HARD.MSG[MSG,LES] - www.SailDart.org

PDheader:1975-09-25 13:09:00-07:00 bdefa6ea821a3634e4c40b36d2f0980a ∂25-SEP-75 1309 BPM,BPM Display Buy

Price info on Datamedia Elite 2500A, UDS 1200/150 baud modem, etc. is contained in page 2 of TERMIN.RPT[PUN,BPM].

PDheader:1975-10-30 16:39:00-08:00 5012ccb4946d902d92cd6e479abc6765 ∂30-OCT-75 1639 BPM,BPM Datamedias To: TED, TAG, LES

Datamedia board exchange policy works like this. Air mail a bad board to them and at the same time call them and tell them which board. They immediately airmail a good board out. Then they fix or replace the bad board (at their option) for $50. (All 6 boards cost the same.) SRI receives a good board within 36 hours of the time they notify Datamedia of their need.

PDheader:1975-12-05 11:03:00-08:00 26ef1f1e9a0215a16a1f59093087fb11 ∂05-DEC-75 1103 REG Ampex disk capacity

6.5*10↑9 bits 6.5*10↑9 bits, as formatted

PDheader:1975-12-10 15:07:00-08:00 0fc33618cf8d694eeb0f3050eb1e3de3 ∂10-DEC-75 1507 LES Arm Controller Development To: BES, TAG, TOB, VDS CC: JMC, JBR, REG, TED

This is my understanding of the conclusions we reached in a review of arm controller development needs this morning. We need two copies of a new controller. It should be modular, with (at least) 3 kinds of card, each separately interfaced to a Unibus. This will permit the controller to run either under the PDP-11/45 or an LSI-11 (not both at the same time). The cards that we have identified are as follows. MOTOR CONTROL (possibly 2 cards each) - 7 needed for each controller. Accepts a digital command and mode (velocity or force servo) and produces a motor control signal, based on speed and torque sensors (the latter probably analog). It will not provide position sensing except through digital commands. POSITION SENSING - 7 needed. Accepts either analog pot signal or optical encoder output and continually produces digital position. This position, as seen on the bus, will be no more than 50 microseconds old. MULTIPLEXED A/D - 1 per controller. Accepts multiplexed analog signals from strain gauges, etc., and produces digital value on the bus. The next step is that Gafford will produce a fairly detailed specification by January 15. Scheinman will examine the tradeoffs on optical encoders vs. pots for the various joints and will make recommendations. After review, the design and construction will be undertaken, with projected completion by June. Who will do the construction is currently an open question.

PDheader:1975-12-11 08:37:00-08:00 feaafbc7ada2cb884c8562855d732e8f ∂11-DEC-75 0837 BES NEW ARM INTERFACE

In addition to the 7 position sensing cards, we will need similiar modules to continuously read the arm tachometers. These velocity readings are necessary when we are torque servoing. When we use the hardware velocity servo, it would be good if we could also read the commanded motor current. However, this third feedback signal is optional and could even be multiplexed to reduce the number of modules necessary.

PDheader:1975-12-16 01:03:00-08:00 a8cf2941ed7d3663d7d11f4e054ded31 ∂16-DEC-75 0103 TAG at TTY10 0103 PDP-11 Disk Channel

PHIL PETIT MAKES A CHANNEL THAT LETS A PDP11 TALK TO AN IBM-STYLE STORAGE CONTROLLER FOR LESS THAN $10K. THE ONLY DEVICE I CAN GUARANTEE THE EXISTENCE OF IS AN INTERFACE DIRECTLY TO THE DISK SPINDLE, WHICH IS NOT REALLY WHAT WE WANT IF WE WANT GENERAL PURPOSE ACCESS TO THE DISK FROM THE CONSOLE MACHINE. THE ONLY OTHER 11 CHANNEL THAT I KNOW OF IS SYSTEM CONCEPTS' SA11, WHICH COSTS GROSSLY TOO MUCH MONEY ($30K OR SO). I'VE TALKED TO REG SINCE I PROPOSED THE DIRECT DISK INTERFACE TO YOU, AND HE SAYS THE 11 SHOULD REALLY TALK TO THE CHANNEL. THE MICROCOMPUTER IN THE SPINDLE INTERFACE THAT I WAS TALKING ABOUT HAS PLENTY OF PROCESSING POWER TO HANDLE IBM INTERFACE PROTOCOL, BUT IT WOULD REQUIRE DEVELOPMENT. THE COMPANY WHICH MAKES THIS MICROCOMPUTER CONTROLLER WOULD BE WILLING TO MAKE SAIL A VERY GOOD DEAL IN RETURN FOR USE OF SAIL'S 11 AND STORAGE CONTROL TO DEVELOP THE PRODUCT, BUT IF YOU WANT A PRODUCT WHICH ALREADY WORKS AND COSTS LESS THAN $10K, PMP'S DEVICE IS THE ONLY ONE AROUND.

PDheader:1976-01-29 12:08:00-08:00 9647dfea720556845931ec6cee1f4522 ∂29-JAN-76 1208 LES IMLAC & air conditioning To: TED CC: JMC, REG, TAG, HVA

The Dendral IMLAC is now ours. It is located in a temporary building back of Serra House and suffers from a flakey memory. To pick it up, contact one of the following at ext. 74878: Bill White, Bob Engelmore, or Ray Carhart. As you know, Plant Services is concerned about the additional heat load of the KL10 and are asserting that we need more air conditioning capacity. Of course, they were not competent to engineer the initial installation and I have seen nothing recently that gives me greater faith in their judgement. I propose that we undertake a small study and perhaps seek outside consultation. I suggest that you or someone add up the expected heat dissipation of the system and compare it with the sum of the green monsters' capacity and the "normal" heat load of the computer room. To estimate the latter, consider the fact that the computer room was originally intended to be an auditorium. Estimate the seating capacity and multipy by 100 watts/person. If we actually do need more capacity, I think that we can argue that it is available from the central air conditioning system, since it is not being called upon to cool the entire building, as originally designed (i.e. the full circle). In any case, we need to know what the expected load is.

PDheader:1976-01-29 12:47:00-08:00 fd1c293b1760609b0b1e41a97a9d6932 ∂29-JAN-76 1247 JMC To: LES, TED, REG, HVA

I suggest that we simply stall until the KL-10 is installed and see how hot the room gets and extrapolate to summer conditions.

PDheader:1976-02-24 15:03:00-08:00 23db36d03e00ba1841f903259904c3af ∂24-FEB-76 1503 REG KL10, PDP-6

Jim Stafford called you, but talked to me. 1. He's waiting for Dick Devlin in Marlboro to send him a TWX stating a delivery date for the KL10. Stafford called Devlin who said that he'll send the TWX tomorrow. 2. Devlin wants a letter from you stating that DEC wil never hear about the Stanford PDP-6 again. Send this letter to Stafford.

PDheader:1976-09-07 15:04:00-07:00 96cb29c00fd12a9590690ba586f8edd6 ∂07-SEP-76 1504 BPM

PDheader:1976-09-07 15:04:01-07:00 6acd8f3bf833db37f093bc46c83fbc75 ∂07-SEP-76 1504 BPM UDS modems Here is what we have to ask permission to buy:

Modem Purchase Order Information (second order) To: Mr. Ken Larsen Inter-Link Systems 10601 South Sunnyvale-Saratoga Road, Suite 212 Cupertino, California 95014 Copy to: Universal Data Systems 2611 Leeman Ferry Road Huntsville, Alabama 35805 UDS Stanford 1200/150 auto-answer modem (must conform to attached "UDS-202C Data Modem Specifications", with the following exceptions: updated as per changes discussed by Al Wegeman of UDS and Tom Gafford of Stanford; sized to mount in UDS RM-16 multiple modem rack; must be compatible with Bell CBS Data Access Arrangement; must be compatible with Prentice P 1200/150 originate-only modem; when talking to either Prentice P 1200/150 originate-only modem or UDS Stanford 1200/150 originate-only modem, error rate must not exceed that of two Prentice P 1200/150 modems in equivalent situation) automatic answer-only PC card $ 275 ----- ----- TOTAL per unit $ 275 x 2 units ---------- ---------- TOTAL $ 550 All equipment above shall operate as specified within a temperature range of +55 degrees to +100 degrees F. UDS shall supply three sets of documentation on all items listed above sufficient for maintenance and interfacing with other devices (e.g., a display terminal).

PDheader:1976-11-19 04:49:00-08:00 dcb07961d0aaf4119466b31bab59e77a ∂19-Nov-76 0449 LES Random errors To: TED, JBR

I have been getting quite a number of unrepeatable errors this evening, both ILL MEM REFs and "PC exceeds mem bounds". E.g. one Pub run took four tries before it completed successfully.

PDheader:1976-12-01 23:06:00-08:00 4e235e8596e4a26ea6d7f460aa659d23 ∂01-Dec-76 2306 DPB To: DM.DIS[1,DPB]:;

There is some confusion about our new DataMedia terminals, modems, pc boards, etc. The following table summarizes what I know/believe: NO. Purchaser Notes 1 MYCIN Will be wired to SUMEX. No modem. 1 AI-Lab Modem ordered. LES Earnest will install modem. 8 CSD Modems ordered. Nick Veizades will install modems. 2 MOLGEN Terminals, modems have arrived. Nick is installing or has installed. 2 Feigenbaum Terminals are now running, connected to DSL's PDP-11. Carolyn Taynai has ordered modems. Nick will install. 1 Feigenbaum Terminal not yet ordered. Purchase order for terminal and modem is in process. (Carolyn Taynai and Betty Scott.) Will need installation by Nick. Please contact Denny Brown (dpb@sail or dbrown@sumex) if there are any problems in the above summary. -Denny

PDheader:1976-12-21 18:30:00-08:00 87483f876107c7e15918fdb850f27c5c ∂21-Dec-76 1830 BPM Using Datamedias via AMES-TIP To: DM.DIS[PSI,BPM]:;

The procedure to follow is given in DM.ME[UP,DOC]. Other information about DM hardware and software is there also. You should also do HELP DMKEY to get a crib sheet for your keyboard if you haven't already.

PDheader:1977-01-06 15:39:00-08:00 b0441cbb37a6bdb2ea0e6ca1afe33c97 ∂06-Jan-77 1539 LES Home phone lines To: DRB, TOB, JJ, CCG, EK, DCL, BPM, DCO, JP, ALS CC: JMC, HVA

Because of recent telephone tariff changes and the resultant costly charges for connect time on business lines, we have told the telephone company to disconnect all the business lines in peoples homes (including yours). Disconnect should happen within about 10 days. Assuming that you wish to continue using a home terminal, you may either use your regular home phone or have another personal line installed. If you use your regular home phone, the lab will cover half the cost of 60 unit service. If you have an additional line reconnected, the lab will cover the cost of installation and the full 60 unit service. Any additional charges other than toll charges for lab business will be your own. In either case, it will be necessary to bring the phone bills in to be reimbursed out of petty cash. Since the legality of this move may be open to question, please try to avoid provoking Ma Bell. In particular, before the telephone remover or installer appears, please disconnect the computer terminal and any jacks that have been installed. It might be better still to put the terminal somewhere else temporarily. Since you can now order a new phone with a jack at no additional cost, I suggest that you request it that way since it makes the terminal connection easier. I regret having to complicate your life this way, but the current policies on business lines are discriminatory and we can't afford them. The question of whether or not these lines should in fact be business lines is arguable (e.g. is a graduate student doing research in business?), but I would rather avoid the argument if possible. Les

PDheader:1977-01-17 12:56:00-08:00 67aaa9cf53469be6e12a9e51331f4f9e ∂17-Jan-77 1256 DPB To: DM.DIS[1,DPB]:;

Ten DataMedias have arrived. They are now in the Polya storage room. Modems have not arrived. Latest inquiry says 1-2 weeks. The ten terminals are allocated: 8 to CSD, 1 to Randy Davis, 1 to Les Earnest. -Denny

PDheader:1977-01-18 04:01:00-08:00 63402407de05b217cf906085c7bdc0eb ∂18-Jan-77 0401 BH via SUMX INTECOLOR 8001 TERMINALS To: TTY.DIS[1,BH]:;

I HAVE BEEN RESEARCHING THESE TERMINALS AS A POSSIBLE ALTERNATIVE TO DATAMEDIAS FOR IRCAM, AND THEY MAY ALSO BE OF INTEREST FOR ANY FUTURE SAIL HOME TERMINAL PURCHASES. THEY OFFER ESSENTIALLY THE SAME FEATURES AS DATAMEDIAS FOR ESSENTIALLY THE SAME PRICE, BUT WITH COLOR DISPLAY AND 48 TEXT LINES. THEY WILL SUPPORT THE FULL SAIL CHARACTER SET. ALSO, FOR ABOUT $500 MORE, THEY WILL PROVIDE GRAPHICS (160 X 192 DOTS) AND A LIGHT PEN. ANOTHER $1000 WILL BUY YOU 256 DIFFERENT CHARACTERS, GOOD FOR APL FOR INSTANCE. THE MAIN THING IS THAT SINCE THEY ARE BASED ON THE 8080 MICROPROCESSOR, IT IS IN PRINCIPLE POSSIBLE TO HAVE THEM USE A GENUINE MICROSWITCH SAIL KEYBOARD AND GENERATE THE TWO-BYTE DATAMEDIA SEQUENCES FOR META AND SO ON. I AM PRESENTLY CORRESPONDING WITH MR. CHARLES MUENCH AT INTELLIGENT SYSTEMS CORP (4376 RIDGEGATE DR, DULUTH, GA 30136) ABOUT THE HAIRY KEYBOARD REQUIREMENTS AND AM AWAITING A PRECISE PRICE QUOTATION. IF ANY OF YOU ARE INTERESTED YOU CAN GET IN TOUCH WITH HIM TOO. (BY THE WAY, FEEL FREE TO ADD NAMES OR REMOVE YOUR OWN TO/FROM TTY.DIS[1,BH] IF YOU ARE OR AREN'T INTERESTED IN THIS.) FOR OUR OWN PURPOSES, IF ANY OF YOU HAS ACCESS TO THE BIT DEFINITIONS FOR THE DATA DISC VERSION OF THE SAIL CHAR SET, I'D APPRECIATE A LISTING THEREOF. TNX.

PDheader:1977-02-09 04:41:00-08:00 9aed036a41324925404de5f9f3abcdc4 ∂09-Feb-77 0441 BH via SUMX MICROSWITCH KEYBOARDS

PMF SAYS THAT YOU SAID THAT THERE IS SOME DIFFICULTY ABOUT GETTING SAIL KEYBOARDS FROM MICROSWITCH. COULD YOU ELABORATE? ALSO, DO YOU HAVE A NAME OF SOMEBODY TO CONTACT AT MICROSWITCH? THE LATEST WORD FROM INTECOLOR IS THAT THEY THINK IT'D BE EASY TO REPROGRAM THEM TO USE THE SAIL KBD BUT THEY DON'T WANT TO DO IT THEMSELVES, SO WE MAY USE ONCE AGAIN THE TECHNIQUE OF STEALING MANPOWER FROM STANFORD. (DPYSER, BY THE WAY, IS MOVING RIGHT ALONG... IT ALMOST ASSEMBLES, AND WE ARE PREDICTING A TRIAL RUN THIS WEEKEND. IT WON'T WORK, OF COURSE.)

PDheader:1977-03-05 13:11:00-08:00 f3c4d1ff9adbaec9237d1a901d8d8def ∂05-Mar-77 1311 LES Librascope Auction Results To: DBA

$151 to David Shaw, $150 to Geoff Goodfellow (SRI), $130 to Vic Scheinman, and $129 to L. Earnest. They are all fools who lose control at auctions. Les

PDheader:1977-03-11 12:33:00-08:00 5f0cecad79662e3de6c5ae8838cacaf3 ∂11-Mar-77 1233 DPB dm's

Les, We now have 6 dm's with modems installed. None work exactly right yet. Dale Pearson at SUMEX is having some trouble with the SAIL mods. Four of the six are almost right, and in fact have a disjoint set of problems. We should have things ironed out on Monday (I hope.)

PDheader:1977-05-19 22:30:00-07:00 b8c2375d6887d47d55c6062a7f4d6811 ∂19-May-77 2230 100 : TODD GLASSEY CLOCK

HELLO CLOCK IDEAS TO FOLLOW 1 WWVB RESET EVERY 5 TO 10 MIN. 2 DEVICE GETS TIED TO THE KL10'S I/O BUS 3 DEVICE GETS A 6(SIX) DIGIT LED DISPLAY (HH:MM:SS) 3 DUAL POWER SUPPLIES FOR POWEROUTS (BATTERISE ARE TRICKLED CHARGED) 5 BEST OF ALL DEVICE GETS BUILT ON A FLEXABLE PC BOARD (THESE ARE CHEAP AND RELIEABLE AND THE LAYOUT CAN BE DONE ON THE XGP WHICH MAKES IT EVEN CHEAPER) 6 FULL DOCUMENTATION INCLUDING SCHEMATICS WILL BE DRAWN UP MOST OF THE SCHEMATICS ARE ALREADY DRAWN UP , SO ILL TRY TO DROP BY AND SEE WHATYOU THINK SOMETIME NEXT WEEK. THE REASON FOR THE FLAλEXABLE PC BOARD IS SO YOU CAN MAKE THE WHOLE THING LOOK LIKE A WRIST WATCH WITH 3 OR SO WIRES AND A PIECE OF RIBBON CABLE HANGING OFF OF IT. TODD Clock In response to your message of 19 May: 1. I see no point in displaying the time-- no one is going to look at it. 2. Hours, minutes, and seconds are not enough. We have to know the date and fractions of a second (on the order of a millisecond). 3. Given that this device will survive power failures, there seems to be little point in the WWVB tie-in. It would be much simple to have an ordinary binary counter driven by a crystal oscillator. Conversion to date, hours, etc., would all be in system software. The only control function would be a RESET, which sets the clock to 0. This should be a contact that gets grounded. 4. Since this is to be a one-of-a-kind device, it would be silly to put it on a PC card. Wire wrap is clearly the right thing. 5. Please review the design with TED before proceeding.

PDheader:1977-05-20 12:21:00-07:00 008e071ddeda9f48416873201a70413c ∂20-May-77 1221 TOB head-per-track discs

Les I talked with the Sales Manager of Amcomp. The price for a 38 Mbit disk is $10k. The price for a PDP-11 interface and controller is $4.5k. He said that he is interested in the PDP-10 market and would be interested in a short description of the hardware to which it would be interfaced. Tom

PDheader:1977-05-27 14:13:00-07:00 3ee555f76fd4a1fe3864fab02ca36d13 ∂27-May-77 1413 PAM Cheapy Terminals

I found out recently from the folks at LOTS that they are looking at a new terminal by Elmer Perkins called the Fox 1100. It has a microprocessor in it, displays 24 lines of 80 characters, each 9x12 bits, and claims to have full cursor control. It cost $925 in qualities of 25, and looks to me like it might be made to look like a Data Media, thus avoiding the need to program the system and ETV to deal with yet another terminal type. If you can forsee the possibility of the Lab buying any of these turkeys, let me know. I'd be interested in investigating the effort necessary to make them look like DMs, and perhaps doing the programming in the terminal to make it compatible. My ulterior motive is that I'd like to have a home terminal that can run E, and I think DMs are overpriced. Let me know; the basic question is whether the lab has a chance to buy any terminals; if so, we could presumably get 2 1/2 times as many for the same bucks. Paul $

PDheader:1977-05-27 15:56:00-07:00 d8550f1a0204a9ebd12a4646d9fbe315 ∂27-May-77 1556 LES MGW -- U.S.ROBOTICS ANSWERING MODEMS FOR $105 EACH !!

Date: 2 MAY 1977 1538-EDT From: MITCHELL WOLRICH ( MGW@MIT-AI) Sent-by: MGW at MIT-ML Subject: CHEAP MODEMS!!! To: PBARAN at USC-ISI, REM at MIT-MC THIS IS TO INFORM ANYONE WHO MAY BE UNAWARE OF THE FACT THAT THIS MONTH AN INEXPENSIVE AUTO-ANSWER MODEM WAS INTRODUCED. THE MODEM CAN RUN AT SPEED FROM 0-600 BAUD. IT IS EQUIPED WITH AN RS-232 AND A 20MA TTY INTERFACE. IT IS ALMOST COMPLETLY COMPUTER CONTROLLED (I.E. THERE ARE ONLY THREE CONTROLS ON IT, (1) POWER ON/OFF (2) ANSWER/ORIGINATE AND (3) A CARRIER DETECT LED. IT IS PRICED AT $105.00 AND IS AVAILABLE FOR IMMEDIATE DELIVERY. A 1-YEAR SERVICE CONTRACT (FOR THE NON-HARDWARE HACK) IS AVAILABLE FOR $25. THE ADDRESS OF THE MANUFACTURE IS LISTED BELOW: U.S. ROBOTICS P.O. BOX 5502 CHICAGO, ILLINOIS 60680 (312) 528-9045 YOU CAN ALSO WRITE TO ME AND I WILL TRY TO ANSWER ANY QUESTIONS YOU MAY HAVE MY NET ADDRES IS: MGW@MIT-AI OR AT HOME: MITCHELL WOLRICH 8 BRUCE CIRCLE RANDOLPH, MA 02368 OR SEE THE ADD ON PAGE 159, APRILL 1977 BYTE MAGAZINE. ∨

PDheader:1977-05-28 19:02:00-07:00 87f89368642316e9b6219e1e40e92919 ∂28-May-77 1902 BPM Jarvis terminal To: PMF CC: LES

Pentti Kanerva is interested in final specs for it. IMSSS may want to go in with SAIL and CSD to produce a bunch of them. You should call him and discuss this.

PDheader:1977-06-02 14:19:00-07:00 b9951c43a126a359456d7678fbd4130b ∂02-Jun-77 1419 LES Computer clock To: tk at MIT-AI CC: TED

Rumor has it that your machine has a clock that servos on power line frequency but keeps on ticking through power failures, floods, and tidal waves. Since we would like to build one with similar properties, it would be nice to have a look at any relevent documentation that you have. Is there some?

PDheader:1977-06-02 15:36:00-07:00 c79d0a0c05199c87ff9ace06366515ba ∂02-Jun-77 1536 FTP:TK at MIT-AI (Tom Knight ) Computer clock Date: 2 JUN 1977 1835-EDT From: TK at MIT-AI (Tom Knight ) Subject: Computer clock To: LES at SU-AI CC: TED at MIT-AI

Well, Les, that's a good rumor, but unfortunately not true. You see, it gets wet during the tidal waves. Actually, all we have is a 36 bit counter driven from an independant power supply which never gets turned off, even if computer power does, but it is certainly not immune to power failure. We keep the offset between real time and 36 bit time on the disk, and that number gets checked for "reasonableness" when the system is brought up. The result is that we have to reset the offset when we have power failures, but that is no great burden, at least in Cambridge. These days, of course, a CMOS counter with battery backup is obviously the right thing to build, with a quartz crystal driving it. Likely that is better than line frequency. If you wanted to be hairy, I suppose we could receive the 60 khz WWVA phase modulated time signals....

PDheader:1977-06-16 22:32:00-07:00 0d954e2cfbdd4f1d97ef678e47dc0166 ∂16-Jun-77 2232 HPM Heathkit TV repairs To: TED, LES

It would be easiest for me to proceed at my dizzying pace if the following replacement parts, most of which the set will not need, were bought. Description Heathkit part # 1970 price zener diode 56-79 1.60 known bad (special HV) vertical transf. 51-158 4.40 transistor 417-233 .50 2N3643 transistor 417-126 .45 2N3565 transistor 417-234 .60 2N3638A transistor 417-239 5.00 MJ3029 _____ 12.55 since being awake at the right time and driving are my weakest skills, I would appreciate if somebody else were delegated to do it. Unused parts will be incorporated into the cart's electronics.

PDheader:1977-06-28 20:48:00-07:00 d4bbedc9bb9d34d3fb63e3c40bf1ff30 ∂28-Jun-77 2048 FTP:PEPIN at USC-ECL KL 10 Date: 28 JUN 1977 2048-PDT From: PEPIN at USC-ECL Subject: KL 10 To: les at SU-AI

Tom ellis at isi told me to send you a note asking about the speed yove seen out of your kl 10 vrs the ka. Here in engineering at usc we are getting a kl to replace a ki ten. We have talked with ellis some and he says you have seen some disturbing speed stats for the kl vers your old ka. Can you give me some ideas as to what he was refering to. jim pepin mgr ecl -------

PDheader:1977-06-29 18:20:00-07:00 994cd7cdabf050a091d83b6efa00c1bb ∂29-Jun-77 1820 LES KL10 vs. KA10 To: Pepin at USC-ECL CC: Ellis at USC-ISI

In terms of CPU time for user programs, the KL10 is about 5 times as fast as the KA10 (up to 10 times as fast for programs that use a lot of byte operations). Our total thoughput is not up that much because we became heavily disk bound, having lost our swapping store at about the time the KL10 arrived. We have ordered a second channel and disk controller so as to double the disk transfer capacity and balance the system. It should be going by late summer.

PDheader:1977-06-30 10:57:00-07:00 63f7b1a595bb45b44fe23fc92a719b2c ∂30-Jun-77 1057 MJL terminals

hi Les, we are taking an inventory of DataMedia terminals recently purchased. Would you plz send me the serial #'s on the DataMedia's and their keyboards that Denny recently sent up to you? The terminals and the keyboards each have their own serial #. Thanks alot - this will be a big help. Moira

PDheader:1977-06-30 12:21:00-07:00 d8e487461578f34c634ab9a6efd02b29 ∂30-Jun-77 1221 LES To: HVA

PDheader:1977-07-15 15:19:00-07:00 d123df28f8243f5c1d7aa64c9bc4e5b7 ∂15-Jul-77 1519 FTP:REM at MIT-AI (Robert Elton Maas) Holofile Date: 15 JUL 1977 1817-EDT From: REM at MIT-AI (Robert Elton Maas) To: LES at SU-AI, JMC at SU-AI, TVR at SU-AI, PBARAN at USC-ISI CC: REM at SU-AI

.MAIL LES,JMC,PBARAN%ISI,TVR Subject: Holofile Type message followed by <CONTROL>Z I have written to Holofile Industries Ltd., distributor for the TRW holographic 4" by 6" 200-megabit storage medium, and received a prompt reply saying that there is no more printed material available at this time but answering most of my questions. The basic reader, described in the advertising brochure, costs $5000 (not $500 as in the brochure). There is no comment on that, so I don't know whether the brochure or the letter has a typographic error, although surely one or the other must. If the cost is $5000, it is too expensive for hobbyists, but reasonable for SU-AI or IMSSS to replace one disk drive or whatever 200 megabits amounts to. The writer is about $250,000 but the company will offer the service of converting magtape to holofile for a very low charge, available in key cities. [SU-AI MAIL PROGRAM CRASHED HERE, REMAILING FROM MIT-AI] Delivery expected first half of 1978. Anyone wishing to read full text of the letter, from James A. Case, Vice President (of course, like most companies, probably at least 10 v.p.'s, maybe) should contact me (REM). Major application I envision is storage of system software and other files that aren't actively being changed but which are being used (read) online. All files older than a certain date, or 200 megabits thereof, could be copied to magtape and written on holofile, with extra copies at $.07 each appx., then disk service could be modified to point from directory to places on this device, and after verification of correct data the disk version could be purged. Whenever the file gets edited or otherwise rewritten, the pointer to the holofile could be scrapped and a copy made on the disk. Of course a study for cost-effectiveness would have to be done first.

PDheader:1977-07-26 17:07:00-07:00 34a3719bbafada8daa597d2b9e122bb0 ∂26-Jul-77 1707 JBR Ampex CU

Lacks "Theory of Operation" and "Maintenance Drawings". The main problem we have been having is parity errors in the microstore. The CU reports check 1 errors of the form "early 10".

PDheader:1977-10-14 00:11:00-07:00 bf5078629932aa1812307d9f9e0162cc ∂14-Oct-77 0011 JMC Display system To: LLW, LCW, TM CC: JBR, LES

I have heard that both Intel and Zilog will be coming out early next year with PDP-11 speed, 16 bit word, micro-processors with 32 bit addresses. Making the display system with a PDP-11 was always a big strain, because with a 512 by 512 array, there was only 12K words left for programs and fonts, etc. without going to secondary storage. The new micro-processors will make possible much more flexible display systems. Therefore, I think we should wait for it, but if you have any sources of accurate advance information, it would be very useful.

PDheader:1977-11-07 19:56:00-08:00 06c8b912b726e179cb4947ebe6ff5450 ∂07-Nov-77 1956 BPM DM for Lab To: LES CC: TED

Ted and I both think it would be a good idea to get a DM to keep at the Lab. Below are some ways in which it would be used. Normally it would be connected to the TIP for TIP testing, providing another terminal for SAIL users, and providing ARPAnet access when SAIL is heavily loaded or down completely. The Model 33 TTY currently connected to the TIP, besides being slow and cretinous, seems to have a new problem every week (e.g., CONTROL key sticks, input characters garbaged, leaking oil, etc.). Ted and Don Coates could also use the DM to test out broken boards, modems, etc. that people bring in. This would save people having to take different spare boards home to find out which one is really losing or, alternatively, bring in their entire terminal to be diagnosed. We could also dial up the various split-speed lines to test them out locally. Marty and others could use the DM for debugging new or changed system code or programs such as WHO, E, and DMDRD which know about DMs. Right now, only people with DMs somewhere away from the Lab can debug DM-specific software. The lack of a DM at SAIL also inhibits other people from playing with DMs and dreaming up new hacks. Finally, the larger characters of DMs make them more useful than DDs as a videotaping source. To videotape a recent demo for Bill Carlson, Lou Steinberg had to take the videotape recorder home. (Alternatively, he could have dragged his DM up to SAIL.) Having a DM at SAIL would save a lot of shuffling. I'm sure that other uses will become evident once we get one.

PDheader:1977-11-09 11:48:00-08:00 878d3149f3e813b4eeca9a901398006b ∂09-Nov-77 1148 JP MicroSwitch keyboards To: LES, TED

Since MIT's keyboards are so similar to ours I thought you might be interested (replace,add,etc new keyboards at SAIL) KEY BOARDS: BAKER@MIT-AI 11/05/77 13:24:41 Re: Group order for Microswitch keyboards --More-- I am thinking of placing an order to Microswitch (the Mercedes of keyboard manufacturers) for a batch of MIT AI-style keyboards within the next month. There are two options being considered: the exact replica of the AI keyboard, and a smarter version (with an Intel 8048 single chip microcomputer inside) with exactly the same key arrangement, keytops, etc., but which tells the computer not only when a key goes down, but also when it comes up. Details are in AI:BAKER;SMART KB. The price of the AI replica will be approximately $200, while its smarter brother might be as much as $50 more. The delivery time on these keyboards is approximately 12 weeks, so you probably won't see your keyboards until March. If we place this order, a large deposit will be required, on the order of $100/kb. Please let me know if you are interested, and how many. Henry Baker (BAKER@MIT-AI) P.S. Those of you on foreign hosts, please pass this message on. ∨

PDheader:1977-11-11 07:41:00-08:00 1ae3f89c2d756ca9d3bba291742586d3 ∂11-Nov-77 0741 JRA gasp!!!!

did you realize that there is a significant distortion in the XGP output. eg a basl30 line which measures 29 pica-4 points counts out at 30 pica -2 points when you use the character widths and col-inc output from PUB?? (or am I mis-interpreting the chr width info. in the font files??) if it is xgp distortion please don't change it yet, but it might influence other publishing enterprises around here. john

PDheader:1977-11-23 14:10:00-08:00 9bff99404a76b4d306cdc48ab87de43f ∂23-Nov-77 1410 LES C1 payoff To: DWP CC: JBR, TED

Here is a slightly revised version of your letter. It changes your date arithmetic, reduces the penaly accordingly and adds a list of deficiencies that are to be corrected. -------------------------- Foonly, Inc. 22 Nov. 1977 Les Earnest Stanford Artifical Intelligence Project PENALTIES AND PAYMENT FOR C1 CHANNEL We propose that the penalty for late delivery of hardware ($100/day) should apply to those periods when Stanford's utilization of the C1 was delayed by non-delivery. The contractual delivery date was 31 July (90 days from 2 May). The C1 was made available to Jeff Rubin for system debugging on 6 Aug. Only one channel was working, but this occasioned no inconvenience because there was no intent or provision to connect the second channel to a controller at this time. The second channel was available before it was needed. After Stanford began to use the C1, several bugs were found, resulting in some additional delay to Stanford. We estimate this delay as three more days, including lost system debugging time and system down time. Documentation was due within two weeks after the equipment due date, i.e. by August 14. This was actually completed on 22 Nov., or 99 days late. Therefore we propose that Stanford's total payment to Foonly, Inc. be 20000-(6+3)*100-99*25 = $16,625.00. I acknowledge the following minor deficiencies, which will be corrected within the next two weeks. 1. Rear door is missing. 2. Strain relief and routing on flat cable to channel will be improved. 3. Third connector to IBM will be installed. 4. Light panel vector board will be replaced with something more substantial. If you concur with this analysis, please consider this letter to be our invoice for this amount. Sincerely, David W. Poole, for Foonly, Inc.

PDheader:1977-12-01 19:36:00-08:00 fdeb25832f9413e705bbff331f424cd9 ∂01-Dec-77 1936 SGK

Borrowed portable VTR for the evening.

PDheader:1977-12-03 21:35:00-08:00 1e78be4eaa581fe73771c9a5ddb20153 ∂03-Dec-77 2135 HPM Tape drives and dumps

Restoring a file for DEK revealed that tape T29 is almost unreadable on either drive, containing many dozen bad block all through its length. The priority for new drives should be incremented. Our file backup system is in jeopardy. Tape T24, the only other place this file (ROMAN.SAI[1,DEK]) was dumped was even worse. After uncountable advance commands and multiple backspaces and retries we managed to get the file off T29. Dozens of tries on T24 failed, and we finally gave up on that tape. It looks to me as is our drives are quickly becoming write-only. We already have simpler methods for that function. The Datacomputer, despite the cretinous local program (DFTP) is somewhat more reliable; the main hassle is DC inaccessability at times, not loss of data that has been successfully backed up. DFTP is not a good program, but unfortunately I do not have the time for writing the program that needs to be written. Perhaps one of the hot LOTS kids who is itching for something to do can be drafted to do this? -- mrc

PDheader:1977-12-04 20:22:00-08:00 4dc677283f5359189c0dcf2ad7181be5 ∂04-Dec-77 2022 MRC

PDheader:1977-12-13 05:51:00-08:00 cc0247a50238a62c02d8b2f62d8cf696 ∂13-Dec-77 0551 LES SAIL character set on Datamedia To: PMF

It is probably too late to bring this up, but have you considered getting the LLL Datamedias with a ROM containing the full SAIL character set rather than IMSSS? I would like to modify the rest of our terminals along that line.

PDheader:1977-12-17 16:53:00-08:00 7ae659eb67096e1758cad47feeedb676 ∂17-Dec-77 1653 LES Noise on DM output To: TED CC: ME

While running on TTY-3 (497-2776), the "all white" symbol is printing frequently (once every couple of minutes). I presume that this is caused by spurious "0" characters being received. Seems to be nothing wrong with keyboard side of link. ME - The box character is 177, not 0. I have noticed it occasionally both at home and here at the lab with a DM connected directly.

PDheader:1977-12-17 17:51:00-08:00 1bce2ff607d2e5240a04f1b7e88aeac6 ∂17-Dec-77 1751 ME To: LES, TED

PDheader:1978-01-09 14:36:00-08:00 0b432b37fa4f7e16bff08ef921130319 ∂09-Jan-78 1436 FTP:Tucker at SUMEX-AIM TAPE RACKS Date: 9 JAN 1978 1436-PST From: Tucker at SUMEX-AIM Subject: TAPE RACKS To: LES at SU-AI

LES, I just thought I should try to close the loop on the Mars Tape racks in the basement. We have packed in boxes all the Mars Tapes and put them in the Genetics storage area. Likewise the racks have been put in the storage area. I had talked to Ted and he said that all of your tapes are in Cannisters, not the hanging tape seals which these racks require. He thus said he had no use for the racks. You people are more than welcome to have the racks. From Hershey's records they appear to have been purchased under the JPL contract with the Genetics Dept. and thus must be properly turned over to you (whatever that means). I will attempt implement any plan you prefer. Of course E. Levinthal has the final say at this end if it is Genetics stuff, but I am quite sure that he would be glad to see you make use of them. Bob Tucker ------- .<< Core RFP >> .maxtemplate←10000; .macro core(addressee) ⊂

PDheader:1978-02-26 18:39:00-08:00 a6ebd74ed39f9e9910cf81c0d3fd0986 ∂26-Feb-78 1839 DEK a thought re Dover

If Dover would be used to make CS reports, it would be necessary either to have a collating attachment or to recompute each page each time. In either case it is hard to say that the cost of the memory usage is negligible. The Alto II that goes with Dover seems to have 4 extra hardware cards to make something computed fast, I dunno what, but maybe we should think twice before deciding not to have it. My main thought was that the collating and binding part of printing costs is not negligible, and should be considered when estimating how much money we'll "save".

PDheader:1978-02-27 21:32:00-08:00 a5b41cbe6dcb126a7926c119bc829c34 ∂27-Feb-78 2132 DPB New DataMedia package available To: BPM, TED, LES, JMC, JMB, SSO, JLH, EJM, BS, GIO To: kanerva at SUMEX-AIM, rindfleisch at SUMEX-AIM To: buchanan at SUMEX-AIM, wilcox at SUMEX-AIM To: kahler at SUMEX-AIM

Martin Hardy of Product Associates in Redwood City has been working on software for the new micro-processor based DataMedia terminal. He is prepared to deliver the following package immediately (modulo the normal 60-day DM lead time) : DM 3025 CRT $1700 (less educational discount) -340 Keyboard (similar but not identical to the SUMEX keyboard) plus his program $ 200 Prentice 1200/150 modem $ 325 Installation of modem, switch, cable $ 75 ------ Total price (approximate) $1960 The software will make the 3025 look like our modified 2500A DMs both to the user and to the host computer software. (I.e. Bold face, no protection, no parity instead an EDIT key, etc.) These terminals will run TVEDIT and E without change to host software. The keyboard will be a modified version of the standard keyboard. That is, in addition to the main key pad, there will be a top row of special keys. We can specify what codes we want under those. Therefore, what is now the special row to the left of the main pad on the SUMEX kbd could be included in the upper row. The normal numeric pad will be replaced with the same pad as is normally on the SUMEX keyboard, except we will have a few more keys to "play with." Some (Ted Panofsky at SAIL, Pentti Kanerva at IMSSS) have suggested that in the long run we will want to design a new "smart" keyboard. Hardy is interested in building what we want. In the short run, however, we will order keyboards close to identical to the SUMEX keyboard. This message is being sent to as many people as I know who are or might be interested is this package. Please forward to any other interested parties. -Denny Brown (DPB@SAIL)

PDheader:1978-03-03 14:49:00-08:00 801312a76236923e509f7375cc2f536c ∂03-Mar-78 1449 FTP:Feigenbaum at SUMEX-AIM LISP timing experiments Date: 3 Mar 1978 1443-PST From: Feigenbaum at SUMEX-AIM Subject: LISP timing experiments To: jmc at SAIL, les at SAIL

Date: 3 Mar 1978 1422-PST From: Rindfleisch Subject: FYI - LISP TIMINGS To: SWEER, LEDERBERG, FEIGENBAUM, BUCHANAN, SMITH Mail from SRI-KL rcvd at 3-Mar-78 1025-PST Date: 3 Mar 1978 1025-PST From: Lynch at SRI-KL (Dan Lynch) Subject: Lisp Benchmark Timing across many Machines To: Hart, Sacerdoti, Raphael, Heathman, MMcM, Victor, To: Ellis at ISIB, McKinley at ISIB, Rindfleisch at SUMEX-AIM, To: Lynch Recently I got timing information from DEC on the 2020. I had them run the Lisp "benchmark" job that I have been running for years on various machines. That prompted me to run the same test on as many machines as possible (by type) to get comparison information. The table below summarizes the simple throughput data that were collected. All tests were run on unloaded machines. Notice the difference between 256K and 512K of memory on the KA machine. There may appear to be some anomalies in the peformance data on the 2020. I have no explanation. It looks like the 2020 is at least as good as a KA. Here is the data: Lisp Benchmark Timing Data KA KA 2020 KI KL-1090T 256K 512K 512K 512K 1024K Function Build (ms) 31.8 29.9 25.0 16.7 5.1 (ratio) (0.94) (1.00) (1.20) (1.79) (5.86)

PDheader:1978-04-01 19:07:00-08:00 d85514edd9c610b504cb2ef25402d18d ∂01-Apr-78 1907 CBF at MIT-MC (Charles Frankston) Terminal Demonstration Date: 1 APR 1978 2206-EST From: CBF at MIT-MC (Charles Frankston) Subject: Terminal Demonstration

There will be a demonstration of a Concept 100 display terminal Wednesday 5 April. Likely location for the demonstration is some office on the 8th floor of 545 Technology Square, although if the Multics people want, it might move someplace else part of the day. The exact time is not known yet, but I anticpate most of the day. For those who don't know yet, the Concept 100 is a 24 by 80 line display terminal, featureing all the usual features found in a VT52, HP2645, DM2500, etc. plus support for windows and option for 1 level of overstrikeing. A concievable MIT quantity 25 price for the unit would be $1261, or $1400 with the overstrike option. APL is also available, but you cannot have both Ascii and APL overstriking capability. The quantity one non-educational price is $1575 or $1750 non-overstriking. Much more info is available from the ITS info subsystem, under the "Terminal Support" node (>Term>Term>Term>Concept 100). If anyone who cannot get to this information desires it, tell me and I will mail it to you. Please feel free to forward this to anyone I didn't think of, especially people who are not reachable via Arpanet mail.

PDheader:1978-04-01 19:12:00-08:00 18048ad3c4e497540c8650774cd4811b ∂01-Apr-78 1912 BH Candidate for home terminal for SAIL? To: LES, ME, TED, BPM

You might want to check out this message.

PDheader:1978-04-07 02:09:00-08:00 b8d3aaefa717a06bef498b1a92fda921 ∂07-Apr-78 0209 ROZ

D7-Apr-78 0139 Mail account, UDS modem Our DCA contract is terminating soon, we are making a big effort to finish it. Please open for John Newkirk, one of our new RA's, an account. I suggest NUK as programmer name. Could you advise the secretaries so that Larry Stewart (LY) and John Newkirk (NUK) could get keys for the front door, thanks. We have a 300bd UDS modem for sale, are you/AI interested? Its new, never used! One split speed UDS modem (also automatic answer) might be available. We are in the process of modifying Hazeltine terminals, to look like AI type DM's, Ralph Gorin gave us one to modify already. Are you interested? Stanford pays $860.- for Hazeltines. We are also developing a cheap (comercial LSI chip set based) modem for 300bd and hopefully split speed. The DCA DM's are required back soon. The two UDS modems mentioned above could replace the two DCA UDS modems. I already talked to Jeff Rubin about the DM. I also suggested then to him to get UNIX for the 11/45, its cheap ($125) and we are very pleased with it. -andy --------------- Thank you for your efforts. Martin Morf % ROZ ----------------- Return-Path: <unix-wizards-request@BRL-TGR.ARPA> Received: from BRL-TGR by SUMEX-AIM.ARPA with TCP; Wed 17 Oct 84 14:49:57-PDT Received: from ll-xn.arpa by BRL-TGR.ARPA id a006584; 17 Oct 84 16:45 EDT Received: by ll-xn.ARPA (4.12/4.7) id AA24683; Wed, 17 Oct 84 16:45:53 edt Date: Wed, 17 Oct 84 16:45:53 edt From: Glenn Adams <glenn@ll-xn.ARPA> Message-Id: <8410172045.AA24683@ll-xn.ARPA> To: david@COLUMBIA-20.ARPA Subject: Re: Tightly-coupled multiprocessor UNIX Cc: ted@COLUMBIA-20.ARPA, unix-wizards@BRL-TGR.ARPA Dave, I would call your attention to the recently released Balance 8000 system from Sequent Computer Systems, Inc., Portland, OR. They have taken the 4.2BSD system and modified it extensively to operate in a tightly coupled environment using a processor pool architecture. Their system uses the 32XXX series of CPUs from National Semiconductor with the FPU and MMU from that same series. In addition, they add an 8-KByte, 2-way, set associative cache and a high speed 8 Kbyte local memory used for read-only kernel data. Two autonomous CPUs are packaged per board with up to twelve supported in the system. A custom VLSI chip, the System Link and Interrupt Control (SLIC), handles interrupt handling contention and implements mutual exclusion semaphores. A distributed arbitration, 40 Mbyte/sec, synchronous bus (an enhanced VAX SBI?), handles communication between processors, memory, a multibus adapter, and a SCSI Bus Adaptor/Ethernet Module. A separate synchronous serial bus ties each SLIC together, each unit on the main bus possessing a SLIC. Software-wise, you get a 4.2BSD system with its Virtual Memory system completely replaced with one of Sequent's design. A new model was employed in its design that results in processes paging against themselves, rather that against the system at large, etc. Their performance figures show near linear performance up to the full complement of twelve processors over a variety of benchmarks. They are currently shipping with 10Mhz 32016's (half a VAX750), but soon hope to be using the 32032 as soon as they can get reliable, i.e., full temperature range, chips from National. With the latter, the system is fully worth a dozen VAX750s. If some of the above sounds VAXEN like, or an improvement thereof, it is probably due to their V/P of engineering, Dave Rogers, who was on the VAX780 architecture team at DEC. Personally, I am going to watch this company carefully. I think they have a handle on a good architecture and have did all the leg work to produce a working system. If there is anything lacking, it may be the limitations (!?!) of the 4.2BSD network system in producing a viable networked file-system. However, given that multiplexor files are still warm in the grave, I think that a lot can be done with the generic network architecture in the 4.2BSD implementation. This reflects little on Sequent's system and I urged Mr. Rogers to continue in the same vein and add a good networked file-system on to the Balance 8000. For further information, see Electronics Design, September 6, 1984, pp. 153-168. Glenn A. Adams glenn@LL-XN.ARPA MIT - Lincoln Laboratory ------- ------- -------

PDheader:1978-05-15 06:49:00-07:00 45c58357b9b788828fbde1b9a1811486 ∂15-May-78 0649 FXB via AMES-TIP#14 terminals

What is the status of departmental terminal projects, if any? I guess I would like to get involved and do something although it would likely be something that didn't quite fit your view (as I have understood it in the past). Are you thinking about this subject these days? As you know, I have wanted bit map terminals but found your old terminal too expensive; a centralized memory seems to be cheaper. I have delayed the project in the AI Lab, at LOTS and in the Department for two reasons. Mainly, the new 16 bit large address microprocessors forthcoming from Intel and Zilog has made me want a larger address, because squeezing all the fonts, etc. into the 12K address space that remains in a PDP-11 after 16K is used to address the bit map seems marginal. A larger address space makes the size and shape of the bit map more flexible, and we can contemplate a system that will grow to high resolution and even color if we have enough address space. A secondary reason is that the price of 16K RAMs hasn't declined as rapidly as was expected. In short the project is delayed but not forgotten. Johnathan King has been pursuing money, and he has different ideas. I propose to argue for my ideas when money arrives or the technology is ready - whichever is first. It seems to me that agreement will be easier, when we are ready to go financially, because some of the options will be clearer.

PDheader:1978-05-15 09:33:00-07:00 88ed5f3180715787ef4a966c93e6880f ∂15-May-78 0933 JMC To: FXB CC: LES, JJK, feigenbaum at SUMEX-AIM

PDheader:1978-06-01 09:47:00-07:00 bc8c4ff6dab8f92c20d55e90793a4781 ∂01-Jun-78 0947 JJK To: MAIL2.SYS[CSD,JJK]:

1. I have not heard from many people about their planning ideas for the file EQUIP[CSD,JJK], the subject of my last message. Please contact me if you haven't already. Your ideas will be helpful to everyone on the list. Thanks.

PDheader:1978-06-14 06:10:00-07:00 f348ebab7a5e3b64fafa0562d0fa7b70 ∂14-Jun-78 0610 FB via AMES-TIP#66 VAX To: MAIL2.SYS[CSD,JJK]:

I am going to visit Digital on the 26-28 of June to try to pursuade them to donate a VAX to CSL for research purposes. I have a list of projects that I want to pursue with such a machine that would seem to have short and long term benefits to Digital. I am listing some of those projects below and asking you to think of projects of your own that might have some appeal to DEC. You might even write down a line or two and we can perhaps discuss such ideas next week. My overall strategy is to try to convince DEC that they can very nearly capture one of the brightest and newest systems groups in the country for lots of VAX work much like the 10 did years ago. Synchronous Bus Study: The SEL 32, VAX 11/780, and all the four medium scale 370 plug compatible CPU's are built on a high speed synchronous bus. It is time we had a more scientific understanding of that design space. Even IBM's new E series is rumored to be built on a synchronous bus. A VAX 11/780 would provide one in depth example for such a study as well as providing a vehicle for measurements of a llive system using a particular design. (Long Term Benefit) Pascal: The Pascal P compiler and our experience with three different P code translators (370, Cray-1, S-1) should make it possible for us to bring up a Pascal on the VAX and one which is quite respectable. In fact, we should be able to beat Univ. of Washington and, with the P Code optimizer that R. L. Sites is doing, we should be able to do a better job than UW. (Short Term Benefit) Optimizing Loader: The P code compiler and the P code optimizer make use of a table of basic block execution counts to give the user an execution count annotated listing and to give the optimizer the information it needs to make the right decisions about code motion. A loader could also use that information to construct load modules that minimized page faults in instruction space. Such a loader would do a much better job than all previously proposed and studied schemes. (Short Term Benefit) Instruction set architecture: I would like to get a slightly modified VAX that allowed complete replacement of the microcode much like the 11/40E that CMU devised and had DEC make. The machine has an 8K word microcode address space, 4K of which is used. High speed 4K RAMS are available now that would compete in board space and speed with the ROMs in the current microcode so the engineering looks straightforward. A softer machine structure would allow a number of studies ranging from appropriate targets for optimizing compilers to appropriate targets for operating system structures. It might also be possible to investigate whether the machine definition can be changed in such a way as to make a high speed implementation more feasible. (Long Term Benefit) Knuth's Text Processing Software: Obvious. Networks and Software Development Environment: The machine would need to be connected to other machines and would presumably be quite useful for microprocessor software development work. Barth and Hennessey? Computer Aided Design: vanCleemput? I see three possibilities: 1. Digital gives in and we win big. 2. Digital says "Why doesn't Stanford contribute x% and DEC will contribute 100-x% (Where do we get x?) 3. Digital says "Why don't you take the prototype that was used for VAX design and development work. This possibility is the one that would be most difficult to deal with and decide about. I will give more details next week. In the meantime, if you think you can contribute, think of some flashy things for me to tell them.

PDheader:1978-06-14 21:04:00-07:00 8242771ed9585543388e1fb43b93e07c ∂14-Jun-78 2104 Mclure at SRI-KA (Stuart Cracraft) Datamedia Date: 14 Jun 1978 2106-PDT From: Mclure at SRI-KA (Stuart Cracraft) Subject: Datamedia To: bpm at SU-AI

I am looking to buy a personal Datamedia Elite 2500A like SRI, SAIL, and SUMEX all have with the special keyboard. Can I buy one through SAIL and perhaps get the educational discount (I am an undergraduate at Stanford) along with the discount for large batch orders that SAIL gets when ordering groups of DM's?

PDheader:1978-06-15 09:08:00-07:00 1d60ebf6956368451f8b9b9df1e0eff0 ∂15-Jun-78 0908 BPM DM order parasite?

What sez ye?

PDheader:1978-06-28 11:50:00-07:00 11e40d7bada40f34771ec713d8b15426 ∂28-Jun-78 1150 FB via SU-TIP#3 keyboards

How do I order an AI keyboard? I heard some rumor that Stanford was switching to the MIT keyboard or vice versa to simplify life for Microswitch. Do I get specifications with the keyboard (assuming I can get one)? What is the lead time? If you could do it for me, I can give you a Stanford account that could pay for it.

PDheader:1978-06-28 12:17:00-07:00 f3a2216a2ddf25e30283ffe0e05db2b9 ∂28-Jun-78 1217 FB via SU-TIP#3 terminals

I had a thought for a box to go between a clutzy terminal such as a Hazeltine and the line that would make it more usable with somethine like E. It would be a box with a microcomputer and perhaps a modem (1200) and topped with a keypad. I call it an E box. Does that sound like a good, bad, or indifferent idea?

PDheader:1978-06-28 16:13:00-07:00 58c1db3bfe29d99d09badd67954c4cac ∂28-Jun-78 1613 LES Keyboards & terminals To: FB

We do not intend to switch to the MIT keyboard, and never did, because it is a step backwards. Ted has been trying to get Microswitch to bid on an improved keyboard, containing a microprocessor with either serial or parallel output, but is having great difficulty getting their attention. We may have to go someplace else. You E box idea sounds like a good idea if the economics work out. We can now buy a microprocessor-based Datamedia that does the same thing and includes a 1200/150 modem for just under $2000.

PDheader:1978-06-29 07:13:00-07:00 c6b29eb8987357ceefd26ae52c546383 ∂29-Jun-78 0713 FB via SU-TIP#1 keyboards

I take it that it is currently impossible to get a Stanford keyboard? Is the MIT keyboard a step backward from a Datamedia keyboard? You might try Cherry; I have been impressed with the quality of there current line. Tektronix says it is more reliable than a microswitch keyboard. The feel is not perfect but the keytops are lovely and can be ordered with two piece keytop. The two piece keytop has an inner part that attaches to the plunger and on outer shell that fits over the inner part. The outer shell is transparent (non glare) and removes so that any ledgend can be put on top of the inner piece. Is there a pressing need for more keyboards at your place or is this just a low level activity that may drag on for a long time?

PDheader:1978-08-01 17:20:00-07:00 b355ff49b7a1f309d677afa170f7f3b8 ∂01-Aug-78 1720 MGF To: DPY.DIS[1,MGF]:

Below are suggested requirements for the vision group image display system. Items preceded by a * are optional and considered negotiable. Comments, suggestions to @DPY.DIS[1,MGF]. File DPY.DOC[1,MGF] contains this and more display information. Requirements Memory - 512x512x8x3 - read and write operations; addressable from PDP-11 - X, Y register and increment mode - expandable for more memory planes and larger display size (1024x1024) - NPR (DMA) direct from disk over Unibus TV input - digitizer system - TV input at 10MHz, 8 bits (direct to video memory) - broadcast TV compatible mode (480 lines) Overlays and graphics - cursor - ability to use overlay channels as intensity channels - character generator - writing into memory - vector generator * non-jaggies * conics Display capabilities - display at standard TV rates (30 frames/second) - broadcast TV compatible mode (480 lines) - zoom (prefer interpolation) - X, Y scroll (independent for 3 color channels) Interfaces - PDP-11/45 Unibus * TI-990 (?) Input devices - (2)joysticks or (2) trackballs - response rate to track ball (60 times per second or interrupt) Outputs - (2) color monitors - lookup table for each D/A - ability to use overlay channels as intensity channels - flexible switch to select overlays - switchable - 512x512x8 stereo - 512x512x8x3 color - 256x512x8x3 color stereo - 6 4-bit users - 3 8-bit users Options for larger system - image array processor - general purpose digital switch - map any plane to any D/A - interpolation zoom - extra memory planes

PDheader:1978-08-04 20:07:00-07:00 5f07adc527b322e46a71c4dc34d59468 ∂04-Aug-78 2007 TED To: DPY.DIS[1,MGF]:

I have read the suggested requirements for the VISION GROUP IMAGE DISPLAY SYSTEM. It all seems fine to me with the following two comments: 1) in order to use the display for the drawing system a vector generating capability would be nice. 2) I think maybe this is a little too glorious, but you can always try.

PDheader:1978-08-09 16:51:00-07:00 a77dbcc70b081b80e26c7e9a2d5533a0 ∂09-Aug-78 1651 BS Purchase of Disc File

Additional information, Les. One-half the cost should be charged to 2-FCZ161 (Knuth Unrestricted) and the other half to 1-FCZ00l (CS Operating Account). Please call me if you have questions and please route the requisition through here. I would like Ed Feigenbaum to countersign it. Thanks. Betty

PDheader:1978-08-14 16:20:00-07:00 d25c557f3a931203c34360bdf506b1cd ∂14-Aug-78 1620 LES To: HVA

PDheader:1978-08-15 16:32:00-07:00 67df649f70a21c11a77104764b008b1b ∂15-Aug-78 1632 DPB modems

EAF may be able to acquire "unlimited numbers" of 1200/1200 terminal end modems. NIH is apparently trying to unload a bunch. He thinks that we could get 100 or so, for the asking. Interested? -Denny

PDheader:1978-08-17 18:58:00-07:00 b3c8c0ee96224cf5f5267a2de9bcc118 ∂17-Aug-78 1858 RPG Terminal

If the DM3025's finally meet lab specs, I would be interested in obtaining one. Perhaps we should get together sometime to discuss terms before any are ordered. If, however, the old style DMs (2500) are ordered, I would not be interested. Thanks. -rpg-

PDheader:1978-08-23 13:00:00-07:00 474c9da962d65af3ab9325a416a7e9f5 ∂23-Aug-78 1300 TOB planned to hold up payment on the device.+[Kw

PDheader:1978-08-23 13:00:01-07:00 fbdf33b525d700c961c2de56403d5b1b ∂23-Aug-78 1300 TOB To: DPY.DIS[1,MGF]:

Addresses and Phone numbers Interpretation Systems Inc 6322 College Blvd Overland Park, Ks, 66204 913-642-8700 Lexidata Corp 215 Middlesex Turnpike Burlington, Mass, 01803 Matrox Electronic Systems Ltd, P.O.Box 56 Ahuntsic Sta. Montreal, Que, H3L 3N5 514-481-6838 [373] Ginesco [415] 941-0240 W Joe Jones 2680 Bayshore Frontage Rd Mountain View, Ca 94043 Ramtek [408] 735-8400 Mark Sayer 585 N Mary Ave Sunnyvale, Ca 94086 [system 9400 available September or later] Grinnell [408] 988-2100 [1pm Wednesday] John Schlueter, Vice President 2986 Scott Blvd Santa Clara, Ca 95050 De Anza Systems, Inc. [408] 988-2656 Chuck Masters (SRI contact) Wayne Hepler Charles Nordby 3446 De La Cruz Blvd Santa Clara, Ca 95050 [SRI has system from them They also have TI 990 interface] Evans & Sutherland Computer Corp. 792-5604 [Wednesday 3pm] Taylor Wright 38024 Martha Avenue Suite C Fremont CA 94356 [have reputable frame buffer] Comtal Corp. [213] 793-2134 Dick Jones 169 N Halstead Arcadia, Ca. 91107 Stanford Technology Corp 737-0200 Larry Dorie Esmond C. Lyons, Jr. 650 N. Mary Ave Sunnyvale, Ca [systems at UC Davis(Ralph Algazi), NBS] [article in recent Digital Design describing system for LANDSAT applications] Rohde & Schwarz Sales Co [201] 575-0750 14 Gloria Lane Fairfield, NJ 07006

PDheader:1978-09-14 18:28:00-07:00 2b0892c344560a69de58623456b1e7d1 ∂14-Sep-78 1828 BH disk delivery

I talked to someone named Ken Jones at Ampex, who says that things are not entirely settled about whose delivery is when, but for sure not before Oct 15 and almost certainly not after Oct 31. He'll call me back when he has more detailed info.

PDheader:1978-12-12 10:54:00-08:00 26c60fab524639dbeb21e81d82ff1239 ∂12-Dec-78 1054 TED Varian interface To: LES, DSB

The Varian interface book is inside the plotter. There is a copy of the PDP-10 interface handbook in the rack on top of the KA-10 and another in the bookshelf in "my" office. However, you will find it a lot easier to connect to the TTL IOB Extension rack (assuming a IOB connection) because it eliminates level conversion and device selection. I will be glad to advise. It probably would be a good idea to look at the file VARIAN[PLT,TED] before proceeding.

PDheader:1979-01-05 16:28:00-08:00 9001f1e2ba8488b01054fe9699eff7cb ∂05-Jan-79 1628 TW To: LES, HVA, TW

Les, I just talked to Betty Scott who said that you were going to order my Datamedia. I need a non-standard one which I have discussed with Ron Bondurant at Product Associates. I want to use both SAIL and MAXC systems. The wording he said to put on the P.O. was: Datamedia 3000/xerox with integral Prentice Modem, with exchange blink-bright switch on front panel (as per conversation with Ron Bondurant). If there are any problems with this, I will be here in Polya (7-2780) for a while this afternoon and for most of the day on Monday. Thanks --t

PDheader:1979-01-08 14:10:00-08:00 fdd1f50912eb81109f158c14359ce78b ∂08-Jan-79 1410 TW To: LES, BS

From your short note, I assume you will go ahead and order it right away as a separate order. If there is any more I need to follow up on, let me know. --t

PDheader:1979-01-26 19:35:00-08:00 646a18cdbe6152bf12b1ac8a6542e2ff ∂26-Jan-79 1935 PMF

For some reason the modems aren't supposed to arrive until Late Feb. AUTO ANSWERMODEM 266210 4117409 2/28/79 KEYWORD: REQ.NO: P.O.: REQN.DATA 11/6/78 DU DATA:2/28/79 BUYER: GRISWOLD EX 29287 COST: 550

PDheader:1979-02-03 00:46:00-08:00 f833cdbdc6b02934cdb5f1d353e1b18d ∂03-Feb-79 0046 MITCHW at MIT-AI (MITCH WOLRICH) NEW S-100 BUSS MODEM Date: 3 FEB 1979 0332-EST From: MITCHW at MIT-AI (MITCH WOLRICH) Subject: NEW S-100 BUSS MODEM To: INFO-MICRO at MIT-MC, INFO-PCNET at MIT-MC

A NEW S-100 BUSS MODEM WILL BE ANNOUNCED IN AN UPCOMMING BYTE MAGAZINE (MARCH 1979). THE MODEM IS BEING OFFERED BY POTOMAC MICRO-MAGIC INC. DESIGNATED THE MM-103, HERE ARE SOME OF ITS NOTEWORTHY FEATURES: ORIGINATE/AUTO-ANSWER, 61-600 BAUD (LOCAL), 61-300 BAUD (ANYWHERE), FCC REGISTERED (REG # BJ686B-67773-MD-R), -50 DBM SENSITIVTY, SOFTWARE CONTROLLED (ON EVERYTHING, AUTO-DIALING (PULSE), ANSWER, BAUD RATE..) ASSEMBLED AND TESTED, ADVERTISED PRICE WILL BE $319. IF YOU SEND IN A "RESERVATION CARD" PRIOR TO FEBRUARY 15TH, YOU CAN GET THE MODEM AT THE PRE-ADVERTISED RATE OF $259. PMMI HAS STATED THAT IT DOES NOT WANT MONEY UNTIL THE PRODUCT IS AVAILABLE, BY SENDING IN A RESERVATION FORM, YOU ARE MERLY STATING YOUR INTEREST IN OBTAINING A MODEM, YOU ARE NOT COMMITED (THEY SAY FIRM ORDERS WILL BE ACCEPTED IN THE ORDER THE CARDS ARE RECEIVED.) IF YOU CALL THEM AND TELL THEM YOU WANT ONE, I'M SURE THAT WOULD SUFFICE. THERE ADDRESS IS: POTOMAC MICRO-MAGIC, INC. P.O. BOX 11149 ALEXANDRIA, VA 22312 (703) 750-3727 P.S. - I AM JUST PASSING THIS INFO ON, I MYSELF HAVE HAD NO DIRECT EXPERIENCE WITH THIS MODEM, EXCEPTING THAT I HAVE ONE ON ORDER WHICH WILL BE USED IN BOSTON CBBS, IN MID-MARCH IF YOU WANT TO TEST IT.. ALSO NOTE, THESE GUYS ARE THE SAME WHO DESIGNED THE IDS MODEM... THEY SEEM QUITE COMPETENT.. ANOTHER SHORT NOTE, THEY SAY THEY HAVE PROVISIONS FOR INTERFACING VOICE-RECORDER-ANNOUNCEMENT EQUIPMENT AND HOOKING UP ALARAM AUTO DIALING EQUIPMENT AND "POWERING" UP THE COMPUTER ON RING (YOU SHOULD SEE MY ELECTRIC BILLS THESE LAST 2 MONTHS!) I WILL BE GLAD TO RECIEVE COMMENTS ETC ON THIS MODEM IF ANYONE OUT THERE HAS ONE YET... - MITCH P.S.S. - THEY SAY THEY'LL ACCEPT RESERVATION FORMS UP TO FEBRUARY 15TH.

PDheader:1979-02-06 11:47:00-08:00 7d17d175b2eb76044c533d2d9a0c90ac ∂06-Feb-79 1147 TOB To: LES, HVA

Would you make out a purchase order to Unimation, Inc Shelter Rock Lane Danbury, Conn,06810 203-744-1800 for two Unimation 600 six degree of freedom manipulators? Charge to old NSF contract

PDheader:1979-02-14 15:52:00-08:00 a3926aa39ef876ddab71654fcf1fcfe5 ∂14-Feb-79 1552 LES DM terminal

We are about to order the Datamedias finally (really!). The specs are in my office, if you are interested. The unit cost is $2406 plus tax. Please confirm that you wish Stanford to buy one on your behalf and at your expense. Ok, however, the only possible hangup could be the procedure if the thing loses and needs fixing due to initial screwups (things such as original defects). Is there some "warranty" period? In most resonable answers to this question my answer is yes. -rpg- LES - Warranty is 90 days after delivery. ------- End of Forwarded Message

PDheader:1979-02-14 16:06:00-08:00 47e867978b1e78c5127836c9b47a1a05 ∂14-Feb-79 1606 RPG

PDheader:1979-02-14 16:09:00-08:00 580d0186cbb713fec73679c064386c11 ∂14-Feb-79 1609 LES To: RPG

PDheader:1979-02-14 21:27:00-08:00 af0b692a6f5dd67106713e8d06c0ed8a ∂14-Feb-79 2127 JMC ka system To: ME CC: LES

The musicians may get the KA as a separate system with one disk and controller, Petit channel, tapes and possibly data disk but not the switch and 256k memory. Is there a system that can readily be resuscitated to run on such a configuration. ME - I wouldn't say "readily". The code to run the Petit channel and the code to make the system work on a KA are no longer around, though they could be restored from DART tapes and worked into a useable state. Running the DD without the video switch would also require a bit of reworking of the display service, essentially back to what it was like before the switch (with each monitor viewing a fixed DD channel). Out of curriosity, when might this happen and which 256K of memory are you thinking of? It certainly wouldn't be impossible to do; it might take a month. This assumes the hardware still works (the channel).

PDheader:1979-02-15 19:49:00-08:00 9ce1727270d4ab4054c9ad02751cb13a ∂15-Feb-79 1949 ME ARM-10M interleaing tests To: LES CC: ME

I ran the worst-case interleaving tests of the ARM-10M with >the KL-10 caches off, >the system code and data itself located in the ARM-10M memory, >both disk channels running as fast as user programs could run them, >a CPU-bound user program designed always to miss the cache, >a user program running the DD as fast as possible, >a user program displaying on a III. Here are my conclusions. The CPU-bound user program didn't seem to make much difference, nor did the III usage. With 2-way interleaving, the system works only if EITHER (1) the system itself is in FAST core (e.g., the MG-10), OR (2) the KL-10's caches are ON and the Data Disc is not doing much. With no interleaving, the system works reasonably only if both (1) and (2) above are the case. The various modes of lossage are: (a) SBUS errors detected by the KL-10, sometimes including NXMs and/or parity errors (although no real memory parity errors exist), (b) C1 PC different from expected halt address, and (c) DD programs timing out because of slow response from memory. Mode (b) probably means the C1 couldn't read its own program out or the ARM before it was too late. I don't know exactly what is causing mode (a), but probably it is from the DD interface becoming unhappy; the C1 also may have caused that error, but I would suspect the DD. Mode (c) is tolerable in the small quantities that were observed when (a) and (b) were not occurring.

PDheader:1979-02-17 03:59:00-08:00 bea308210d5b8d8d3bbbbeedacd00868 ∂17-Feb-79 0359 ME KA system To: JMC CC: LES

PDheader:1979-02-20 09:03:00-08:00 1f7d79deb16a38ed226e8cb59edba6e5 ∂20-Feb-79 0903 Feigenbaum at SUMEX-AIM musicians/mj hall space Date: 20 Feb 1979 0859-PST From: Feigenbaum at SUMEX-AIM Subject: musicians/mj hall space To: jmc at SAIL cc: les at SAIL

John, I am eager to take the next step. I owe John Chowning a response to the long shopping list of equipment he said he needed. Our response was to be given in terms of a KA-10 system configuration, but you and Les were going to think that one through and send something to me. I would appreciate closure on that. We are pursuing the other manufacturers, such as Synapse, Sierra, and Sequent. -rpg- Ed -------

PDheader:1979-02-20 16:09:00-08:00 4a3393ea5e0d1482e062fc6c03b87a5c ∂20-Feb-79 1609 JMG To: LES, MUZ, JMG, KS, TVR, LCS CCRMA needs for independence from SAIL:

Hardware KA10 & BBN pager & 512K memory (Suppes+Ampex+256k) $100k + 4 mo. PDP6 & 128K memory (old Ampex + old DEC) 10 (cables) Versatek or equivalent $10 + 2 mo. 3 3330 dsk drives + 2 @ $14k 28 + 3 mo. & 6 packs in addition to ones we have now 6 @ 600 3.6 Displays: 8 vector, 8 raster/graphic (dd equivalent) 96 (Imlac) & 8 tty scanner lines 19.2 (DMs) 4 1200/150 modems (800 + 4 x 250) 1.8 (UDS) TTY scanner 14 (DCA) Disk channel, contoller 16 (Foonly) Line printer 10 (Printronix) 9 track magtape drive 20 Fast connection to SAIL on campus 2 (Dialnet) ----- Software $330.6 System (modified current SAIL system for KA10) Paging software

PDheader:1979-02-20 16:09:01-08:00 1dbc62b1a531f29ae2c7ee26598a088c ∂20-Feb-79 1609 JMG To: LES, MUZ, JMG, KS, TVR, LCS CCRMA needs for independence from SAIL:

PDheader:1979-02-20 16:53:00-08:00 cc47fc7998dc798236dfb785ebad5716 ∂20-Feb-79 1653 TVR Other considerations To: KS, MUZ, LCS, JMG, LES

It should be noted that there is also the cost of machine conversion. I have been careful not to use KL-10 instructions in my code, however i do see them is lots of user programs. This is a conversion which must be taken into consider, that is, that all SAIL programs will have to be recompiled and many FAIL programs will also require modification. Most libraries will have be at least partially reconstructed. This will be especially painful where source maintanence. Perhaps it might not be a bad idea to put a moratorium on KL-10 instruction usages within the Music Group. Comments?

PDheader:1979-02-20 16:53:01-08:00 13fdfc406c024fea71e54a0ca2e97328 ∂20-Feb-79 1653 TVR Other considerations

PDheader:1979-02-20 16:53:02-08:00 54aafd360017d003bef4936ea73d26ad ∂20-Feb-79 1653 TVR Other considerations To: KS, MUZ, LCS, JMG, LES

PDheader:1979-02-20 22:42:00-08:00 28c559267a7329b0c1854c447550ea5a ∂20-Feb-79 2242 JMG CCRMA separation

JAM suggests the addition of 2 modems to the list. Basicly JC and JAM support the separation, if indeed we are able to obtain our needs.

PDheader:1979-03-01 07:58:00-08:00 85e2a6856877422acd5d6a7c1e308c0e ∂01-Mar-79 0758 PHY Alphatye dimensions

34" wide 28" deep 48" high. You have to open it on at least one side. In the same darkroom we will have Agfa developing unit, whose dimensions LTP knows. Also a sink or something needed (LTP knows).

PDheader:1979-03-02 15:53:00-08:00 209b3ab32d2bfe57e77d153fbd72bc35 ∂02-Mar-79 1553 LES To: JMG

PDheader:1979-03-06 16:57:00-08:00 47cda5baa2bae9cd8f7b094de6fd9bb7 ∂06-Mar-79 1657 LES Ethernet and friends To: Feigenbaum at SUMEX-AIM CC: JMC at SU-AI

The Xerox offer of a Dover, Ethernet, etc. looks quite attractive. I note that nothing has been said about dates, though. It appears that some new equipment that SAIL plans to acquire might best be connected to the KL10 and each other through an Ethernet. This includes the new computer donated by TI and an image display system that Binford and Levinthal plan to buy. If we are going ahead with the Xerox deal, it would seem advantageous to start with the Ethernet as soon as possible.

PDheader:1979-04-18 15:50:00-08:00 3353a667d2e730f61231df2774e8ef8f ∂18-Apr-79 1550 JMC conversation with Lowell Wood To: LES, DCL

He returned my call. We can expect the Mark I sometime next winter, but it won't be usable as a time-sharing machine. The second Mark 2A is scheduled for NBS, and we will get one of the following batch. Time-sharing is one-and-a-half to two years away on his schedule. ARPA money could advance that, but Lowell is skeptical as to whether Carlson has money to back up his ideas. I told him that I thought the S-1 project should pay a larger share of the operating expenses of the KL, and we agreed to discuss it next week. We need cost information by then.

PDheader:1979-04-30 19:48:00-07:00 ef4334a7a367ec60a8d6ba2e685876bb ∂30-Apr-79 1948 Mark Crispin <Admin.MRC at SU-SCORE> [G.DACRUZ: DEC's LCG Strategy] Date: 30 Apr 1979 1744-PDT From: Mark Crispin <Admin.MRC at SU-SCORE> Subject: [G.DACRUZ: DEC's LCG Strategy] To: Les at SU-AI, ME at SU-AI, JBR at SU-AI, Geoff at SRI-KA, Moon at MIT-MC

I thought you might be interested... --------------- Date: 30 Apr 1979 1355-PDT From: G.DACRUZ Subject: DEC's LCG Strategy To: Admin.JQJ, Admin.MRC Cc: G.DACRUZ From a DEC internal memo from Gordon Bell, et al, in case you haven't heard about it (eat this message): "Since there have been so many rumers (sic), discussions and suggestions, pertaining to the large computer business in the last few months, we felt it important to clarify our positions on this activity. "In regard to engineering, we are pleased to confirm that we are working on the development of a highly competitive follow-on product for the KL10. In addition, our engineering support for existing systems will ensure their on-going competitiveness. This will not only include hardware and software enhancements but also the very important new maintainability features for the KL10. [some blather about the commercial group] "The technical group is strongly product and applications focussed and wishes to maintain a pure LCG specialty for our high-end business. We will gradually increase our cross-product training in future years as the VAX family grows in both hardware and software capability. All this marked "U R G E N T" and dated April 26. What all this means, according to our sales rep, is that there will be at least one more big 36-bit machine: not the Dolphin (which was to be 2 tightly-coupled -20's fed by a throng of 2005's) - they decided it wasn't cost effective - but something they won't tell us about until next month.. In any case, the 36-bit architecture won't die for a few more years yet. By the way, we're upgrading to a 1 MW 2060 this summer; we took advantage of their quantity discount on memory and ordered 1.75 MW (the balance to be put on some unspecified machine at some future time). -Frank ------- --------------- -------

PDheader:1979-05-05 22:12:00-07:00 f38cab7047140fa7afbf5cba716b9f61 ∂05-May-79 2212 DD at MIT-AI (David Dyer) Date: 6 MAY 1979 0106-EDT From: DD at MIT-AI (David Dyer) To: jmc at SU-AI

This message should be redirected to the systems programmer responsible for your C-1 data chennel made by foonly inc. We at III have the next iteration of the design for the C-1 channel. Be warned that your channel will not work without some hardware changes should you ever decide to use it to run magtapes. It also will not run without some microcode changes if you ever connect more than one controller of any kind to either channel. To contact me, use DD@AI, which is my mail drop, or you can call me at Triple-I.

PDheader:1979-05-09 01:27:00-07:00 99ce5f524865688d960f9567f893e4ad ∂09-May-79 0127 JMC

PDheader:1979-05-09 01:27:01-07:00 6287a26770cc1964e159dcb883e5f294 ∂09-May-79 0127 JMC C-1 channel To: LES, ME

PDheader:1979-05-23 13:17:00-07:00 5916f104c5e8c177b502be2a947d57c3 ∂23-May-79 1317 JMC Rosse memo To: feigenbaum at SUMEX-AIM CC: LES at SU-AI I propose the following counter argument:

Initially the musicians were entirely free loaders. There was no real justification for letting them on a machine entirely paid for by the Defense Department, but I, John McCarthy, felt that the arts should be treated more generously than other scientific projects (who were only admitted if they contributed to the goals of the ARPA project) because of their difficulty in getting support. Music turned out to be compute-intensive, and the few musicians we admitted initially turned out to have lots of friends and students. While they sometimes were co-operative, they always pressed their luck and have been the largest single user of the computer most of the time. There were many complaints from other users, but we kept hoping that the problem would go away, and with improvements in our facilities, and then there arose the willow-the-wisp of the expected Yamaha money. When ARPA support diminished, it became necessary to look for additional sources of support for the machine. Unfortunately, we were not and aren't a cost center and could not charge directly for machine use. Therefore, we asked other projects using the machine to buy us hardware instead. This has proved to be an unsatisfactory solution. as the present situation shows. As I understand it, you propose that the musicians owe nothing for their use of the approximately $2,000,000 in hardware contributed by ARPA and others. Well I can't bring myself to continue with this memo to Rosse but there is also the fact that we don't want their damn disk packs and we wouldn't have needed the second disk controller if we had kicked them off as it now turns out we should. See if you can make something of this, but if I talk to Rosse I will start to shout.

PDheader:1979-05-24 08:37:00-07:00 6894e21aa4fece2cbfc4e921ad06422b ∂24-May-79 0837 Feigenbaum at SUMEX-AIM Re: Rosse memo Date: 24 May 1979 0836-PDT From: Feigenbaum at SUMEX-AIM Subject: Re: Rosse memo To: JMC at SU-AI cc: LES at SU-AI

In response to your message sent 23 May 1979 1639-PDT From connecticut en route: John, I have not seen the Rosse memo to which you refer, but your line of reasoning is identical to mine. I will have to deal with this when I return (to orrow). If he's trying to get more money out of us, then I may reduce our offer from 25k to zero. Ed -------

PDheader:1979-06-21 10:59:00-07:00 e629f9426ffc6750e03acf93277c6ac6 ∂21-Jun-79 1059 AVB graphics To: "@SUN.DIS[1,AVB]" at SU-AI Sirs and Madams:

A raster-scan graphics system has been designed at Stanford that features: - a monochrome display of 1024 by 800 points - vector writing speeds of 1 pixel per microsecond - raster manipulation at a rate of up to 16 pixels per microsecond - a parts cost of less than $2,000. Eventually, this graphic system will be integrated with a Z-8000 host processor, an Ethernet connection, and optional file storage to form a personal computer system for standalone or network use. The entire system has been nick-named SUN (an acronym for Stanford University Network, but also reflecting a nice property of the West Coast). A mailing list has been established for prospective users of the graphics and the personal computer system. If you wish to manipulate this list, send me a note or edit the file SUN.DIS[1,AVB] yourself. The graphic system is described in detail in the files G.TEX[1,AVB] or G.XGP[1,AVB], respectively. This document, including some figures that cannot be printed on the XGP, has already been distributed to some individuals. See me in person for figures or hardcopies. Comments are solicitated on all aspects of the system, including functionality, definition, design, engineering, and presentation. Specifically, I would like to collect input on the following: - the nature of applications - speed and resolution - color versus monochrome - the choice of video monitors - green versus white phosphor - interfaces at all levels - to host-processors - to the network - the choice of the Intel Multibus as backplane standard. The current state of the project is as follows. The first monochrome frame buffer has been wirewrapped (courtesy of Zilog Corporation) and should be working shortly. The graphics processor has been designed, it is being readied for wire-wrapping. The color frame buffer is not designed yet, and all other parts of the system are in the state of definition. The Motorola monitor was running with a test-pattern generator for several months at SLAC. The current schedule for milestones is as follows: July 1: Monochrome frame buffer working with Motorola Monitor. This will give us the first "real" picture of the visual qualities of the Motorola Monitor. The frame buffer will interface to a Z-80, thus we will not be able yet to judge the speed of the graphics system. Aug. 1: Graphics Processor working with the monochrome frame buffer. This system will demonstrate the entire design, including its performance. Fall: Given that we are satisfied, we will then layout printed-circuit boards for the graphics and produce them in a small assembly line. Initial users of the graphics system will be the VLSI and the TEX project. Prospective users include the Music people, Robotics, Design Automation, and maybe the entire department. Please send me your comments or questions. Sincerely Yours - Andy Bechtolsheim -

PDheader:1979-06-25 22:57:00-07:00 be1f3833ae36e4745bb435865756eb5a ∂25-Jun-79 2257 ME LLL drive fixed To: LES, JBR, LCW, HWC, TM, ROB

Al Barrett of Ampex came out and replaced head 17 of the LLL drive, and now it works. Unfortunately, we need to replace or fix the even head preamp of the same drive, since the orginal one in that drive was swapped into another drive a little while ago to keep the other drive going. So the LLL drive isn't up yet, but it will be as soon as it has a good even head preamp (Rob is taking care of that -- he may try to fix it, having gotten some good clues from Al Barrett).

PDheader:1979-07-18 12:43:00-07:00 94560caf0baf5b4fe763ec0f43fffba8 ∂18-Jul-79 1243 JLS Tapes:

------ KENNEDY:(Don Mandle, 324-4877) Drive: 9300-3 125 in/seg, dual density $6800 Formater: 1219-3 $2850 ------ $9650 AMPEX: (Tom McQuillan, (408)/(56) 733-2900) Drive: TME 75 in/seg, dual density $7025 (with formater?)

PDheader:1979-10-04 05:10:00-07:00 1031024ec7209f19638b7fb088a10167 ∂04-Oct-79 0510 RICK.RASHID at CMU-10A Ethernet Date: 4 Oct 1979 0803-EDT From: RICK.RASHID at CMU-10A Subject: Ethernet To: ROB at SU-AI

Robert Poor, I am the CMU faculty member in charge of efforts to integrate the Altos and Dover into our environment. Currently we are working on connecting both our KL-10 and VAXen to the net. We alreay have our 10 set up to transmit and receive on the Ether. The trick we have used is to buy an 11/34, hook it up to the EtherNet, and then connect the 11/34 to the 10 by means of a DA28F to DA28C linkup. Software for the DA28C (the 10 side of the link) was written in macro for CMU's version of TOPS-10 and should be appropriate for any vanilla TOPS system. I dodn't know if SU-AI could incorporate it into the Stanford version of TOPS, though. The 11 code is simple and is written in BLISS-11. We are currently only accessing the net via one 10 job and other jobs gain access via IPCF requests to the controlling job. In addition, one of our graduate students is working on a text file to press file conversion program and on a printing spooler for the Dover. Our VAXen are running version 7 of Unix. Work is just now starting on linking them into the net. We are working with the Univ. of Rochester on the VAX software. If you need any more information, just send a message to Rashid@cmu-10a. -Rick

PDheader:1979-10-04 08:07:00-07:00 a78ec2e3aecedf7077b39572f6867c04 ∂04-Oct-79 0807 ROB Ethernet info To: LES at SU-AI, ME at SU-AI, REG at SU-AI

PDheader:1979-10-09 21:32:00-07:00 c722253beffc641a04ef516767654272 ∂09-Oct-79 2132 CSD.BOSACK at SU-SCORE Re: a question

The problems you describe all were fixed in some past rev of the KL. The traps problem happens most often when an interrupt or another trap or page fault happens before the trap in question. It can happen that one of the trap conditions is just plain lost (dropping one on the ground once in a while can look like one happening 'late' - the same condition tends to reoccur). The ECO changes the way the trap flag stack is popped - it now ors up from the bottom. IO page fail for no apparent good reason was fixed (several disjoint bugs) in both the uCode and the PI system. In short, Yes, your problems are familiar and believed to be fixed. I think the current KL rev is around 12, although I think rev 12 was just one of my RH20 bugs being fixed. I suggest you get the ECO sheets from DEC and we can go over which fix goes with what bug. Maybe a wholesale upgrade to the current rev would be easier, but with the greater downtime risk of a moby change. ------- ME - Rob, can you get (or do we already have) the ECO sheets? (Maybe you're already getting them?) Sounds like we should definitely get some new Revs, although how many remains to be seen.

PDheader:1979-10-09 23:08:00-07:00 60f49ec35ef4aaf94db395317b49a838 ∂09-Oct-79 2308 ME KL-10 hardware bugs and Revs To: ROB, LES, RPG

PDheader:1980-01-17 16:33:00-08:00 7381aac9b17bf5f8f9dc7490cae6c7e9 ∂17-Jan-80 1633 JMC Suppes offer for KA.

He offers $20,000 for KA and BBN mapper. The mapper ain't ours, but I suppose he has enough Government support so that ARPA would let him use it. I didn't accept his offer, and an opinion that the KA is worth more is tenable. Perhaps we should hold out for the $25K in commemoration of the divorce settlement. Feigenbaum's money for displays is burning a hole in his pocket, so I am now inclined to act sooner than June on displays. Who is the relevant person to talk to at Grinnell and his co-ordinates, and when and at what number is the best time to talk to Panofsky?

PDheader:1980-01-17 16:40:00-08:00 e0af18ef193ea77723b092eafdfcc589 ∂17-Jan-80 1640 JMC

Would you talk to Suppes? ------ Yes. And perhaps to American Used Computer Corp. But remember that we need the KA until the Canon or some other XGP replacement is proven.

PDheader:1980-01-20 14:00:00-08:00 ce634509026112d8de209b3238bb8ef2 ∂20-Jan-80 1400 Mark.Sherman at CMU-10A Datacomputer going away Date: Sunday, 20 January 1980 1658-EST From: Mark.Sherman at CMU-10A Subject: Datacomputer going away To: S1 @ SU-AI Message-ID: <20Jan80 165858 MS40@CMU-10A>

As per the following notice, the Datacomputer facility at CCA is going away. All those with files stored there should retrieve them in the next couple of weeks and move them to mag tape (assuming you want to still save them). While some people are making efforts to keep the Datacomputer on net, I don't think they will succeed. The S1 root is named, appropriately, "S1". -Mark Sherman Date: 20 JAN 1980 1124-EST From: JZS at CCA (Joanne Z. Sattley) Subject: Scheduled termination of Datacomputer Service I have been requested by ARPA to distribute the enclosed message. ***** begin enclosure To All Datacomputer Users: The TBM-based Datacomputer has become too expensive to operate. We are faced with a combination of increasing maintenance costs and a funding deficit which was created when a major user community's need for the service ended. The Ampex TBM hardware is obsolete, and no compatible follow-on product is planned. Therefore, with much regret, we have decided that Datacomputer service on the ARPANET must be terminated. We want to make the transition off the Datacomputer as painless as possible for existing users. On the other hand, we want to phase out the service as quickly as possible to avoid needless expense. Our plan is to stop accepting any additional files for storage, effective immediately. The Datacomputer will continue to operate for retrieval only during the transition period. We would like users to retrieve their private files by March 15, 1980. DFTP public files will be saved automatically by the CCA staff. We apologize for any problems which this decision causes you. If there is something specific we can do to ease the transition, we will do our best to help. Send a message to JZS@CCA if you have questions or need assistance. Bill Carlson Program Manager DARPA/IPTO ***** end enclosure It has been a genuine pleasure working with all of you. For the Datacomputer Staff, -- Joanne

PDheader:1980-01-21 02:05:00-08:00 887ff773144b1726231d3c68c0ec2545 ∂21-Jan-80 0205 ME via SU-TIP magtape drives

This is to inform you that, for the time being, there are no Dart backup dumps being done because the magtape drives don't work. Drive A has intermittent problems at the end of the tape, and can almost be used, except that it requires a large amount of human intervention to get around these end-of-tape problems. Drive B, which has recently been suspected of needing alignment, has just dropped dead, with a troubled tape of the last dump very near the end of the tape, with the drive refusing to move the tape in any mode. Rob having just gone on vacation, there is no one to fix the drives at all. Perhaps you will figure out what to do. I might suggest hiring Ted to look at the drives, if he is available. Rob seemed to think there was no need for aligning drive B, but it seems to need something, probably alignment. Perhaps a decent burial. I'm sick of worrying about useless hardware when I have enormous amounts of software to work on.

PDheader:1980-01-25 17:21:00-08:00 f254c2c002c34805be188285b85c4574 ∂25-Jan-80 1721 REG Status of tapes, updated To: ROY CC: LES, REG

The data control now works. Tape unit A reads and writes tapes. Tape unit B reads and writes tapes. Tape unit B can read the tapes written by unit A. Tape unit A can NOT read the tapes written by unit B.

PDheader:1980-03-22 08:13:00-08:00 23772ed97a9a37a45918ccf29affd223 ∂22-Mar-80 0813 ADMIN.MO at SU-SCORE (Maurice Bizzarri) Possible tape drive source for SAIL.... Date: 22 Mar 1980 0808-PST From: ADMIN.MO at SU-SCORE (Maurice Bizzarri) Subject: Possible tape drive source for SAIL.... To: les at SU-AI, admin.gorin at SU-SCORE

If you are interested, I found out last night that DARCOM, which has a KA-10 at SRI, is probably getting a 2040 or a 2060 to replace it. They have these CALCOMP tape drives and disks they got less than a year ago brand new, and they are IBM channel compatible (through an SA-10). Gary McGreal at SRI doesnt know exactly what is going to happen to the equipment, but if there is any interest in the tape drives he can refer us to the correct people at DARCOM... Mo -------

PDheader:1980-03-29 16:20:00-08:00 85f20919abbc9e4dc49a26cd5f1693e8 ∂29-Mar-80 1620 LES SRI's tape drives To: REG

I assume that the tape drives Mo mentioned come with a controller. Indeed, they could be bought for SAIL. We would have to find some unrestricted funds (e.g. JMC's) to buy them with, then repay the money by the depreciation allocation from the cost center. Have you considered the alternative of putting tape stations on the Ethernet? As you know, that offers somewhat more flexibility, not to mention the opportunity to start another engineering project.

PDheader:1980-06-03 20:58:00-07:00 064986d511fb55c595e8e95416d2b83d ∂03-Jun-80 2058 LES Flakey TTY lines To: ROY, ROB

I dialed in a few minutes ago via 0084 and got a seemingly solid line -- I didn't pay attention to the line number. After a few minutes work, it suddenly disconnected. I called back again on 0084 and got 0085, which was so flakey that it was unusable. Nevertheless, I was able to determine that my former job was sitting there detached. I hung up and dialed 0086, which answered but wouldn't synchronize. Tried again on 0088 and got an extremely weak tone that wouldn't work at all. Finally, in desparation, I dialed 0081 and got 0083, which is supposed to be flakey, but is far better than any of the other lines. Perhaps my modem is developing strange tastes.

PDheader:1980-06-28 00:26:00-07:00 caf4a0fbcbe256d70ef90fdb9a0d10d4 ∂28-Jun-80 0026 ADMIN.JKS at SU-SCORE Ethernet Interface Date: 28 Jun 1980 0023-PDT From: ADMIN.JKS at SU-SCORE Subject: Ethernet Interface To: [SU-SCORE]<SU-NET>SU-NET.DIS.3: cc: Swinehart at PARC-MAXC, Boggs at PARC-MAXC

Andy and I have completed debugging of a prototype Ethernet interface. This 30 chip TTL device essentially performs the function of a UART -- transforming Ethernet serial data into a 16-bit parallel port with appropriate handshaking. To incorporate Ethernet into a particular machine architecture it is now only necessary to build a controller from the host machine bus to the 16-bit port, rather than to the Ethernet itself. The low-level details of Ethernet communication are now less of a barrier to new interface designs. Our next step is to build a microprocessor based controller that provides more standard interface architectures such as GPIB, 50Kb serial link, and MultiBus. Many of you are probably interested in adapting this design to fit your particular interface requirements. We hope to see interfaces built for the PDP-11 Unibus, IBM Series-1, and PDP-10 and -20 I/O bus. In addition, Ethernet servers such as terminal concentrators (TIPs), gateways, repeaters and printing servers should be built. Full specifications will be available shortly -- send me a message if you'd like a copy. Estimated prices: (1) lowest price development system: $250,000 to $300,000 (2) reasonable 2 processor system: $500,000 to $600,000 (3) additional CPU's (including 8 to 16 megabytes more memory each): about $200,000 It seems that it is fairly high priced for just fiddling around with, but I am not sure how the performance on a uniprocessor would stakc up against other machines. I would assume the overall speed would be twice a Symbolics 3600 or better.

PDheader:1980-07-09 10:15:00-07:00 1788fd3a16b63471bc9a6c108fcbda86 ∂09-Jul-80 1015 LES Keyboard krock To: REG at SU-AI, ME at SU-AI CC: JMC at SU-AI, MO at SU-AI, ROY at SU-AI, ROB at SU-AI

I do not understand why you have permuted the numeric keys or why you did so without public discussion. The SAIL configaration is clearly superior to that of conventional typewriters. The argument that it should be stadardized the older way to simplify training overlooks the point that almost no one touch-types numbers. This issue was carefully examined and resolved locally long ago (correctly, I believe). From your viewpoint, I would guess that all calculators have been improperly designed in that they put "0" next to "1", contrary to standards established by The Telephone Company. If you believed that a change was needed for some reason, you should have had the guts to discuss it openly, rather than undertaking a sneaky coup. The only remarks about this that I heard came from ME, just after he returned from that hotbed of conservatism, MIT. I am aware of no concrete proposals or discussions of this issue. I am disappointed both in what has been done and the manner of execution. I recommend that you restore the keyboards to the prior configuration, then make a proposal and defend it. FCC is allegedly a two year old Dallas-based public corporation. They have a multiprocessor system with a number of different internal busses with transfer rates up to 48 megabytes/s and with uniform access times to memory. Their product is aimed at realtime and scientific computing applications. First delivery was December '84 and they allegedly have delivered about a half-dozen systems, including one each to Purdue U. and NASA-Langley. Their system currently uses 32032 processors. (They claim that no one else, including Sequent, is delivering 32032-based systems yet.) They claim that they will also support the 68020 and will demonstrate a mixed system (32032 & and 68020) this summer. They have three operating systems: Unix System V, MMOS, and Interim. We didn't discuss that latter two much. They also have "concurrent" versions of C and Fortran, effected by adding preprocessors to standard compilers. They expect to have a concurrent ADA later this year. Caron said that of the machines we are considering so far, they were most like the Encore. I requested literature. They stand ready to tell us more. One of the founders will be in this area next month and could be programmed to visit.

PDheader:1984-10-22 09:19:00-07:00 6a2a9d029e449be3b7ea51c6dd615627 ∂22-Oct-84 0919 FEIGENBAUM@SUMEX-AIM.ARPA [Bruce Delagi <DELAGI@SUMEX-AIM.ARPA>: [Andy Freeman <FREEMAN@SUMEX-AIM.ARPA>: [Glenn Adams <glenn@ll-xn.ARPA>: Re: Received: from SUMEX-AIM.ARPA by SU-AI.ARPA with TCP; 22 Oct 84 09:19:02 PDT Date: Mon 22 Oct 84 09:22:19-PDT From: Edward Feigenbaum <FEIGENBAUM@SUMEX-AIM.ARPA> Subject: [Bruce Delagi <DELAGI@SUMEX-AIM.ARPA>: [Andy Freeman <FREEMAN@SUMEX-AIM.ARPA>: [Glenn Adams <glenn@ll-xn.ARPA>: Re: Tightly-coupled multiprocessor UNIX]]] To: jmc@SU-AI.ARPA

Mail-From: DELAGI created at 20-Oct-84 07:36:24 Date: Sat 20 Oct 84 07:36:24-PDT From: Bruce Delagi <DELAGI@SUMEX-AIM.ARPA> Subject: [Andy Freeman <FREEMAN@SUMEX-AIM.ARPA>: [Glenn Adams <glenn@ll-xn.ARPA>: Re: Tightly-coupled multiprocessor UNIX]] To: nii@SUMEX-AIM.ARPA, brown@SUMEX-AIM.ARPA, feigenbaum@SUMEX-AIM.ARPA, davies@SUMEX-AIM.ARPA, andy@SU-SCORE.ARPA, rpg@SU-AI.ARPA Based on my work experience with Dave Rogers and Dave Schanin (the guy I imagine who is responsible for the ideas in Encore's machine) the Sequent machine seems worth knowing more about. (I mean no derogation of Schanin: it's just that Rogers -- at least when I knew them both -- was the deeper technically, ranged over a broader set of design issues, and had better human interaction skills. If these qualities showed up in the design of the Sequent, it's a machine we should understand better)......./bruce --------------- Sort (ms) 178.7 140.8 158.3 85.2 30.9 (ratio) (0.79) (1.00) (0.89) (1.65) (4.56) Mail-From: FREEMAN created at 17-Oct-84 14:57:51 Date: Wed 17 Oct 84 14:57:50-PDT From: Andy Freeman <FREEMAN@SUMEX-AIM.ARPA> Subject: [Glenn Adams <glenn@ll-xn.ARPA>: Re: Tightly-coupled multiprocessor UNIX] To: delagi@SUMEX-AIM.ARPA Slosh (ms) 123.2 112.9 99.7 69.0 21.7 (ratio) (0.92) (1.00) (1.13) (1.64) (5.20) Properly programmed, this may be a decent multi-processor simulator. ------- ------- -------

PDheader:1984-10-29 13:48:00-08:00 714feb335bcfcf68c2e36d7fa8515835 ∂29-Oct-84 1348 RPG Subcontracts

There was some discussion among the ARPA folks and me about Stanford subcontracting the Common Lisp work at Stanford to Lucid. This would involve the blue pages work and the yellow pages work. I think we're in a pretty good position to take that on. How shall we proceed? We have been looking at the various pieces of multiprocessor hardware, and we've pretty-much decided that the way to do the HEP is with a byte-code compiler and interpreter. That is, we will compile Lisp down to some abstract machine code, and we will hand-code an interpreter for that language in HEP assembly language (or PASCAL). The interpreter will run in the program memory. Unless we do this, it will take a very long time to get a Lisp running.

PDheader:1984-11-12 15:22:00-08:00 4a7bdd87f4fc15d6a580955cbc0136fd ∂12-Nov-84 1522 RPG Multiprocessors To: JMC@SU-AI.ARPA, JJW@SU-AI.ARPA, cheriton@SU-PESCADERO.ARPA CC: rpg-q@SU-AI.ARPA

The following companies are `in the running' as far as I'm concerned: Denelcor Synpase Encore Symbolics BBN Sequent Let me discuss the pros and cons of each: Denelcor: We all looked at the HEP and have a pretty good idea of what it can do. Pros: 1. Friendly and co-operative staff: they believe in what we're doing; 2. cute architecture; and 3. they are giving DARPA a `good deal.' Cons: 1. The hardware implementation of the architecture make it very difficult to do a cheap, fast Lisp, and 2. Lucid is less interested in doing it than some of the others, because it is probably not a `typical' multiprocessor, such as we might see in the next 10 years in common use. Encore: I talked at length to Gordon Bell, Ike Nasso, Steve Amarant, and Henry Weigart from Encore in Massachusetts. I think that of the bunch-of-micros-on-a-bus machines, Encore's has the highest potential in terms of being relatively high performance and expandable in the near term to more processors than either of Synpase or Sequent. This is because they have a 100 megabyte per second bus, with separate address, data, interrupt, and parity paths. There are 4 different board types in the machine: processor, memory, IO, and Ethernet. I saw a physical example of each board on the workbench, with logic analyzers hooked up to each. That is, they have gotten board back and are now debugging them. They use NS32032's, and appear to have them; they are running at 6 Mhz instead of the nominal 10 mhz. The memories and bus support 32-bit addresses, and the memories are 4-way interleaved Each processor board has 2 cpu's and a shared cache between them. The cache is 32kbytes. The memories support test-and-set. In the future each board will have 4 processors, and National is going to a 2.5x faster chip in a year or so. They have a software group that it porting Berkeley 4.2 to it. They have several functional emulators (made up of NS32016 boards) on which they are doing software development. They claim that the first 2-processor (1 cpu board) machine will be put together this week, and having seen the state of the hardware as of last week, I think it might be possible. In addition, they claim that they could get us a prototype machine mext April-June. In order to do a Lisp on that machine, Lucid would have to port our Sun Lisp to the 32032. I think this could be done using a uni-processor between now and next summer. With the uncertainties of the DARPA contract, the Encore time schedule might not be a problem. Pros: 1. 32032 is a nice machine; 2. fast bus makes for future expansion; 3. Bell has a strong interest in the project and has been associated with multiprocessors for many years; 4. demand paging; 5. Unix; and 6. Lucid would want to do it. Cons: 1. the Encore schedule could slip; and 2. Encore may collapse, though I don't see this right away. Synapse: I agree with everything that Cheriton said, except for one. That is, I think that Synapse is a little closer to the brink than he thinks. Synapse shipped machines too early a while back, and the investors have slapped their wrists. They may be on a prove-yourself-or-die cycle. I want to emphasize that Synapse will have *no* interest in helping us at all unless we convince them that there is a market they could attack with a parallel Lisp. Otherwise, I think, they will proceed in the database area until their investors are happy. Pros: 1. 68000 Lisp is a simple port of Lucid's Sun Lisp; 2. basically a good architecture; 3. they will accept a purchase order today; 4. reliable hardware; 5. nearby; and 6. Lucid would want to do it. Cons: 1. They will not help us in any way unless we change their marketting strategy to include AI; 2. operating system is not something Lucid wants to do; and 3. perhaps they're not long for this world also. BBN: BBN appears to be about to get strategic computing money to put a Common Lisp on the Butterfly along with a multiprocessing version of it. Poor lads, they thought I (rpg) was going to do the Common Lisp for them for free. I talked to a panic-stricken Don Allen last week, and he strongly suggests we use the Butterfly. He and Randy Rettberg, a manager, are willing to fly out to Stanford for a show-and-tell. They hope that we select the Butterfly and get DARPA to pay for Lucid to put Common Lisp and the low-level primitives for Qlambda on the Butterfly. They would then take the Lisp and the primitives and do something different from Qlambda for DARPA. It sounds like there are a few things wrong with their suggestion: 1), they are suggesting that DARPA pay twice for some of the work, namely the multiprocessing primitives, and 2), I think DARPA is more nearly sold on Qlambda than it would be on a parallel Lisp to be specified later. Nevertheless, the Butterfly is a 68000-based machine and it exists. Pros: 1. 68000-based machine; 2. Sun Lisp easy to port; 3. butterfly switch is not so bad; 4. machine exists; 5. BBN *wants* to co-operate; and 6. Lucid is not against doing this. Cons: 1. Unknown operating system; and 2. the switching network may need to be micro-coded to handle a synchronization primitive. Let me elaborate on this last point. Each cpu board has a micro-codable memory-watcher. This memory-watcher implements the butterfly network, and there is more control store in which to do other things. For example, references to certain memory locations can be interpreted as other sorts of instructions by this watcher. I'm not sure how much programming of it must be done. Symbolics: I talked to them about multi-processors, and I was surprised when they said that they had proposed a multi-processor to Arvind, which would be a memory-bus or multi-ported memory scheme. Another possibility would be a gigabyte serial port amongst some 3600's. Symbolics volunteered to send out someone to Stanford to chat about it. Pros: 1. The 3600 already has a Lisp. Cons: 1. I'm not at all sure whether they could deliver this in a reasonable length of time. Sequent: They are left to look at. They have delivered, and use 32032's. Gordon Bell says that except for a slower bus, they are similar to the Encore machine. Tomorrow I am starting my sequence of inquiries which ought to land us some information soon. -rpg-

PDheader:1984-12-05 00:14:00-08:00 3132ed8ab299b70c75243d01e1646e7e ∂05-Dec-84 0014 JMC [Bruce Delagi <DELAGI@SUMEX-AIM.ARPA>: [Andy Freeman <FREEMAN@SUMEX-AIM.ARPA>: [Glenn Adams <glenn@ll-xn.ARPA>: Re:

PDheader:1984-12-05 00:16:00-08:00 a702559befdedbfcdb5a0ac5ecdceed8 ∂05-Dec-84 0016 JMC

PDheader:1984-12-05 13:47:00-08:00 686f1ebaa84ae02dd8f5ef703dbcee05 ∂05-Dec-84 1347 imagen!les@su-shasta.arpa HEP

Thanks, Stuart Cracraft ------- From: shasta!JMC@SU-AI.ARPA Received: from SU-AI.ARPA by Shasta with TCP; Wed, 5 Dec 84 00:18:05 pst Date: 05 Dec 84 0015 PST From: John McCarthy <shasta!JMC@SU-AI.ARPA> To: LES@SU-AI.ARPA

PDheader:1984-12-30 18:35:00-08:00 a54e33ca25e65a3b2eb15f440113d819 ∂30-Dec-84 1835 fateman%ucbdali@Berkeley info on Elxsi multiprocessors.. Received: from UCB-VAX.ARPA by SU-AI.ARPA with TCP; 30 Dec 84 18:35:42 PST Received: from ucbdali.ARPA by UCB-VAX.ARPA (4.24/4.40) id AA07360; Sun, 30 Dec 84 18:28:38 pst Received: by ucbdali.ARPA (4.24/4.40) id AA11972; Sun, 30 Dec 84 18:34:56 pst Date: Sun, 30 Dec 84 18:34:56 pst From: fateman%ucbdali@Berkeley (Richard Fateman) Message-Id: <8412310234.AA11972@ucbdali.ARPA> To: clt@su-ai Subject: info on Elxsi multiprocessors..

From taylor@ucbkim Sat Dec 29 21:34:48 1984 Received: from ucbkim.ARPA by ucbdali.ARPA (4.24/4.40) id AA06331; Sat, 29 Dec 84 21:34:43 pst Received: by ucbkim.ARPA (4.24/4.27) id AA15492; Sat, 29 Dec 84 21:37:14 pst Date: Sat, 29 Dec 84 21:37:14 pst From: taylor@ucbkim (George Taylor) Message-Id: <8412300537.AA15492@ucbkim.ARPA> To: fateman@ucbdali, taylor@ernie Subject: Re: Elxsi Maintenance 1 full time hardware 1 full time software

PDheader:1985-01-02 16:08:00-08:00 9eae71c6a674384d130f8f841cfe61df ∂02-Jan-85 1608 CLT info on Elxsi via fateman To: RPG, les-here, JMC, JJW

Space Current or equivalent

PDheader:1985-02-04 21:16:00-08:00 7dc77957dfdc3206995049d1f01b6011 ∂04-Feb-85 2116 JJW Sequent To: JMC@SU-AI.ARPA, LES@SU-AI.ARPA, RPG@SU-AI.ARPA, CLT@SU-AI.ARPA,

cheriton@SU-PESCADERO.ARPA I don't think anyone has given a report of our trip to Sequent yet. They are definitely a real company, and are producing and selling their machines. The hardware is approximately the same as Synapse and Encore, but is available (unlike Encore at present) and runs Unix (unlike Synapse, whose operating system isn't as good), though Synapse has a more sophisticated cache and both of the others have a faster memory bus. Sequent seems as interested in our work as Encore, much more so than Synapse. It's my understanding that we intend to propose buying two Sequent machines to ARPA, since Synapse, being more expensive, might reduce the chance of the proposal being accepted. We were given a revised price quote that is about $290K for 12 processors and 16Mbytes, and $310K for 8 processors and 24Mbytes. (In each case this is the maximum amount of memory that can be put on the machine.) The exact figures are on a sheet that I gave to Rutie, and I asked her to make copies for JMC and LES. Note that these prices include $65K for a Dynix source license. I got the impression that they would actually rather not sell us the sources, since they are trying to keep their multiprocessor improvements to Unix confidential. If we could run without operating system changes, or if we could get them to make any necessary changes before shipping it to us, we would save a substantial amount. The only changes to the kernel that I could think of were for IP/TCP (Stanford subnet routing). Jeff Mogul has told me that we may be able to run without changes because of a feature in our Ethernet gateways, and eventually there will be a subnet routing standard that Sequent may be convinced into supporting. Our discussions about implementing Qlambda did not uncover any necessary changes in the operating system. Shared memory between processes is now available as a system-wide resource (i.e. you could have one multiprocessor Lisp job running at any time), and by September they will have a more flexible version of shared segments so that several users can do it simultaneously. When deciding how many processors and how much memory to get, keep in mind that the virtual address space of any process is at most 16Mbytes (24 bits). A single user running Qlambda might benefit more from 12 processors vs. 8 than from 24Mbytes vs. 16, since the extra memory couldn't be part of his address space. (It would allow the entire user program as well as the operating system to reside in memory. I forgot to ask how much memory Dynix takes up.) The prices don't include any terminals, nor an Ethernet transceiver. We might want to buy only one terminal for an operator console, and use Telnet as the primary means of access. (Similarly for the machine to be used at Lucid, I assume.) I forget whether their quote was for a 1/4" cartridge tape or a 1/2" tape, but we might want to change that as well. These are all rather small items however.

PDheader:1985-04-29 10:06:00-07:00 fad9db33878ef27bd7bb297e5a85f90d ∂29-Apr-85 1006 RA Ray Caron

Ray Caron from Flexible Computer Corp. called (415) 593 7044; referal from DARPA. John ------- -----------------

PDheader:1985-04-29 10:11:00-07:00 0c171ae6786a4d214fb25cd90216f167 ∂29-Apr-85 1011 JMC

Please find out what he has to sell.

PDheader:1985-04-29 11:06:00-07:00 6f52552cafc1f2885b4d5a738ab3f8fa ∂29-Apr-85 1106 LES FCC To: JMC, RPG, CLT

PDheader:1985-05-01 15:06:00-07:00 682d61d83d70d63b1868ade539f6ff58 ∂01-May-85 1506 JMC To: CLT, RPG, LES

Motorola proposal to DARPA in my parallel machine file.

PDheader:1985-05-30 13:35:00-07:00 a2bf5cd44d8a2d2fefb819c767b097a2 ∂30-May-85 1335 fateman%ucbdali@Berkeley Re: DARPA feedback on Qlisp Received: from UCB-VAX.ARPA by SU-AI.ARPA with TCP; 30 May 85 13:35:47 PDT Received: from ucbdali.ARPA by UCB-VAX.ARPA (4.24/4.46) id AA09455; Thu, 30 May 85 13:32:39 pdt Received: by ucbdali.ARPA (4.24/4.46) id AA22012; Thu, 30 May 85 13:35:50 pdt Date: Thu, 30 May 85 13:35:50 pdt From: fateman%ucbdali@Berkeley (Richard Fateman) Message-Id: <8505302035.AA22012@ucbdali.ARPA> To: LES@SU-AI.ARPA, RPG@SU-AI.ARPA, Fateman@Berkeley Subject: Re: DARPA feedback on Qlisp Cc: clt@su-ai, fateman%ucbdali@Berkeley

I suggest that the following company be among the prospective vendors: Alliant Computer Systems Corp 42 Nagog Park Acton, Mass 01720 (617) 263-9110

PDheader:1985-06-07 15:50:00-07:00 480e180eb08cf2484b083055c713d85b ∂07-Jun-85 1550 TOB

Lisp machines I am inquiring about your interest in sharing in a deal for Symbolics Lisp machines. I have reserved three pairs of Lisp machines on a deal that Tom Rindfleisch was originally arranging. The arrangements are: one used 3600, 8 Megabytes of memory, about 150 megabytes of disk one new 3640, 8 megabytes of memory, about 150 megabytes of disk machines are $90k per pair, $45k each. If you are interested, please act fast. I can keep the option open only for a week. Also, terms are payment on delivery. It is assumed that they will go onto maintenance contract immediately.

PDheader:1985-06-13 10:38:00-07:00 a341018db591415e92e7e86f2079aa22 ∂13-Jun-85 1038 RINDFLEISCH@SUMEX-AIM.ARPA Instructional Lisp Machines Received: from SUMEX-AIM.ARPA by SU-AI.ARPA with TCP; 13 Jun 85 10:37:42 PDT Date: Thu 13 Jun 85 10:36:58-PDT From: T. C. Rindfleisch <Rindfleisch@SUMEX-AIM.ARPA> Subject: Instructional Lisp Machines To: Nilsson@SU-SCORE.ARPA, Feigenbaum@SUMEX-AIM.ARPA, Buchanan@SUMEX-AIM.ARPA,

Genesereth@SUMEX-AIM.ARPA, JMC@SU-AI.ARPA, LES@SU-AI.ARPA, TOB@SU-AI.ARPA cc: Rindfleisch@SUMEX-AIM.ARPA, Bosack@SU-SCORE.ARPA, Yeager@SUMEX-AIM.ARPA Friends, as most of you know, the KSL is about to make a large purchase of Lisp workstations for research. As a fallout of negotiating with TI, Symbolics, and Xerox, we have offers of very advantageous discounts for a quantity of machines beyond what the KSL can buy (see below). Noting that the department has already invested in 2060, (micro) VAX, Macintosh, and other hardware resources for various non-AI parts of CSD, Nils is considering buying $100-200K worth of workstations for AI instructional purposes. Nils wants to discuss this possibility among the AI faculty as soon as possible. Can we meet for an hour early next week? Nils is available afternoons after 1:30 on Monday through Wednesday. Please let me know you schedule preferences. Tom R. SUMMARY OF CONFIDENTIAL PRICING: VENDOR # MACHINES PRICE COMMENTS TI 25 $22K/mach Package deal, 25 mach for $540K. DARPA/ISI price $54K. Symbolics 6 or 8 45K/mach Package deal, 6/8 mach for $270/360K. DARPA/ISI price $69K Xerox Variable 14-22K/mach Purchase quantity scalable. List price on 6085 is $18-19K. MACHINE CONFIGURATIONS: TI Explorer with 8 MB memory, 2 x 140 MB disk, and Ethernet Symbolics 50:50 mix of 3640's and 3600's with 8 MB memory, 169 MB disk, and Ethernet. The 8 machine deal would also include one 450 MB disk. Xerox $14K machine = newly announced 6085 with 3.7 MB memory, 80 MB disk, 8K control store, and Ethernet. Also, IBM PC-compatible co-machine internally. (Xerox guarantees this machine will perform as well as the current 1109 or will replace them with the same number of 1109's). $22K machine = 1109 with 3.5 MB memory, 80 MB disk, 12K control store, and Ethernet. PS: To date, TOB has expressed research interest in piggy backing on this buy but JMC/LES have not. If this changes, I need to know by early next week. The KSL will buy a mix of about $800K of this hardware. -------

PDheader:1985-12-06 11:20:00-08:00 862e711baf5ae1bff5310244fd8e1bd7 ∂06-Dec-85 1120 CONTRERAS@SU-SCORE.ARPA message Received: from SU-SCORE.ARPA by SU-AI.ARPA with TCP; 6 Dec 85 11:20:51 PST Date: Fri 6 Dec 85 11:18:05-PST From: Tina Contreras <CONTRERAS@SU-SCORE.ARPA> Subject: message To: Les@SU-AI.ARPA Message-ID: <12165003143.20.CONTRERAS@SU-SCORE.ARPA>

Rich Stam of Convex phoned, regarding a Shared Vector super on campus funded partly by Dean Gibbons. Please call. 408 275-0844. Tina -------

PDheader:1985-12-06 13:52:00-08:00 f4e463f3f0f081bc3214f4b16eb07035 ∂06-Dec-85 1352 BOSACK@SU-SCORE.ARPA [sequent!dave@decwrl.DEC.COM (Dave Rodgers): Re: Path test] Received: from SU-SCORE.ARPA by SU-AI.ARPA with TCP; 6 Dec 85 13:52:09 PST Date: Fri 6 Dec 85 13:45:07-PST From: Len Bosack <BOSACK@SU-SCORE.ARPA> Subject: [sequent!dave@decwrl.DEC.COM (Dave Rodgers): Re: Path test] To: LES@SU-AI.ARPA Message-ID: <12165029910.19.BOSACK@SU-SCORE.ARPA>

[Dave Rodgers is an old friend who is also Engineering VP for Sequent.] --------------- Date: Fri, 6 Dec 85 11:29:24 pst From: sequent!dave@decwrl.DEC.COM (Dave Rodgers) To: BOSACK@SU-SCORE.ARPA Subject: Re: Path test I'll let you know when things happen. We assume you will have a 32032 or 32332 based version of your machine sometime soon. While the 32016s would work for a proof-of-concept, the faster PEs would make the development effort go faster. We have been shipping 10 MHz 32032 based systems since June, 1985 but are just now completing the upgrades of older 32016 based systems to 32032 (at no charge). We are at work on a more powerful system. Dave -------

PDheader:1986-01-13 11:54:00-08:00 d4c108984b4b6b1266f2393b1d931953 ∂13-Jan-86 1154 JJW Machines for Qlisp To: JMC, LES, CLT, RPG

People in Cheriton's group (Berc, Stumm) tell me that a multiprocessor based on the Motorola 68020 would be much nicer than one based on the National 32032. They're currently designing one called Nebula, but it won't be available as a usable system for several years. They didn't know whether any commercial 68020 systems currently exist.

PDheader:1986-01-13 12:03:00-08:00 9f43ef83b892ca78a02db6c036a11e3c ∂13-Jan-86 1203 RPG Machines To: JJW, JMC, LES, CLT

I would agree. You can cut 3-4 months off of Lucid's time to implement Qlisp if we have a 68010 or 68020 based multi. -rpg-

PDheader:1986-01-14 10:49:00-08:00 4a488e2fbdf5beff5b2a75922620db49 ∂14-Jan-86 1049 RPG Alliant To: LES, JJW, CLT, JMC

I met with the Alliant folks yestersday, and I think their machine is possibly the best one to start with. It is somewhat expensive, but I think we can get up and running on it very fast. Here is a brief description: It has up to 8 68020-like processors that can operate concurrently on a shared address space. They are 68020-like in that they are made from custom CMOS gate arrays and operate about twice 68020 speeds. In addition they do vector operations and concurrency operations. These guys share 2 caches (4 machines each). Each cache is 64kbytes, and the cache to processor bandwidth is 376mbytes per second. The caches talk to up to 64mbytes of real memory at a rate of 188mbytes per second. The basic mode of operation is to do a CTART instruction which grabs n processors and starts them all with a common instruction stream. Then each stream can pick up tasks from a global queue. There is no messing around with an operating system to do this. They implement exactly the 68020 instruction set except for the 6 register compare and set, which would have been the right instruction for having concurrent processes do reference-count GC. The 68020 added that instruction after Alliant was done negotiating with Motorola. In addition, there are n 68012's (12 mhertz 68010's). These guy talk to users and devices. The disk drives are Eagles. The price is high, $270k to buy into the parallel version of the machine (the FX8). That gets you 1 68012, 1 68020, 1 cache, and 8mbytes (?) of memory. Then the prices are approximately: 49k per processor 40k for the second cache 40k per 8 mbytes of memory (!) This is high, but it's high-speed memory, and does some test and set operations. We could operate with a 4 processor machine for a while. It runs berkeley 4.2 and will switch to berkeley 4.3. We can get delivery in late spring, and they've talked to Squires about it. The nice part is we can probably get the uniprocessor Lisp going in less than a week on this thing, and it will probably run much faster than a 3600 in uniprocessor mode. Jack Test, who was the Alliant guy here yesterday, is willing to come back to talk to us at the end of this month. His best days are Jan 27, 28, Feb 3, 4, Jan 29, in decreasing order of ease. Let me know the date you'd like to get a presentation from him. -rpg-

PDheader:1986-01-14 11:14:00-08:00 9c668c19a336a2df3d9d3cf47b0fa687 ∂14-Jan-86 1114 JMC re: Alliant To: RPG, LES, JJW, CLT [In reply to message from RPG rcvd 14-Jan-86 10:49-PT.]

Unless there is a strong reason to do otherwise, I still want to wait until the contract at least clears DARPA.