gofaster because we cannot get the power out of thechips anymore. And so we’re going to have to figure out ways to deal with thatand I think some of those things are pretty exciting because it’s going torequire looking at systems in slightly different ways thanwe’ve been doing for the last 20-30 years, is to say we’ve been quiteincremental. So now we have to do a lot more thinking out of the box.

 

Kossow: So what do you think of the convergent devices or you know, mobile phonesand that kind of thing?

Thacker: I think they’d be nice if they worked. I haven’t seen any that I reallylike. Usually people who try to build fancy phones forget that a killerapplication for a telephone is making phone calls and also, it’s

very hard to have a sort of convergentdevice world when the market is controlled so heavily by the carriers. This issomething people frequently don’t realize, which is that you can’t walk up toanybody’s cellular network with a different phone and necessarily run it. Sothey control very carefully what goes on their network and that while that ishappening it’s a some what different thing than happened in other sort ofcomputer related industries. I’m not quite sure how it will play out but we’llsee.

 

Kossow: What advise do you have forcurrent and future generations?

Thacker: Well, that’s an interesting question. I’m very worried about the fact thatfewer and fewer kids are being, are interested in technology. And I’ve tried tofigure out ways to help with that but I don’t know, I don’t know exactly what to do. I meanwe don’t have the very, we almost don’t have you know, the ham radio operatorsand the radio experimenters that happened when I grew up. You can no longer buy Heathkits. So what is going to be themotivator for young people? And of course for awhile, people thought it wouldbe the PC. But once you’ve written a few Basic programs, what then? You knowand you’ve manipulated the sprites on thescreen and written your little video game, what then? Well, one possible answeris Robotics. Allowing a kid to build something that actually does something isa pretty interesting thing. And I know Lego hasbeen relatively successful about, with that. We’re actually making a roboticstool kit to try to make some of that simpler, but I don’ know, we need a sortof a killer app. for sucking people into technology and ofcourse when I grew up, when I was in, Sputnik happened when I was in highschool and of course that was a galvanizing moment for the nation. And so a lotof people who were deciding on what to do made adecision to you know, to basically go out there and beat the Russians. Wehaven’t had anything like that. So,--

 

Kossow: Well, there have been galvanizing things that have happened but theyhaven’t been positive.

Thacker: Well, but they haven’t been, it’s not the same kind of thing. It’s notsomething that you would actually aspire to.

 

Kossow: Well, this has been fun.

Thacker: Yeah, thank you, my pleasure.

( The End)