15 May 1967


Quarterly Technical Letter Report 4

Covering the Period 8 February through 8 May 1967

Stanford Research Institute Project 5890









D. C. Engelbart



Contract NAS1-5904



Prepared for

National Aeronautics and Space Administration

Langley Research Center, Langley Station

Langley, Virginia 23365, Mail Stop 126



Copy No. 18


ARNAS QPR4, February-May 1967


1 (qpr4) Quarterly Progress Report No. 4, ARNAS Project, covering period 8 February through 8 May 1967


2 General


2a The planning and designing for the new multiconsole system has occupied most of the effort.


2b However, producing the first annual Interim Progress Report slipped over into the start of this quarter as a fairly significant effort, as well as the following:


2b1 We have developed separate Technical reports on COPE, the 3100 NLTS, and MOL940 (reports that we expect to have cleaned up for publication within several months), as well as on the 3100 FLTS (which we expect to redo and publish sometime early next Fall).


2c The additional ARPA-supplied funding for the multiconsole system became available for use--the contract having been executed on 8 May, 1967 as an extension of our prior RADC Management-Aid contract (our Project 5919).


2c1 We now have ordered our equipment (some of these orders had already been tentatively accepted by the suppliers). Details, including expected delivery dates, are given below.


3 3100-system developments--COPE and FLTS have had no additions or modifications during this period, but we have done some work on NLTS:


3a We have oriented our design approach to adding:


3al Pattern-Matcher Marker Setting feature, as described in the Interim Progress Report, Chapter IV, Section 4e (i.e. IPR IV-4e).


3a2 Generalized Hop Commands, as described in IPR IV-4g.


3b We have also placed in operation the user-action monitoring system, so that now a user cannot operate NLTS without mounting the monitor-record tape.


3b1 Our early analyses are interesting to us, but not conclusive in any respect.


3b2 We simply plotted out event-density and event-interval distributions--we expect to do considerably more, but this particular activity is currently pushed down on the priority list




ARNAS QPR4, February-May 1967


below finishing the above 3100-system changes, and also below the developmental needs for our initial 940 system.


4 New-Facility Hardware


4a The 940 Computer


4a1 There has been no change in the delivery schedule. Our computer is now in checkout at SDS in Los Angeles, and we expect delivery of the complete system about 1 June.


4b Bryant Disk File


4b1 An order has been placed through Chandler leasing for a Model 4061 Disk File (approximately 32 million-word capacity) and for a controller to interface to the 940. We expect delivery sometime in October.


4b1a The controller design is patterned after specifications developed at Berkeley.


4b2 We want to be able to use as much of Berkeley's disk-handling software as possible.


4b2a Thus we have been keeping in close touch both with Project Genie at Berkeley and with Bryant regarding the controller specification, and we will follow closely the progress of Bryant's design of the controller.


4c Line Printer


4c1 An order has been placed for the Potter HSP-3502 printer with 96 printing characters. We expect delivery in October.


4d Workstation System (See IPR, Figure 2, p. 29)


4d1 Display Generators


4d1a An order has been placed with Tasker Instruments for the first display-generator system (six displays) with delivery expected by 1 August.


4d1b An order will be placed for the second (identical) system in the next few weeks, with delivery in November.


4d2 Television System


4d2a For this system, we have selected and ordered 12 General Electric Model TE-21 cameras and Convac monitors, scheduled for




ARNAS QPR4, February-May 1967


delivery in June.


4d2b This is an 875-line system (in comparison with other cameras of higher scan-line rating it presented a sharper picture because it has less noise).


4d3 Interface and Controllers


4d3a Logical design is complete for the common control and display controller, and wiring will begin within a week.


4d3a1 We are using a logical-equation approach to the hardware design, and we are making extensive use of the on-line text system to handle and edit the equations.


4d3a2 The output phase of the text system has been adapted so that we can punch the equations on cards, which then are used directly as input to a program running on the SRI central computer (B5500) to do load checking and to produce wire lists.


4d3b The input-devices controller and the line-printer controller have been specified, and their design will begin in a few weeks.


4d3c Most essential components for the system have been ordered, and we expect to have at least the display portions checked out by the time the display-generator system arrives (about 1 August).


4d4 Remote Consoles


4d4a Keyboards for the consoles have been ordered from Friden. The first units will be delivered by the end of May.


4d4b A new binary keyset has been designed and we are experimenting with the first model.


4d4c A contract has been let for twenty of the new version of the mouse. The first of these will be received by the end of the month.


4d4d Within the next week, we will have 17-inch and 14-inch television monitors, and we will begin designing and experimenting with physical arrangements for the workstations.


5 New-Facility Software


5a MOL




ARNAS QPR4, February-May 1967


5a1 Programming


5a1a At the time of writing the Interim Progress Report (IPR), the MOL was approximately 50 percent coded and debugged. It was able at that time to parse an input stream but had no code-generation facilities.


5a1b The initial version of the MOL is now almost completed and can at this time compile small programs.


5a1c Future Work


5a1c1 All of the work on the MOL is being done on the APL 940. Because of hardware problems on that machine, work is progressing at a slow pace. At the current rate, it should take about two weeks to finish the initial version of the system.


5a1c2 New features and better optimization heuristics for code generation will be added as the need arises.


5a2 Documentation


5a2a A complete specification of the MOL was produced at the time of writing the IPR.


5a2b A User's Guide for the MOL is being prepared.


5a2c These two documents will be combined to form a technical report.


5b Meta


5b1 As work progressed on the Control Metalanguage, we found that major changes to the syntax occurred frequently. A compiler for an ever-developing language should be as flexible and easy to modify as possible. To implement the Control Metalanguage compiler, we have chosen to first implement a compiler-compiler system.


5b1a The system referred to by us as "Tree Meta" is largely based on a compiler-compiler system first developed by V. Schorre at UCLA, and described in the article on Meta II in the proceedings of the 1964 ACM Conference in Philadelphia.


5b1b The parsing part of our system works the same way as Schorre's, by a bottom-to-top syntax analysis.




ARNAS QPR4, February-May 1967


5b1c The code-generation portion of our system uses a tree-analysis technique first developed in the TRANGEN system of computer associates, and later expanded by the Meta compiler group at SDC.


5b1d A most attractive feature of the Meta system is that the Meta compiler can be written in its own language. This gives the system a flexibility and bootstrapping capacity many other compiler-compiler systems do not enjoy.


5b2 Programming


5b2a In order to develop our Tree Meta system, we first wrote a smaller, less powerful Meta-type system known to us as Meta-2.


5b2a1 The Meta system consists of two somewhat independent parts.


5b2a1a There is the Meta compiler itself. Operation of this compiler is largely a matter of subroutine calls, with very little in the way of computation.


5b2a1b The other portion of the system consists of a set of library subprograms called "Metalib". The initial version of Metalib was written completely in SDS-930 FORTRAN II. The programs were, however, designed to be general enough to use with the Tree Meta system.


5b2a2 The entire Meta-2 system was written in FORTRAN II using the 940 computer available at Tymeshare. This system has been completely debugged.


5b2a3 Our experience with SDS FORTRAN has verified our original conclusions about the speed of programs written in the language and the size of these programs. Even our small Meta-2 system runs very slowly and is very, very large.


Sb2a4 Since this Meta-2 program is to be used only to bootstrap the final Meta system, no documentation will be produced.


5b2b Tree Meta


5b2bl Metalib has been translated from FORTRAN II to ARPAS. It will later be translated to MOL when MOL is running. This ARPAS code is almost thoroughly debugged.


5b2b2 The Meta-2 compiler, which was written in its own




ARNAS QPR4, February-May 1967


language and which produced FORTRAN II output, has been modified to produce ARPAS output. It has compiled itself and has thus produced a checkout ARPAS version of itself.


5b2b2a This ARPAS version involves something like one-fifteenth the amount of program steps, and runs very much faster than the FORTRAN version.


5b2b3 The final Tree Meta compiler has been written in Meta-2 and is currently being debugged.


5b3 Documentation


4b3a A large technical report on the Tree Meta system is being prepared and will be released late this summer. Since there are no general introductory articles on Meta systems available in the literature, we are including in our technical report an extensive tutorial section on bottom-to-top syntax analysis as well as a complete history of all known Meta systems and a thorough guide to our Tree Meta system.


5c The Control Metalanguage


5c1 Programming


5c1a An initial version of the Control Metalanguage has been specified in the Tree Meta language.


5c1a1 This specification is about 75 percent complete. When it is completed, the Control Metalanguage compiler will also be completed.


5c1a2 The completion and extension of the Control Metalanguage will proceed as the final specifications for the initial 940 user-service system are completed.


5c1b Control Metalanguage programs will be mostly calls on an extensive set of library subprograms written in the MOL.


5c1b1 Fairly detailed specifications have been made for the general data structures upon which the above library subprograms will operate.


5c1b2 The actual coding of these programs has begun in the MOL.


5c2 Documentation


5c2a A technical memorandum on the Control Metalanguage was




ARNAS QPR4, February-May 1967


partially completed at the time that the IPR was issued.


5c2b This memo has been extended and refined as the Control Metalanguage itself has been developed.


5c2c This memo will always be dynamic in the sense that it reflects our current state of development. It should, however, reach various plateaus where it can serve as a detailed set of specifications for a display-oriented user-service system.


6 Other Activities


6a We have slowly and steadily been cleaning up our computer-held XDOC files (the 2800-odd technical-reference citations accumulated over the past 7 years in this program), and expect soon to generate a KWIC index for them.


6b We have been doing preliminary planning on setting up the "ARPA Computer-Network Information Center," as discussed at the ARPA Contractor's meeting in Ann Arbor on 11 April.


6b1 All of us in the AHI Program are quite pleased to take on this job, feeling that it fits very well with our interests and research plans.


6b2 We are assuming that we will eventually need more disk storage than what we now have on order from Bryant--but to avoid disturbing the current delivery date we shall not act upon this immediately (besides, the 96-million characters of the initial capacity would certainly suffice until at least a year from now).


6b3 We have been discussing this challenge with Dean Swank and Professor Maron of the School of Library Sciences at UCB, exploring the participation of faculty consultants and of graduate students in the design and operation of this system.


6b3a They seem quite interested, and we plan to proceed, having in mind at least acquiring a couple of Research Assistants from there to work part-time with us throughout the forthcoming year.


6b3b It seems obvious that this activity would benefit from their participation, but it is also quite certain that there will need to be an intensive period of familiarization on their part with the techniques and possibilities of our system--thus it seems wise to try to bring them into the program as soon as possible.


6b4 Our first direct action will probably be to make an initial




ARNAS QPR4, February-May 1967


collection of the reports, citations, abstracts etc. that are more or less readily available from the different contractors.


6b4a We may prepare (in our structured-statement form, as computer-held data) various lists of citations, for which we can produce KWIC indices.


6c We have been providing some support for implementing COPE and NLTS and FLTS on the CDC 3100 system at NASA's Langley Research Center.


6c1 Muriel Jarrett, a NASA systems programmer, spent almost two weeks with us, familiarizing herself with the systems, their implementation and their documentation.


6c2 David Hopper (one of our men) worked closely with her, and is currently at Langley to help check out the systems there. We expect this to take about a week of his time.


7 Visitors


7a A. E. Gribble and J. Page, NASA


7b A. Blue, RADC


7c Lawrence K. Grodman, Human Resources Foundation


7d W. J. Doherty and C. E. Shanesy, IBM


7e Robert Taylor, ARPA


7f Les Ernest, Stanford University


7g R. Robinson, Stanford University


7h Professor D. A. Thompson, Stanford University


7i Dr. Aufenkamp and W. Arntzen, Oregon State University


7j Roger Burnell, Stanford University


7k Arie Dirkzwager, Free University, Amsterdam, Holland


7l Peter T. Bennett, MIT


7m Dr. Alejandro Zaffaroni, Dr. Carl Djerassi, and Kenneth Hansen, Syntex Laboratories


7n Herman C. Osborne, Jr., Lester D. Tally, Charles T. Webb, and




ARNAS QPR4, February-May 1967


David B. Long, Mellonics Systems Development


7o Paul Davies and Robert Clark, Encyclopedia Britannica


7p Group of 6 from U. S. Forest Service, Berkeley, California; H. Lent, R. Russell, D. Claxton, W. O'Regan, M. Powell, and T. Yanke


7q Group of 6 from IJniversity of California, Berkeley, Institute of Library Science; R. Shoffner, C. Sypert, D. Barrett, W. Schieber, M. Bates, and L. Gould


7r A. Pratt, Service Bureau Corporation


7s Dean Swank and Professor M. Maron, University of California, Berkeley, Institute of Library Science.


8 Plans for Next Quarter


8a Most of our effort will be on developing the software and hardware for the new 940 system


8a1 We will aim toward having a first implementation that will provide users with basically the same features as now provided with NLTS.


8b We will develop specific plans for our Information-Center activity.


8b1 We will begin arranging for added staff, e.g.,


8b1a One or two more computer-science professionals


8b1b One or two (half-time) library-science professionals


8b1c Some clerk-typists (perhaps train, for part-time use, some already available SRI employees).


8b2 Begin collecting information from other research centers, setting up initial file formats, transcribing the reference material, organizing it, perhaps producing a KWIC index.




8b3 With this material as a "stationary target" sample, begin working more closely with the other centers to see what more information they might have for the reference file, or what more information they might want to see in it.


Submitted by:


[personal signature]
D. C. Engelbart
Program Head



Approved by:


[personal signature]
Torben Meisling, Manager
Systems Engineering Laboratory