VOLUME 5 NUMBER 1 JANUARY 1980 Page 2 for first mailings of April Fool's people we spoke with in Toronto Day. Those who have made submissions seemed less interested in a seperate to this and previous software distri- organization than in the problems of butions will be receiving release moving tapes across the border. If forms which must be Signed ~ and that is the case, we would be happy returned before we distribute the to contract with some Canadian licen- tape. see to reproduce and mail the distri- bution tapes within Canada. STAFF NEWS NOTICES We are happy to report that the two part-time employees of the associa- This document may contain information tion authorized by the Board of covered by one or more licenses, Directors are aboard. As a conse- copyrights, and non-disclosure agree— quence we hope to be able to process ments. Permission to copy without fee your letters and phone calls with all or part of this material is greater promptness than in the past. granted to Institutional Members of Susan Roth and Lorrin Wong have the Usenix group provided that copies joined Armand Gazes and Mel Ferentz, are made for internal use at the who continue to serve as volunteers. member campus or plant site. To copy While we continue to prefer written otherwise, or to republish, requires inquiries, we will try to handle specific permission. : telephone calls. Please avoid asking for a specific individual when you Editorial material, payments, soft- call, if possible. If you tell the ware submission, subscription re- person answering the nature of your quests, and address changes should call, it will be handled by the be addressed to: appropriate available person. The number is 212-360-1182. If we are Usenix Association inundated with telephone calls, we Box 8 may have to restrict this service to The Rockefeller University Institutional Members but, assuming 1230 York Avenue reasonableness on both sides, we New York, New York 10021 would hope that wouldn't be neces- sary. EXCHANGE ARRANGEMENTS The Usenix Association, while accept- ing members world-wide, has no desire to compete with similar organizations 7 in other countries. In fact, we are most anxious to enter into coopera- tive arrangements with Unix users’! group abroad. We have spoken with the groups in Great Britain and Aus- tralia and look forward to formal agreements with these and other groups that would give each group the right to reproduce for its membership tapes and newsletters of the others. Canada is a special case in that the UKUUG UNITED KINGDOM UNIX USER GROUP R.A. Mason, Department of Electrical and COMMITTEE Electronic Enqineering, Chairman : Alan Mason, Heriot-Watt University Heriot-Watt University, Secretary: S8ruce Anderson, University of Fssex 3-35 Grassmarket, Member(s}: Peter Collinson, University of Kent EDINBURGH. EH 2HT Tel: OJ1 225 6432 Ext: 155 8th January, 1980. Dc. M. Ferentz, Box 8, The Rockefeller University, 1230 York Avenue, New York, NY 10021, U.S.A. Dear Mel, With reference to our recent telephone conversation I am pleased to confirm the tentative arrangements made. We will send copies of all Newsletters produced and of any software generated direct to yourselves (at no charge) as our part of the reciprocal arrangement. This is by far the most sensible method as it allows ug to disemenate only that information which is pertinent to our own communities. On the possibility of operating a similar scheme, I am contacting the Australian group directly but I would be glad if you would inform any other major groups (Canada?) of my eagerness to make contact. The U.K. group, representing 48 local sites and 30 other subscribing sites meets quarterly and although Newsletters are at present bi-annual it ts hoped to step up production. A Newsletter will be produced following each meeting and as well as containing a report on that meeting, witl contain information about the forthcoming meeting. Local UNIX user groups (LUUG's) are being established here go that site representatives may have more frequent contact (monthly). These groups Look after their own internal software distribution and software policy. Thus reducing the number of centres to which we ourselves have to distribute. As the number of installations increases we are being forced to give ourselves a more formal structure. Although I realise that we are constrained by different legal and financial systems, it would be of great interest to me if you could send information on the final structure that USENIX took on (Byelaws ete.). Could you also tell me how you differentiate between academic and commercial sites, particularly with respect to software distribution. A! Mason THE UNIX USERS' GROUP MEETING IN SUMMARY -or- Reflactions of a Scribe UNIX Canference Medical Sciences Building University of Toronto June 20-23, 1979 This report is a summary of the exciting things people hed to eny at tha Toronto conference. Aa | was the only person taking notes for this purpose (although } was helped out by copias of speakers’ transparencies), the noles inevitably refiect my own personal biases and knowledge (or lack of such). | have tried to convey enough infor- mation for those who did not atlend the conference to get a good idea of the principal features and availability of the systems and projects Lhe speakers discussed. At the same time, ] hope those who atlended can use this report to remind them of al) the thrilling and not-so-thrilling things they heard. Ido not guarantee that what I report was actually said. If you want ta be SURE, check with the speaker in question. My apologies Lo everyone whom I bave misquoted. My thanks lo Sandra Wright for taking the noles of the faat session (Saturday morning). Thanks aiso to Jim DelGrande, Dave Galloway, Gerardo Lastra, David Miller, Sanand Pa- tel, Rob Pike, Bill Reeves, Henry Spencer, Dr. Martin Taylor and my wife Simone for proofreading and helpful comments on the draft, June 29, 1979 David Sherman Computer Systems Research Group University of Toronlo Toronto, Canada M53 1AL SBAIBAAUON & tek ik ek te ee te BONVBALONHG NNOAW-H Garoaones OornarrbaA Dh a beAme ee OOD EU UNH COBNAUAGN SOG AEOS OHO SRR Table of Contents Topic : Pascal and EN-1 Pascal and EN-1 Euclid Path Pascal Languages for tha VAX-11/760 C Compiter for the 280 UNIX Version 7 VAX UNIX 32V VAX UNIX - A User Comments KSOS - Secure UNI What's cooking on the Berkeley VAX News from Western Electric USENIX Winter Conference UNIX without a UNIX licensa What’s on the Berkeley tape YASL Core Graphics in C Perception and Information Enhancement GPAC and veat NYIT Graphics Graphics for lots of different terminals Versatec Typesetler Emulator Theories about Office Automation Musical Interlude Software Register proposal Database for a Micro Mini-UNIX on the LSI!-11 Real-time Data Collection and Failure Anelysis How to get more out of your 11/70 Real-time data gathering on UNIX Networking at Purdue Networking at NYIT RT/ENT: an RT-1f emulator on UNIX Software Tools Users Group Report Anyone doing system performance monitoring? USENIX Commniltees 11/40 Kernel Multiple Address Space UNIX for 1100 Users?2!!I Another large UNIX system Office Use of UNIX Implementing C and UNIX more efficiently UNIX on a UNIVAC V'77-600 An Accounting System for UNIX A High Performance UNIX System UNIX on the IDM 370 UNIX in the Undergraduate Lab Environment UNIX in a large educational environment QED, or The Little Ed That Grew News from the U.K. What's happening at BAN Tuning PWB UNIX Screen editors Speaker Andy Tanenbaum Carry Fostel James R. Cordy Richard Qolocca Bill Joy Keith Davis Brian Kernighan Tom London Ed Desautels Howard S. Weiss Bit! Joy (again) Al Arms Lou Katz, Melvin Ferentz John Donnelly Mark Krieger Bill Joy (one more time!) Navid Litienfeld Dennis Mumaugh Martin Tuori Ron Baecker Tom Dui Mike Muuss G.E. Toth David Macfarlane Bill Buxton Lynn Brock John Kornatowski Bob Hudyma Neil Groundwater Dan Gielan Eric Ostrumt Bit) Croft Bill Lindernann Mike Tilson Dennis tall Phil Poulos Lou Katz Alfred Whaley Jan Johnstone Ceorge Goble Walter Lazear Car! D. Howe IInrold Pierson Robert N. Jesse Mike Muuss Steve Vellovin Don Schertz Mike O'Dell Robert Pike, David ‘ifbrook R.).A. Collinson Cart D. Jlowe George Pajari Mark Pearson Name Arms, Al aecker, Ron Bellovin, Steve Holocea, Richard Brock, Lynn Buxton, Bill Collinsen, RPA. Cordy, James R. Croft, Bill Davis, Keith Desaulels, Ed Donnelly, John Duff, Tom Ferentz, Mel Fostel, Garry Gielan, Dan Gobte, George Groundwater, Neil Hail, Dennis Howe, Carl D. Howe, Carl D, Itudyrma, Bob Jesse, Robert N. Johnstone, Jan Joy, Bill Jay, Bilt Joy, Bill Katz, Lou Katz, Lou Kernighan, Brian Kornatowski, John Krieger, Mark Lazear, Walter Lilienfeld, David lindemann, Bill London, Tom Macfarlane, David Mumaugh, Dennis Muuss, Mike Muuas, Mike O'eli, Mike QOslrum, Fric Pajari, George Pearson, Mark Pierson. Ilarold Pike, ftobert Poulos, Phil Schertz, Don Tanenbaum, Andy Tilbrvok, David Trlson, Mike Toth, Gil. Tuori, Martin Weiss, Howard S. Whaley, Alfred Index by Speaker Topic News from Western Electric GPAC and veat UNIX on the 1BM 370 Path Paseal Soflware Regisler proposal Musicat Interlude News from the U.K. Euclid Networking at Purdue C Cornpiler for the 27.60 VAX UNIX - A User Comments Winter Conlerence NYIT Graphica USENIX Pascal and EM-t How lo get more out of your 11/70 Another large UNIX system Real-Lime Data Colleclion and Failure Analysis Software Tools Users Group Report Implementing C and UNIX more efficiently What's happening al BUN Mini-UNIX on the LSE-11 An Accounting System for UNIX UNIX for 1100 Users?7!1 Languages for the VAX-11/780 What's cooking on the Berkeley VAX What's on the Berkeley tape USENIX USENIX Committees UNIX Version 7 Database for a Micra UNIX without a UNIX license Office Use of UNIX YASL Networking at NYIT VAX UNIX 32V Theories about Office Automation Core Graphics in C A High Performance UNIX Syslem Graphies for lots of different terminals UNIX in a large educational environment Reat-Lime data galhering on UNIX Tuning PWH UNIX Serven edilors UNIX on a UNIVAC V77-860 QkD, or The Little fd Thal Crew Anyone doing system performance monitoring? UNIX in ihe Undergraduate Lab Environment Pascal and KM-1 QHD, or The lattle Rd That Grew RT/EMT: an RE-0) ernubator on UNIX Versatec Typeselter Benulator Perception and fnforimation Enhancement KSOS - Secure UNIX 11/40 Kernel Mutlipte Address Space Speaker 12 20 48 26 25 50 de 4 rat 1a 30 40 29 35 42 51 za 44 39 11 18 1a 37 27 15 41 i7 3d 24 iB 45 22 48 at 52 53 43 49 36 47 43 34 23 19 10 38 WEDNESDAY MORNING Session 1: LANCUAGES ON UNIX Chair: Ron Baecker, Universily of Toronto Speaker 1, at &15 a.m. Pascal and EM-1 Prof. Andrew S. Tanenbaum Vrije Universiteil ~ Wiskundle Seminarium Box 7161 1007 KC Amsterdam, The Netherlands Andy Tanenbaum gave a detailed talk on the UNIX Pascal systam developed al the Yrije Universiteit by Johan Slevenson, tans van Staveren and himself. Their aim has been to make the compiler and support soltware as portable as possible, with transport to vari- ous mini's and micro's in mind. They feel that °C is not the beginning and end of the world," and that Pascal's attractiveness lies in its simplicity and readability. The Pascal language implemented is the complete language, and is compatible with the “British Standard,” which has been proposed as an international {ISO) standard. There are also a few extensions, such as assertions, mark/release, exlernat peocedures, and UNIX-style zero-terminaled strings. It is also possible Lo create libraries of separately compiled Pascal, C, or assembly language procedures that can be called from Pascal programs. There are also many debugging and performance facilities, such as warning Messages for unused or undefined variables: run Lime error messages giving source line number; counts of source lins executions; and procedure entry/exit tracing. The system can be used to produce either compiled or interpreted code: "pe prog.p” witl produce interpreted code, and "pe -C prag.p” compiled code. In both cases tha compiler produces code for an abstract stack machine called EM-J, described in CACM, March 1978. If the -C flag is given, the KM-) code is passed through a machine- independent optimizer, and then transiated Lo the UNIX assembly language to be as- sembled and loaded by as and Id. If the -C flag is absent, the EM-1 code is assembled Lo EM-1 binary for subsequent interpretation. There are 64 versions of the interpreter corresponding lo 6 independent debugging options. The system automatically selects the correct version from a directory, if it is found, or creates and saves a new one dynamically, if need be. The optimizer’s capabilities include constant folding (e.g. mepping 10+4 into 14), using special instructions (e.g. increment for i:=i+1), strength reduction {e.g. shifting for multiplication), reordering (e.g. -k/8 becomes k/-8), and many more. The distribution is available as part of the 1979 conference distribution. It contains all the sources to the compiler proper, the optimizer, the EM-t to PDP-11 transtater, and lhe interpreter, as well as various supporl programs. There are also libraries, includ- ing one containing al! the UNIX system calls, so it is possible to call open, seck, fork, exec, etc. from Pascal programa. The tape also contains versions of as and id capable of handling files larger than 64K bytes. Separale | and D space is useful, but nol re- quired. Anyone who installs the system shauld write Lo Andy to be put on the mailing Hist for bug reports (the first of which was available al the conference). As soon as the version 7 Pascal systern becomes available (in the fail) both iL and Ube version 6 one will be dis- tributed. To avoid sending tapes across Lhe ocvan, Lhe Vrije Universiteit bas made an a UNIX Summer Conference : June 20-23, 1979 arrangement with the Pascal group of Intermetrics, Inc. {see below), for those who do nol have Lhe conference distribution. Speaker 2, at 10:00 Pascal and EM-1 Garry Fostel Intermetrics Incorporated 701 Concord Avenue Cambridge, Mass. 02138 Garry Fostel discussed the Pascal compiler described above. He said it is exceptionally well put Logether and documented, It is available in the United States from Inter- metries for « $50.00 handling charge. IL runs on atandard UNIX V6, and there are certain assumptions about file names near the root. There may be a version for PWB saon, caffee Speaker 3, at 10:45 a.m. : Euclid James R. Cordy Computer Systems Research Group University of Toronto Toronto, Canada M5S 1A1 The Toronte Euclid Compiler Project is a joint project of LP. Sharp Associates (better known for their APL services) and CSRG, funded by the U.S. Defense Advanced Research Projects Agency (DARPA) and the Canadian Department of National Defence (DND). itis a follow-up on an attempted implementation by the Systems Development Corporation, which failed in an earlier atlempt to produce a Euclid compiler, Euclid is the so-called “language of secure systems”, where secure means provadly secure. The Euclid language is roughly based on Pascal but has many extensions. It is very strongly typed; it has explicit visibility ‘and scope control, unlike Pascal; it prohi- bits aliasing, or having two identifiers for a single object. This is necessary for program verificalion and extends even, lo assuring that no two actual parameters to a routine address the same element of an array. The heart of the language is the module. A module operates as a single mechanism which manages a set of data structures internally and “exports” only a clearly defined sel of externally visible operations and data, The program external to the module can invoke any of the exported operations of the module and can access the exported data, bul knows nothing of Lhe internal implementation of the module. Thus, the module lovks like a "black box" which provides certain services and information Lhrough its ex- porled operations and data. It is a self-contained “package” of procedures, variables aud constants. Types can be paramelerized in Kuctid. A parameterized type definition defines a tem- plate for a “family” of Lypes defined by different parameter values. In its simplest form, a parameterized type can define a set of range types whose upper and/or lower bound depends on a parameter value. The full power of parameterized types is realized in Lhe parametenzed module, which ‘can define a whole family of similar mechanisms. ets UNIX Summer Conference June 20-23, 1979 Constants cen be “manifest” - like constants in most other languages ~ or "non- manifest”. The value of a non-manifest constant can be a run-time delermined variable expression, but is “constant” in the sense that, once sel, the value remains constant in the scape. Euclid provides pre- and post-assertions for routines, which enable the specifications and assumptions of a routine to be enceded and checked al run-time. Modules may have “invariant” assertions which specify invariant conditions of the module. These also can be checked at run-time when entering and exiting routines exported from the module. Pointers in Euctid are restricted to a "z0ne” of storage maintained either by the sys- tem or by a user module, and to a particular type of object (which may be paramater- ized). The project schedule so far has been: Sept. 1977 - Transliterator of Small Euclid to C - so as to bootstrap by writing the en- tire Euclid compiler in Euclid. June 1979 - Translator - compiter for a large subset of Fuclid. Fall 1979 - Compiter should be finished, It will cover most of Fuclid, and will include all verification features. Implementation of an automatic verifier is nol within the scope of Lhe compiler project. The compiler does not run withoul separate T and D space. There are 4 machinc- independent and 2 machine-dependenl passes to Lhe compiler. Performance: Object code produced is generally as good as or better than C in space and time, Because there ix 30 much modularization in Euclid, procedure calling has been corefully optim- ized. A oall lo a parameteriess procedure with no local variables costs only two in- structions (JSR, RTS). The source code for the Euclid Translalor is about 80,000 lines. To compile the compiler takes sbout six hours (two to three hours for the largest pass; there are six passes) on a4 relatively quiet 11/50. No atlempl has been made to tune the compiler for speed, Euclid ia not available publicly yet, although the current version is being delivered to DND (Canada) and DARPA. When the project is complete, il is hoped that JPSA will distri- bute it commercially (with full maintenance) and that it will be available (without sup- port) to educalional and research institutions from CSRG. If you would like lo be placed on the mailing list for information aboul distribution of the Euclid Compiler when {it becomes available, send your name tu Jim Cordy (address above). Speaker 4, at 11:35 : Path Pascal Richard Bolocca University of Illinois Champaign, Illinois Path Pascal is a superset of Pascal P. It has additional features for concurrency; for date encapsulation; and for self compilation. Jt is used for the teaching of systems pro- gramming at the University of Jilinois. It runs on a Cyber and a PIP-tt, and may be used for real-Lime programming by the NASA space shultle prograin. Path Pascal was developed by Roy Campbell al the University of ittinois. Concurrency: features arc the process, whith is like a procedure bul “goes ils own way": and interrupt processes, Dula encapsulation: there is anew type, called an “ob- +f} UNIX Summer Conference — June 20-23, 1978 ject”, =n like a module in Euctid or Modula; processes can be created dynamically with this. The compiter is seff-compiling and lakes two passes to reach assembler code. The cormnpiler source is about 4K lines for each pass; the binary takes 46K and 32K bytes for the two passes. The second pass compiles itself in about 4-5 minutes. The compiler needs separated | and I space lo run, or non-separated I and D with “segmentation” overlays. The. language currently runs on UNIX but can also run standatone on an LSI-i1. It ia bighly portable: in existence now are compilers for the Z60, all PDP-11's, and Lhe Series/l. Being developed is one for the Prime 500 and similar models. Path Pascal will be available by the fall for a nominat {tape-handiing) fee. Speaker 5, at 11:45 Languages for the VAX-11/780 Hill Joy University of California at Berkeley 964 Riley Drive Albany, California 94708 Bill discussed the various languages and near-languages running under UNIX on the VAX at Berkeley: 1) Pascal. The Herkeley Pascal is well known. Like the front end of the portable C com- piler, its Iron end ix relatively machine independent, The program pi (Pascal inter- preter) is being modified lo become pe, which will output code to be compiled by either pee on . 11 (generating PDP-11 assembler files) or pec on the VAX (generating VAX as- sembler), 2) LISP. The LISP compiter, a.k.a. "Franz Lisp", is written in C. A simple interpreter is running, but the system so far can run only 2/5 of MACSYMA (two of five parts of the MACSYMA compiler which take up 400K and 350K bytes respectively have been run to- gelber ona Kemegabyte machine). Work is now proceeding on a “Byte LISP" for com- pactness, and a compiter. 3) APL. Work is being done on moving Ken Thompson's APL to the VAX. The advantage of running it on the VAX is that one can get around Lhe 65K workspace limitation (with Paging) and run real APL. 4) Moduta. The compiler, originally BCPL code from York (England), was rewritten line for line to compile as C, and has been taken over to the VAX. IL does not yet compile VAX cude. 5) Rigel. This retationat database language was described in the proceedings of the last SICMOD conference (at Boston). 6) MACSYMA, This system is supported by the LISP system described above. The following are languages which the Uerkeley people are thinking about and/or in- turested in: 7).2. Amodified version of FORTH. Sinee FORTH is propriclary, no more need be said... 7% UNIX Summer Conference June 20-23, 1979 8) Algol 68, They would like to move this to lhe VAX. 9) Snobot/Spitbol. It would be especially nice to gel a Spitbol up. 10) STAPLE. This is being worked on. It is a structured applications and programming language for modification of syslem programs. I's based on C, PL/1, Pascal... None of the above languages is available officially (i.e., nothing is packaged) except for the Pascal interpreter. The following can only marginally be considered languages...: 11) Termcap. Makes the screen editor terminal-independent. 12) C Shell. Asbelt with its own command language, much like C. Speaker 6, at 12:15 C Compiler for the 780 Keith Davis Teletype Corporation 4 Mayflower Vernon Hitis, litincis 80061 (219) 982-3619 The C compiler for the Zitog Corp. Z80 is a cross-compiler which runs under PWB. Developed by Interactive Systems. it features everything in the C compiler except floal- ing point. Il generates specific 280 cade, which is 42% larger than PDP-11 code because of instruction inefficiency. The compiler is available, atong with a debugging package {like cdb), for those within the Bell system from Keilh Davis of Teletype. Internal charge: $5,000, Outside the Bell system, it’s available [rom Interactive Systems (Keith can point you at the right peo- ple). LUNCH UNIX Surnmer Conference June 20-23, 1978 WEDNESDAY AFTERNOON Session 2: THE UNIX OPERATING SYSTEM Chair: Bill Reeves, University of Toronto Speaker 7, al 2:00 p.m. UNIX Version 7 Brian Kernighan Heh Labs Murray Hill, New Jersey 07974 A lot of the Version 7 changes were covered in the BSTJ issue on UNIX. The important new things are: the shell; huge files (2* bytes): portability; the portable C compiler; lint (C program checker); stdio package; Fortran 7 compiler; make; lex (lexical analysis phase of yacc); awk (pattern scanning and processing language); sed (stream editor); learn (computer-aided instruction about UNIX); adb {a “complex bul indispensable" C debugger); vucp (UNIX-ta-UNIX comm unications handling). The new shell is much more oriented to programming than previously. Cone is the old "goto"; in its place are “for .. do ... done “, "case ... esac”, "if... lhen ... else .. A". “while”, “until”, and a “trap” command to handle interrupts. Also new are shell vari- abies {including special ones for home directory, mail file and bin path search). Changes in C for portability are ell described in the C book; in addition, structure as- signment has now been implemented, and there is an “enumerated type” feature. Lint is a program checker that goes right through a program (including multi-file pro- grams) and checks for lype violations, portability problems, probable errors, and bad style Lhat may be evidence of error. Make is a command Lo compite according lo instructions in a "makefile”. It knows about a lot of things by default, such es that “ce -c zork.e” will produce "zork.o”, Its only drawback it Lhal it doea nol yet handle ar libraries. ‘ F77 compiles the complete Fortran 77 language, with n few extensions. ft generates Lhe same intermediate code as C, so UNIX 1/0 is accessible. Also on the distribution is strucl, a program Lo convert standard unintelligible Fortran to Ratfor. Awk is a pallern scanning and processing language in which programs ere typically 1 or 2 lines. Initialization and declarations are nol used -- awk decides what an item is by how it is accessed and used. It is useful for such Lasks as switching two fields within « file, or adding up the contents of & particular fieid. Learn interprets a script to teach UNIX. Scripts are available for teaching UNIX file handling, ed, C (not a very good tutorial, according to Drian, who wrote it), and the -ms macro package {the Hell people don’t think anyone should use nrofl directly so there's none for nro). The lessons are streamed and tune themselves to the competence of the student. Brian mentioned that writing scripts for these things is extremely hard. Tar, the new tape program, was discussed by Tom London (below, bul the information belongs here). Files sent lo tape with tp slill work. All code is writlen on the tape in ASCII, so any system ean read it. You can block the tape with huge blocks. Updates on Ihe und of the tape are allowed, and extracts are done on the last instance of a file. DireeLorivs are built for you when you extract files. -9- UNIX Summer Conference June 20-23, 1979 coffee ’ Speaker 6, at 3:30 VAX UNIX 32V Tom London Beil Labs Murray Hill, New Jersey 07974 UNIX 32V is UNIX on a VAX-11/780. Torn described in some delait the mysteries of VAX addressing, with its 31} bit address space. Presently running on the VAX is a full UNIX * V7, with the same file system, shell, commands, tanguage end system interface as the PDP-11, There are just a few incompatabilities: the VAX word size is 32 bits, so pro- grams which know that a word is two bytes are wrong and ints are aligned at 4-byte in- stead of 2-byte boundaries; long ints are encoded in the reverse order from the PDP-11; traps are handled differently; poinlers cannot conveniently be assigned the value -1; external variables declared contiguously won't necessarily be contiguous in memory; and so on. The VAX, with its block-move instruction, performs faster than the PDP-11 in tasks which involve moving large chunks of data at once {allhough it does nol quite Jive up to the manufacturer's billing af being consistently twice as fast). Compiling a C program on the VAX is faster than on the 11/70, even though the peg being used igs Steve Johnson's, which is not a5 finely tuned as Dennis Ritchie's. UNIX 32V has partial swapping ~- pages can be scattered anywhere in memory, and only part of « program will be swapped oul Lo make space for another. These improvements reduced execution times by a factor of 2/3. Demand poging is not yet available, but is being worked on at Berkeley (see Bill Joy's talk below). In summary, the VAX is well worlh using as a UNIX machine. You get a full V7 on a fas- ler machine, a larger address space, and the capabilily of handling huge programs. UNIX 32V is now available from Western Electric (see Speaker #12 for details). Speaker 0, at 4:30 YAX UNIX - A User Comments Prof. Ed Desautets Computer Sciences Department Wisconsin Univeraity 1210 W. Dayton Madison, Wisconsin 53706 The VAX al Wisconsin is now running UNIX V7, it is part of a network of several PDP-tt's and LSI-{1's, hooked up lo each other in various ways. The VAX originally ran VMS (with auto-reboot); the YMS Pascal being developed at Universily of Washington (Seattle) was lesled at Wisconsin, VAX UNIX was found to be easy Lo install, although it is “not yel robust". Projects underway on VAX UNIX include: an inteNigent mass storage system (Dave DeWitt, Tony Klug); compilers and operating systetns for cornputer nulworks (Raphael Finkel, Marvin Solomon); a data base system for Al (Larry ‘Travis); a Conmputer Serenee network for Telenet (Larry Landweber; Ed Desautels); connection wilh a large array of micros for perception research (Len Uhr); ‘new archilecture and syslerms work (Ray Moore, Hob -10- UNIX Summer Conference June 20-23, 1978 Cook, Ray Bryant and Goodman); and facililation of access to computing for the visual- ly impaired, along wilh work on the Teletext/Viewdata systems (Ed Desautels). Speaker 10, at 4:45 KSOS ~ Secure UNIX }oward S. Weiss : U.S. Department of Defense 9800 Savage Road, Room 171 Fort Meade, Maryland 20755 KSOS stands for Kernclized Secure Operating System. It started as an ARPA project in 1977 and is being coded in Modula. (Euclid was the original choice but the Euclid pro- ject was not ready in time.) Written by Ford Aerospace at Palo Alto, it should be com- pleted Lhis winter. KSOS is an operating system in itself. The "UNIX" that runs on it is a UNIX emulator. Other emulators may be written, should Lhe demand warrant it, The UNIX emulator is committed lo performing at no worse then one-half Lhe speed of UNIX V6. KSOS will be available to Lhose within the U.S. federal government system. It is un- known yet whether it will be made publicly available. Speaker 11, at 5:00 What's cooking on the Berkeley VAX Bill Joy (again) University of California at Berkeley Projects underway on the VAX called “Ernie” (4 megabyte) include: 1) paging (Ozalp [tabaoglu, Juan Porcar) -- see below 2) microcode work (Dave Paterson, Richard Tuck) 3) a new Moating point box (George Taylor, V.V. Kahan) 4) a floppy-disk driver (Richard Tuck) 5) swapping to UNIBUS disks (Eric Allman, Bob Kvidle) 6) LNI interface for networking (Ken Birman, Larry Rowe) Recause the hour was tate, Dil! restricted himself to talking about paging. The goal of lbe pager, called PUNJVAX, is lo support large programs, with mostly large text seg- ments {up to 6 megabytes of text). There were some clear archilectural problems to overcome, in thal the VAX has no reference bils and has very small pages (512 byles). The paging system ia well on its way now, however, 2 UNIX Summer Conference June 20-23, 1979 THURSDAY MORNING Session 3: Items of Interest to Users Chair: Met Ferentz, The Rockefeller University Speaker 12, at 9:00 a.m. News from Western Electric Al Arms Western Flectric Company, Inc. Box 25000 Greensboro, North Carolina 27420 Al reminded us all thal Western Flectric selis a UNIX license "as is", with no warranty of any sort and no maintenance or service. He announced that UNIX Version 7 and VAX UNIX 32¥ are on the market. Western Electric's commercial! rates are now: Add. CPU __Binary } Mini UNIX = $12,000 UNIX V6 20,000 6,700 PWB 1.0 30,000 10,000 V7 26,000 9,400 For users who already have a UNIX license, V7 may be obtained for $12,000, and $4,000 for additional CPU's, No such discount is available for S2V, Educational institutions may get an “administrative” license, for internal business uses, al one-third of the commercial rates, The educational research price is $300 for V7 or for VAX 32V; $230 for V@ tape and manuals. PWB Version 2.0, which has the V7? file system and is fully compatible with V7, was just released internally within Bell. Jt will nol be available publicly for some time, MERT will not be released (i.e., it officially does not exist any more). V7 will probably not be released for the Interdala 32-bil machine (8/32). A point of interest: there are currently about 800 V6 licenses, of which 250 are commercial. Speaker [3, at 9:45 USENIX Lou Katz ; ; Melvin Ferentz Columbia University The Rockefeller University New York, NY 1200 York Avenue New York, NY 10028 At the users* meeting in New York last year, a committee of five was nominated to in- vestigate the selting up of a formal UNIX Users’ organizalion to handle such matters as lape distribution, newsletter publication and conference planning, On June 20, 1979, an association was formed called USKNIX. It has a legal existence in the State ot New York. Its Hoard of Directors initially consists of: Lou Katz, President: Lew Law, Vice-President; Armand Gazes, Secretary; Mel Ferentz, Treasurer; Mars Gralia and Peter Weiner, members of the Doard. USHNIX invites inslitulions and individuals to join: it has no exclusive rights” over unylhing asd no-one js obliged to join. UNIX Summer Conference June 20-29, 1979 If USENIX becomes successfully operational, it will set up in business at the Rockefeller Universily, handling the affairs described above. Membership in USKNIX will fail into one of four classes: {a) voting: (b)} individual; (c) public individual; (d) non-voling institute, Voting members will be institutes with a UNIX license who pay $300 for cach vote, lo a maximum of the number of CPU's for which they hold licenses, The educational-institulion rate will be $100 instead of $300. Voting members will delegate individuals to exercise their instilution’s vote. Individu- als will be permilled Lo join for $12 and reccive the newsletter, which will be produced at least 10 times per year. Individuals who work for or are associated with a UNIX license holder and are therefore bound by lhe licensa’s non-disclosure clauses wil! hold regular individual membership and be entitled to receive information about proprietary parts of the sysLem. Those not bound by non-disclosure may hold an “out- side" membership. Non-voling institulional membership will be available to those in Lhe Bett system (who are of course not “ticensed”). USENIX will hire a full-lime employee to handle the newsletter, tape distribution and the maintenance of a data base on the UNIX. The facililies at the Rockefeller University will be avaitable for this purpose. The bylaws and an invitation to join will be distributed to all parties thought to be in- terested. Election of a new slate of directors will be held by mail, probably sround the December. The associalion may be contacted by mail at: USENIX Association Box 8 The Rockefeller Universily 1230 York Avenue New York, NY 10021 Lou announced that the Association is currenUy asking for volunteers to serva on four commitiees: Agenda for the Winter Conference at Boulder; Site Selection for next year's Summer conference; distribution Lape format; and nominations for office for De- cember elections (but see Speaker #37, below}. Speaker 14, at 10:30 Winter Conference John Donnelly Nationa! Center for Atmospheric Research Houider, Colorado A brief announcement that the Winter Conference will be held January 29 - February 2 (Tuesday through Saturday) at a convention centre in Boulder. All users are en- couraged Lo bring Lheir /dev/ski drivers. coffee at3= UNIX Summer Conference June 20-23, 1979 Speeker 15, at 11:00 UNIX witbout a UNIX license Merk Krieger Whitesmiths Limited 127 East 59th Street New York, NY t001t1 (212) 796-1200 Whitesmiths have started from a UNIX binary license and created their own C compiler for the PDP-11. The compiler produces either UNIX or Macro-1!1 assembler. Jt supports the full Version 7 compiler as defined in the Kernighan and Ritchie book. (Extensions lo Ubat syntax mentioned by Brian Kernighan in his talk, such as structure assignment, will be provided for if these became standard.) The compiler runs under any of UNIX, RSX-11M, RT-11, JAS and RSTS, A portable library comes with Lhe C compiler. It tealures alloc and free; char, line and formatted 1/0; and string functions. Stdio is not followed too closety. A PDP-tt machine library provides certain functions that the hardware can't do on some machines, such as foating-point arilhmetic. Currently avsilable is a C compiter for the 8080. It runs on any PDP-11 and takes the same V7 C as input. It generates an intermediate assembly !anguage called “a- natural", The transtalor fram a-nalural to ISIS-!I or CP/M microsoft assembly code is the fourth pass of the compiler. Il generates code only $0% larger than the correspond- ing PDP-11 code {not bad considering the instruction inefficiencies in the 8080}. The machine library for the 8080, written by Sill Plauger, supports all 16-, 32- and 64-bit operationa supported by C. By July 1978 Whitesmiths will be distributing a full “a-natural" assembler, Joader and ji- brarian for the 6080. it will run on the PDP-11 or the 8080 on any of Lhe operating sys- tems named above. By the fall the IDRIS operating system {named after the Persian deity over toolsmiths) will be on the market. It will look exactly fike UNIX and will be runnable on an LSI-t!. Nemory mansgament should be available by the early winler. By early 1980, a native C compiler for the VAX and C compiler for a 16-bit micro should be available. Mark went into some delail aboul the features of lhe “a-natural” assembly language. The C compiler currently available costs $5, 000 {including source). Without Lhe source, it's $500, Speaker 16, af 11:55 What's on the Derkeley tape Bill Joy (one more time!} University of California at Berkeley The tepe includes: Berkeley Pascal version 1.2; ex (display editor which requires separate | and D space) version 2; C shell (runs on either V6 or V7, good while convert- ing): ~me macro package (faster than ~s); a new tail prograin; the Huerkeley net- working (foc machines hooked up by back-lo-back DZ-1 ts; mods be sldio for sirmullarre- ous read/write; miseciianeous ulililics; and a V7 simulator library (system deponden- Pas UNIX Summer Conference June 20-23, 1979 - cies can al! be teft in Lhis lib and changed at once). For each item one will find: full source; ¥6 binaries; V7 makefiles; all manual sections and documents; and a Versatec copy of the printed documentation. Everything except Pascal runs on both VAX UNIX and V7, Everything runs on 11/40's and 11/34's. To get Lhe lape you necd $60, a copy of your UNIX license, and a V7 or Photolypesetter license for the -—me package. You will get a 1200-foot tape at 600 bpi. Speaker 17, at 12:10 YASL David Lilienfeld Johns Hopking University 34h & Charles Streets Valtimore, Marytand 21218 YAS1, ~- Yel Anolher Statistical Language -- was designed because there was a need for a statistical package on a small machine. The reasons existing packages (such as BMD or SPSS) could nol be used were (a) they are clumsy; (b) you “can’t tell them what you want them lo do”; and (c) you need a large machine, such es a DEC-10 ar a COC 6800, ta run them because of the memory they take up. The writers of YASL have emphasized structure in their langusge, which they have tried - lo make a superset of C. Certain [estures useful for matrix statistics have been designed, such as case ranges (e.g., case ‘a'~z':), and declarations for various Lypes of matrices. Matrices can be dynamic and/or virtual. Virtual matrices are necessary to run large regression analyyeg and the like on a machine with raltively little memory. The compiler is currently being written. It is Lhe aim of the authors to have something running by the end of the summer, and a finalized version by June 19B0. At thal point it will become generally available (including the source). LUNCH -15- UNIX Summer Conference dune 20-23, 1979 THURSDAY AFTERNOON Session 4: Graphics, Music nnd Typesetting Chair: Ron Baecker Speaker 15, at 1:55 p.m, Core Graphics in C Dennis Mumaugh U.S. Department of Defense 9800 Savage Road : Fort Meade, Maryland 28755 Core Graphics is an “Independent Craphics System” designed by the writers of Lhe Core Report {see 4CM Computing Surveys, December 1978 fora whole issue on Core). It fealures a set of common subroutines for any language to do all the basic things a graphics language should do. At the bottom level are device-dependent primitives to do such tasks as basic line drawing. Dennis reported Lhat a Core Graphics for C has been writlen, although it is nol yet com- plete, It has been checked on Lhe Version 7 C compiler. The only device driver with it is one for a Genesco graphics system (which might not be of help to anyone, since Genesco itself bas no standard graphics format}. H differs from Core in a few details, such as not having funclion names longer than 7 characters (the maximum dis- tinguished in C). This graphics package js available as part of the conference distribution. The Core manual is not on it; that may be obtained from SIGGRAPH on machine-readable tape. This Lape hag a list of differences between lhe package and the SIGGRAPH rnanual. Some test pragrams and instructions on how to build the driver are supplied. Speaker 19, at 2:05 Perception and Information Enhancement Martin Tuori Defence and Civil Institute of Environmental Medicine (DCIEM) 1133 Sheppard Avs. Weal Downsview, Ontario, Canada M3M 3B9 (416) 633-4240 «204 Nartin Tuori and Dr. Martin Taytor are deing work at DCIEM on humen perception. Their aims are to sludy vision and audilion in complex situations, and Lo develop visual and auditory enhancements which will aid persons working in lhose siluations. Experi- ments will be conducted to demonstrale the effecliveness of such enchancements. The techniques will be of use with information from various sources, including carlh resources satellite imagery. Martin discussed Lhe "IFAC" file formal for storage of picture, sound and olher data, and invited interested persons to conlact him. The format is hierarchical; the atruc- tures inherent in the dala are first presented ina header. This will permit programs to interpret data files of varying structure. He would like lo discuss the IPAC technique, with @ view to working oul a more widuly aceepled slandard. Martin then showed slides, gencrated wilh a Dicomed 048 film recorder, including: (a) -1W- UNIX Summer Conference June 20-23, 1979 photographs of parts of Canaita from LANDSAT satellite, and various techniques by which Lhey can be enhanced; (b) images made with fas, a program written by Mike Til- son al Lhe Universily of Toronto to generate coloured pallerns based on x-y-r-thela functions and & 266-cotour map; and (c} pictures of Mars taken by Viking orbiter and brought from Callech by Rob Pike. Speaker 20, at 2:28 GPAC and reat Ron Baecker Computer Systems Research Group University of Toronto Joronlo, Canada MSS JA1 {416) 978-6768 There is nothing new to be said about GPAC. It continues to be used as the graphics package at the University of Toronto for building highly interactive systems (such as ones for producing animated film, composing and performing electronic music and visualizing the behaviour of simulalion models). It is available and fully documented os part of the Toronto distribution. GPAC was designed and written by Bill Reeves, who also wrote ucal, a Versalec typesetler-simulator which runs as an output filter to troff. Veal (which was used to generate this document) is available free as part of the Univer- sily of Toronto distribulion #3. Speaker 21, at 2:30 : NYIT Graphics Tom Dufl Computer Graphics Lab New York Institute of Technology P.Q. Nox 170 Old Westbury, NY (516) 626-0958 The Computer Graphics Lab at NYIT is in the business of making cartoons, ranging from short spurty to (allegedly) full-length features. They don‘t distribute any of their software. They have lots of hardware: 10 PDP-11's which run UNIX, jots more which don’l, a VAX, 18 frame buffers (512 « 512 x 6 bils each), and, as the ultimate peripheral device....a TV studio, Need we say more? Tom showed a very entertaining 20-minute videotape about NYIT's graphics. coffee Speaker 22, al 4:00 Graphics for lota of different terminale Mike Mitiss Jotus Uopkins University Mectrienl Maginceering Department Bartors Tall Vattimnore, Md. 21218 oars UNIX Summer Conference June 20-23, 1979 Nike Muuss described the “Terminal Independent Graphics Package” at Johns Hopkins (funded by the Army Research Office). The goal of the project was to create a package which would let one create a graphical display only once, and use il subsequenlly on different terminals without re-execuling the original “aout”. The system has a graphics format which lets one add new devices easily. It is also possi- ble lo reprocess Lhe display after generation. The display size is exactly the same for a given picture on all devices, Low level primitives, using the concept of a "virtual pen". have been wrillen for the package: pen up/down, move pen, new form, new origin, sel cotour/intensily. High level routines available include rotation, scaling, and line, symbot and number drawing. 3D perspective is under development. Devices currently supported wilh the package include the HP2B00A series; TekLronix 4008, 4010 and 4014-1; Hewston Instruments “Complot"; Versatec; Diablo 1020 series; and Ramtek Colour Graphios. tt takes one afternoon's work to add a new device to the package. The package is distributed by: Dr. Bruce Henriksen Ballistics Modelling Division Ballistic Research Lab United Slales Army, Aberdeen Proving Ground Aberdeen, Maryland 21005 Speaker 23, at 4:16 Versatec Typesetter Emulator G.E. Toth Johns Hopkins Univeraity Electrical Engineering Dept. Barton Hall VERSET was partially developed at Johns Hopkins University bul completed and de- bugged for the commercial firm which markets it. Jl can run wilhout separale | and 2 spsce and fits entirety on an RK disk. The font file takes up 1321 blocks and covers all troff point sizes and widths. VERSET runs as an oulpul filler to troff without any modifications to existing software: JL can process a page on Lhe Versatec in 20 seconds. Verset is available in binary form for $2000 to commercial users, $500 to educational licensees. For more information contact: 7 RLG Associates, Inc. Suite 16 11250 Roger Bacon Drive Reston, Virginia 22090 (703) 471-1108 -1f- UNIX Summer Conference June 20-23, 1979 © Speaker 24, at 4:25 Theories about Office Automation David Macfarlane Hell-Northern Systema Research H22 University Avenue Toronto, Canada M5C 1W7 David Macfarlane gave us a slide show from BNSR called “An Integrated Methodology for Office Automation”. In the slide show, the merging of the three ficlds of computers, communications and office equipment is discussed, and certain design problems in the conslruction of an “Office Inforrnation Communications System" are explored. The butlom line is that there is a need for an integrated perception for the study, design and imptementation of an automated office system. Nothing has been done beyond the thinking stage as yet, Speaker 25, at 4:50 Musical [nterlude Bill Buxton Structured Sound Synthesis Project Computer Systems Research Group University of Toronto Toronto, Canada M5S 1A1 Dill Buxton “conducted” a half-hour concert generaled by the SSSP music synthesizer running off an LSI-11 wilh floppy disks. The entire system is written in C and takes up about 14.5 Kwords of core. The musical patterns were loaded into core at the start of the perfurinance, and Bill "conducted", sitling ata termina? wilh a bil pad, cursor and function box. The lerminal's addressable cursor was used Lo approximate lhe graphic environment available on CSRG's 11/45 with a full tablet and vector graphic display (Lhose who came (to the demos on Wednesday and Thursday evenings saw the real thing). “Conducting” consisled of varying, as 4 performer, the volume, timbre, articu- lation, tempo and other properties of the patterns. World UNIX Reversi Championship Thursday, June 21, 7 p.m. - 1 a.m., the World UNIX Champion of Reversi was deter- mined. Reversi, as you all now know, is a board game (also known as Othello) played by two players on a 64-square board. Fight entries competed, including one written by the organizer, Bill Reeves. Our hardware masterinind (Guy Fedorkow) set us up with nine terminals in the CSRG vonference room. Dave Sherman directed the tournament as a 4-round Swiss system {along the lines of a chess tournament), in which players with equal records were paired in each round. We ran three games simultancously on the 11/50 (i.e., six pro- grams, of which a maximum of three were active at any one time), and one on the 11/45 in competition with the graphics/music demos going on in the lab. R Dennis Rockwell of Duke University emerged as the winner with a perfect 4 out of 4, beating Tom Duff in the tast round. Hill Reeves finished second, tosing only to rdr in the third round. -~19- UNIX Summer Conference June 20-23, 1979 To ensure that Lhe tournament didn’t last all night, a time limit of 15 minutes per pro- grarn per game was enforced. All games were run with Limte(]), and any program using more than 900 seconds of user+sys defaulled the game. Fortunately, no-one ran over, allhough rdr did finish his second game with only 6.5 seconds to spare! The final scores: rdr (Dennia Rockwell, Duke University) 4 points; bill (Bill Reeves, University of Toronty) 3: td (fom Duff, NYIT), allan (Atlan Wilkes, Princeton), sask (Lar- ry Custead, Universily of Saskatchewan), and ber (lrian Redman, Hell Labs) pit 2 points; uwo (team of three from the Universily of Western Ontario) 1; psl (Peler Langs- ton, New Permanent Wrinkle) 0, Thanks Lo all contpelitors and speclalors for a great tournament. See you in Boulder? [Rumour has it that Bill Reeves is’ planning on holding the frst meeting of SIGREV/USENIX in Fiji in the summer of 1980.] FRIDAY MORNING Session 6: Interesting projects on inleresting UNICES Chair: Hob Hudyma, University of Toronto Speaker 26, at 9:15 Software Regiater proposal Lynn Brock Computer Corporation of America 1600 Wilson Bivd., Suite 903 Arlington, Virginia 22208 Washington, D.C. The Computer Corporation of America does UNIX work as a contractor for ARPA in cy- bernelics tachnology and the like. CCA has proposed to ARPA, and expects funding by October for, a "UNIX Software Register", to be known as USR or perhaps /usr (groan}, an on-line database of information about UNIX-related software. Both commercial and non-commercial materials would be involved, and the list would include programs writ- lenin and compilers for C, regardiess of Lhe machine. The proposed registry would be accessible via an INWATS (900 exchange) [ree phone number, as well as on the ARPAnet and perhaps TELENET. Users would be able to dial in and access information, enter new information, and update infomation entered by them. The entire database would be maintained by the users, with Lhe administrators merely keeping an eye aul for irregularities. USR would nat distribute any software; it would be up to those wishing to get it to contact Lhe individual suppliers. The suggestion was fairly well received by those present. In the light of Lhe crealion of the USENIX association, it would be appropriate for (hese Lwo groups to get Logether to work out how best Lo handle the dalabase. 2h} UNIX Summer Conference June 20-23, 1979 Speaker 27, at 9:30 : Database for a Micro John Kornatowski Compuler Syslerms Research Group Univeraily of Torenlo Yoronto, Canada M55 1At MRS ia Lhe Micro Relational System for information retrieval on UNIX and UNIX-like sys- lems. Ik is designed for fairly small data bases (up to 10 megabytes). It features: easily-defined data bases, a powerful retrieval facility, simple data entry and interac- live data modification, MRS is fully interfaced to UNIX and has been Lested in actual applications. MRS runs on UNIX, Mini-UNIX and LSX, and hay been tested by the designers on LSI-11's and PDP-11'’s of evary size. The dala can be stored on anything from a floppy lo a big disk. The Lotal size of all programs in the package is about 300 blocks, so users can al- locale whatever space they have left for data. John gave us a sarnple session wilh MRS in his slides, to be followed by hands-on demos Friday and Saturday al Lhe CSRG UNIX lab. Simple techniques for data entry, retrieval, updating and manipulation were shown on a rated list of Toronto restaurants. An applicalion being worked on using MRS is the FORM system, designed for handling oflice forms ta be generated and dealt with interactively or automatically. Such opers- tions as filling in a form, copying it, mailing it, discarding it, filing il, atlaching it toa dossier and making an audit trai) of it can be performed. The FORM system is still under development. MRS is available for a $200 distribution charge from the MRS Distribution Manager, CSRC (address above). A copy of your UNIX or Mini-UNiX license and a signed University of Toronte Software Release Form must be provided, (See how important forms are?) Speaker 26, al 9:55 Mini-UNIX on the LSI-i1 Hob Hudyma Computer Syslems Research Group University of Toronlo ‘oronlo, Canada MSS 1Al Rob picked 6 stot in the session he was chairing to tell us how to adapt Mini-UNIX for the LSI-11 and up. The minimeom requirements for the system are: the LSI-11; Extend- ed Instruction Set Chip (optional but. preferred); Serial Line Board DLJ-1 I: 28 Kwords of memory; and al least two ARD or equivalent floppy disks (double densily, DMA, 1200 §12-byte blocks). : i can end up with is a single-user system (it could accomodate 2-3 users per- eae anal ies wilh a fast disk like the RLOL or RKOS}; 18 Kwords of user- addressable meniory (gust enough Lo hold Lhe C compiler); and a system that can han- dle practically all Lhe usual UNIX software (nroff, yace, V6 ce, ed, etc.}. Forks are sup- porled, and pipes arc simulaled via temporary files. The system can compile a eae C program lo acoul wn about do rinutes. Its execulion limes compare favourably wit ISX (which can, however, rus programs aboul 4 Kwords larger). Rob now presented the “lostant Do-It-Yourself Mini-UNJX-on-the-LS1 Kil": yy. UNIX Summer Conference June 20-23, 1979 1) Write a device driver for the AED floppy disk. 2) Replace all references lo PS in the system source by subroutine calls. Supply the appropriate routine in mcb.s. : 3) Eliminate clock references in mein.c (these screwed up the LSI for some obscure reason). 4) Modify init.c for single-user only: remove references to checking the console switches. 5) Change the swap strategy to inctude a “swap flag” so as Lo prevent swapping while a swap is in progress on those slooooc00w floppies. If you send a copy of your Mini-UNIX license, Lhe source for al) this is available wilh the MRS distribution tape (see Speaker #27 above). ; Speaker 29, at 10:15 Real-Lime Data Collection and Failure Anatysin Neil Groundwater Analytic Disciptines, Inc. 8320 Old Courthouse Road, Suite 300 Vienna, Virginia 22140 Neil spoke about work he did when working for the New York Telephone Company on UNIX starling in 1972. An 11/20 was used for analysing failures in the telephone nel- work with a view to locating areas wilh recurring problems (which would indicate some defect such as water getting into a phone cable), Neil described the trouble reporting and analysis phases of the operation. Because of the quantities of data continuously being collected, it was neceysary to filler out unim- portant information. As a result, an “alerter" was written, which processed masses of data. The alerter would take note of certain problems in Lhe system; any possibly criti- cal problem would be sent to a monitored terminal immediately, The system is now in use at N.Y. Telephone, collecting data from 96 simultaneous in- puts, The dala is manipulated by ATOM (Analysis Tool for Maintenance), 8 grapb- criented series of filters. The system runs now on N.Y, Tel.’s 11/70's., but is not avail- able to those outside the Sell System. coffee Speaker 30, at 11:00 How to get more out of your 11/70 Dan Giclan New York Telephone Company 375 Pearl Street New York, NY 10038 Dan described N.Y. Tet.'s collection of 6 11/70's, each with 1 megabyle of memory and dial-in/dial-out facilities. Running USG-3 UNIX, they have hit a fimtit {per system) of 130 simultaneous processes, 300 open files, 320 inodes and JO buffers, Jo gain more throughput they bave inplemenled certain changes: : 1) a swap device driver thal lets the swap space size vary dynamically by overRowing onto another free device when the swap arca is Cull; : ‘on UNIX Summer Conference June 20-23, 1979 2) checks for swaprnap and coremap exceptions (so that when the system crashes you at least. know why...)s 3) the DU-T driver and tly.c were changed so that in raw mode, the "kill" position {n the byLe table is used to denote a “message lerminator’: 4) RPOG drivers were changed by laking an RPO4 driver and adding a search capability; 5) disk space assignment was improved; 6) moda were wrillen Lo speed up Lhe L.P-li driver In design are a replovement for Lhe RSW04, hanging memory off the MASS BUS; and 5 “gmart" DH-VL driver with a 280. Dan had some recernmendations as to what hardware should be obtained when funds are limiled, In decreasing order of priority, they are: 1) 256 Kbytes of memory with FP11-C floating point box; 2) an independent swap device: Y) another 256 Kbyles of memory (Dan differs from the Dell people on this point); 4) separate controllers for the disks (or hang the TE-18 Lape drive off the UNIBUS and [ree up one more of Lhe 4 high-speed slols on the 11/70): 5) add a “smart” dala handler (belter multiplexer); 6) if you STILI, have money to burn, get another CPU and split up Lhe tasks. Speeker 31, et 11:20 Real-time data gathering on UNIX Frie Ostrum Neurological Control Systems Lab Department of Neurology Carnegie Metlon Universily 5508 Walnul Piltsburgh, Pennsylvania 16232 Eric Ostrum talked about the activities al NCSL in the field of research into brain func- tions and patterns and neurological diseases. UNIX is used [or real-time data galbering of eye movernents and gail pallerns. NCSL has developed the following packages: AD (Analog-to-Digital device driver); RA (read and interpret analysis data): PA (ptot analysis data); EA (edit and manipulate the data), The system is available for $50 Lo educational users ($100 to commercial users) on a distribution Lape. Separate ! and D space is not required Lo run it. Speaker 32, at 11:35 Networking at Purdue Hitt Croft Electrical Rogineering Dept. Purdue Universily West Lafayette, Indiana 47907 Bilt Croft has developed a fairly complex networking facility at Purdue, where they have two 11/70's and lwo 11/45's. The 70's are connected to each other by DMC-1!, as are the 46's. The finks are about 1 megabaud bandwidth. A sustained end-to-end throughpul of 250 Kbaud is about the sarne load on the CPU as a disk file copy. The networking package fealurcs arbilrary interconnection: many individual connec~ tions are mulliplexed Wirough the same physical link rather than spooled. As a result, -2b- UNIX Summer Conference June 20-23, 1979 e reasonable amount of interaction is available, wilh simple user commands for virtual termina! connection (con); remote process execution under a “connected shell” (csk); and file/directory transfers. The user commands are all based on a function library which can be used by any C pro- gram to make one’s own network connections. Functions exist for connecting, discon: necting, transferring files, reading, writing and signalling along the network. The net looks like a UNIX device driver: slty's are used lo manipulale connections to other hosts. A connection ia specified by 4 byles (local-hosl, local-socket, foreign—host, foreign= socket: sockets are ARPA-style sockels). Many connections are allowed to the same host/socket pair as long as the 4-byLle name is unique. This vastly simplifies the establishment of connections, The whole package is available free as part of lhe Purdue distribution tape. Speaker 33, at 11:46 Networking at NYIT Bill Lindemann Computer Graphics Lab P.0. Box 170 Old Westbury, NY 11568 Bill spoke about the networking at NYIT, where an 11/35 is used as a front end for mast of the terminals. Connections between machines are via DR-f1C's which makes it seem as Lhough there are serial lines. As Tom Duff mentioned (see Speaker #21), there are Jots of machines to be networked. Speaker 4, at 11:53 - RT/EMT: an RT-1! emulator on UNIX Mike Tilson Human Computing Resources Corporation 10 St. Mary Street Toronto, Canada M4Y 1P9 (416) 922-1937 Mike Tilson spoke about HCR's RT-11 emulator. This system actually does two separate things: run RT-11 binaries unchanged on UNIX; and run an RT-11 command interpreter wilhin UNIX. There is a need for such software because: (a) UNIX provides a good environment for the development of RT-11 software; (b) RT-11 applications prograrns can be brought up very quickly under UNIX; {c) there are atways users who, for reasons of laziness or oth- erwise, do nol want a change in their “operating system” at the aurface level. < RT/EMT system overhead is comparable to RT-11, The features of the ¥3H SJ monitor are supported, and the system runs in user mode, so that minimal or no changes to Lhe UNIX operating system‘ are required. The RI'-11 file system is simulated with UNIX directories conlaining ordinary UNIX files, so UNIX file manipulation prograins may be used usefully. The ru command will run any RT-11 binary, such as MACRO, LINK, LIBR, FORTRAN, PATCH, OMSI, PASCAL, EDIT, or TRCO (wilh the V2 screen). The caummand vfeom puts -24- UNIX Summer Conference dune 20-23, 1979 you into the RT-1t world with the exception that the keyboard-editing conventions of UNIX rather Lhan of KV-11 are supported, {Since many UNIX installations now use the DEC conventions anyway, this is hardly a problem.) The ermulalor works by loading itself into core in a large user area and copying itself lo the Lop. Il lakes aboul 7 Kwords of memory in 26 Kwords of user space. (If thrashing is a problem, il can easily be reconfigured for a smaller memory size.) Other than the startup overhead, the emulator is comparable to actual RT-11 in speed. RT/EMT is available wilh full source and documentation, along with a year's worth of bug reports and updates, for US$1950 lo commercial users ($540 for each additional CPU); the price for non-profit educational institutions ia US$A10 ($315 for each addi- tional GPU). There is a small additional shipping charge. Speaker 35, at 12:20 Software Tools Users Group Report Dennis tlall Lawrence Berkeley Lab University of California Berkeley. California 94720 Dennis summarized the mecting of the Software Tools Users Group held on June 19, 1979 at the Westbury Hotel, Toronto. Of the 96 attendees, 48 were Software Tools types only and 50 were UNIX people as well. Dennis will be writing up the meeting in 8 forth- coming issue of ;login-. LUNCH FRIDAY AFTERNOON Session 6: Improvements to the UNIX Operating System Chair: Tom Buf, NYIT Speaker 36, at 1:45 Anyone doing system performance monitoring? Phil Poulos Computer Systems Research Group University of Toronto. Toronto, Canada M5S 1A} Phi) announced that he is doing some work on systems monitoring. He would tike any- - one who has done evaluation and/or prediction of UNIX system performance to contact him, -25- UNIX Summer Conference June 20-23, 1979 Speaker 37, at 1:45 ; USENIX Committees Lou Katz (speaker #13) announced that committees have been formed wilh the follow- ing chairmen: Site committee {for next surmmer's meeting): Ben Woznick, BBN, Cam- bridge. Agenda of the Boulder meeting: John Donnelly, NCAR, Boulder. Tape distribu- tion: Reidar Bornholdt, Columbia University, New York. The Nominations committee for the USENIX Board of Directors has not been formed yet. Speaker 36, at 1:47 11/40 Kernel Multiple Address Space Alfred Whaley Department of Computer Science University of Illinois U-C Urbana, Illinois 61801 The UNIX installation at Urbana ran into a shortage of buffer space and lack of room for device lable entries. The problem was caused by "too good !/0 equipment”: one fast powerful multiplexor which handles 128 DMA full-duplex channels, serial and parallel lines, printers, readers, and machine-machine communication. The immediate goal was to put more code and data into the kernel than will normally fit. What was developed instead was a package of sollware which makes it possible to reconfigure the syslem easily wilh any space problem. There have been alternpts al this sort of thing before, but MERT is "too much and loo stow", and the Calgary buffer system lots one move only buffers oul: when those are gone, then what? The multiple addressing system developed works with separate "kernels", each with its own virtual addressing, text, data and BSS. Into “SYS” goes all the usual stuff from the sys directory (sleep, wakeup, system calls, scheduler, etc.): “A” devices are block dev- ices such as disk drivers; “C” devices are character devices such as mern, tly's, and others: "BEGIN" contains the main() routine. A few programs had to be changed to correspond to the kernel changes, notably ld, which now has a -v flag which combines wilh the “-o oulfile” option to split up the kernet into its various pieces. A change Lo cpp (C preprocessor) gets around having lo modify existing C source to run under the new kernel. The system seems to have worked a0 far. It's.available on tape from Alfred Whaley. Speaker 39, at 2:05 UNIX for 1100 Users? 7#il lan Johnstone Australian Graduate School of Management University of New South Wales P.O. Box 1 Kensinglon 2033, Australia The University of New South Wales has been running UNIX for four years {since Version 5). They have 15 PDP-t1's running UNIX. The particular sile which lan described is an 11/70 wilh: FP11-C floating poinl; 640 Kbytes of core, lwo b4 Mbyte UNIBUS disks (which are Ampex DM9100 and NOT recommended); eight DA-11 multiplexers; 40 VIU's at 2400 baud; 10 DECwrilers; and lots of line printers, curd readers, Lapu drives and so -uG- UNIX Summer Conference June 20-23, 1979 on. The system currenlty supports 1100 (yes, eleven hundred) users with 13,000 con- nect hours per month with general access only on weekdays. Between 7 a.m, and 8 p.m., at least 37 terminals are logged in 50% of Lhe lime; 43 Lerminals 25% of the time. Most of the workload consists of undergraduates using the Berkeley Pascal and BM. To support this kind of tse, two types of changes had to be made; those to accomodate a huge user populalion; and those to improve response wilh mony simultaneous processes, Changes to accomodate 1100 users include: a restructured passwd file, in binary rather than ASCH, with extra information slored for each entry; resource control (disk space quotas, process quotas, terminal reslrictions, terminal booking. and proper account- ing); dump and reslor [or a large file system; error logging in a file rather than the con- sote; recovery from power fail and parity errors; dcopy, a disk compactor; ond printf as asyslem call. : Changes Lo improve response include: an unrestricted number of buffers with faster buffer lookup via hashing and a poo! of headers for raw 1/0; a fully optimized DZ-11 driver; bulter process handling; disk driver optimizing: reduced swap activity; im- proved scheduling: and lols more. A comparison between the virgin V8 UNIX and the current UNSW UNIX shows: maximum number of simultaneous users increased from 37 (with intolerable response times) to 44 (with satisfactory interaction); average Lime to compile and execute a 2-second Pas- cal program down from 158 seconds to 32; compile and link /unix (on an idle system) down from 10 minutes Lo 8. The first two results were obtained on a fully active system under norma! operaling conditions. System monitoring shows that the percentage of CPU time spent in “user” rather than “sys" mode has increased, on average, from 15% to 31%, The entire UNSW system-improvemenls package is available on the conference distri- bution tape. The file /usr/sys/defines.b contains full information about the changes. Speaker 40, at 2:30 Another large UNIX system George Goble Flectrical Engineering Department West Lafayette, Indiana 47907 Ceorge described the 11/70 at Purdue with a user population of 1400. Last year the system supported 83 al once al Lhe worsl lime: this fall they expect 60-85 simultane- ous users. With Lhal amounl of computing to be done, compiles and nroffs will probably be forked off Lo olher machines wilh Bill Croft's networking {see Speaker #32). Be- cause of the number of users, GID'’s are not uscd. The kad address is extended to fit in extra buffers, procedures und so on: belween BSS and the kernel there is a new area, slurling al virtual 120,000. This whole region is less than 64 Kwords. The system has a dual swap Jevice, with a fast primary disk (RSO4) and slow secondary swapping. Small progroms are, of course, swapped on the fast disk. There is a new system cath, meer(I), which will let one change the priority on another process. The sel-gid bit us used (since Lyere are no groups) as a “game” bit — the pro- cuss runs al rock-botlore priority, There is also a “research” bil, for programs run in background mode which compule only whea Lhere is no interaclive process running. Power failure recovery has been implemented. Automatic file syslem recovery is per- my UNIX Summer Conference June 20-23, 1979 , formed by bootchk, e program written in 1977 by Mike Tilson al the University of Toron- le. The Purdue peopie have had a lot of problems with Lheir UNIDUS disks and recommend that one should never pul 4 disk on the UNINUS. Use the MASS BUS. The entire collection of syslem changes, atong with lols of other goodies (including the MiT dungeon) is available from George for free. Send him a tape and return postage if you want a copy of the Purdue distribution. Speaker 41, at 2:50 Office Use of UNIX Waller Lazear Air Force Data Servicea Center The Pentagon, Room 1D888 Washington, D.C. 20330 (202) 695-6161 The Air Force Data Services Center has four 11/70's for administrative support, han- dling text processing and similar tasks. The machines are up 22 hours a day, 7 days a week running prectically nothing but nroff, troff and editing. Each 11/70 runs VB with 40 simultaneous users. In addilion, there is an 11/95 for the programmers to play with, which has compilers and so on. All machines are equipped with auto-dislers, so they cen call each other or anyone else. Waller described modifications to UNIX made at AFDSC. One such mod is the proper implementation of exit codes, including a specific code for “bad format”. When the shell receives this exit code from a process, it looks up the correct usage in Lhe manual sec- tion and informs the user. Furthermore, 8 consistent user interface has been developed for argument formats in ali programs. Other changes include: a read-only root file system. separate from the user file system; a system call telling you what machine you're running on; and a stialegy for solving the problem of bad disk blocks, particularly in lhe swap area. The last 48 blocks of each file system are reserved as “alternates”, and a table in the super biock (the last 48 words) holds the bad biock list for that file system. Mkis has been modified Lo read and write every block in the file system, note all Lhe bad blocks, and only then mountit. As a result, mkfs takes about 45 minutes lo make a 56 Kbyte file system; however, once made, file systems run much better. The critical area is reduced Lo only two blocks {boot and lhe super block), and even programs which know aboul the inlernals of the file system (tike df) will still work. Since installing the change, a total of 18 system- months have been logged on three systems without disk error. The overhead for the process is leas than 1%. The collection of AFDSC mods is available from Walter on a distribution tape. coffee - ti UNIX Summer Conference June 20-23, 1979 Speaker 42, at 3:50 Implementing C and UNIX more efficienlly Carl 0. Howe Bolt, Heranck & Newman, Inc. 10 Moulton Streel Cambridge, Mass. 02199 (817) 491-150 x3642 Carl deseribed the environment at BAN, where they have 6 DEC-10's going full blast (and work for lots more). They have an 11/70 running UNIX with 150 users. There is, however, a demand for a machine with the performance of the 11/70 (running UNIX) at much less cost. The BBN Computer Company (a subsidiary of BBN) has designed the Microprogramm- able Building Block (MBB), a fast microprogrammable processor to be used es the basis for special purpose computer systems. This processor features a 135 nanosecond mi- crocycle time and up Lo 1BK of 32-bit microcode memory words. The machine also has 1024 fast hardware registers, Hs architecture ‘makes emulation of already-existing in- giruction sels easy and efficient. Furthermore, it is inexpensive lo build and easy to service. The MBB has been built and is being used for other projects, It emulates the Honeywell 416 one and one-half limes faster than the 316 runs its own code. The MBB has a macrocede which is distinct frorn its microcode. The macrocode is an inlertmediate code produced by the compiler. It is not designed as a users’ assembler and hence is often ugly but efficient. The objective was to optimize the instruction set in the direction of efficiency at Lhe expense of aesthetics. ‘ In constructing a C compiler for the MBB it was desired to avoid some of the problems with the PDP-11, such as Lhe 16-bit address limil, Lhe limit of Lhree hardware registers available to the user, the slow speed of subroutine calls and the inefficiency of local variable addressing. These problems have, it seems, all been solved. With 20-bil words, larger processes may be Tun on the machine. The abundance of hardware registers makes register saving unnecessary (except with large amounts of recursion, in which case 64 are saved at once). The macrocode can call the microcode to do certain things; there is Mlexibilily in deciding what will be implemented in microcode. And the micro- code has an instruction set with many different addressing modes, so that doubles and longs can be dealt with directly. The assembler for Lhe MAB was written in 3 days (using yacc). The compiler Is being written and compiled with Lhe phototypesetter-version of cc. The microcode has been written and the hardware already exista although it is not yet generally available. Onee C is working. RAN will begin implementing UNIX on the MBB. UNIX will feature: a 20-biL MUD processor; 120K 20-bit words of memory; two disk drives of 40 to 300 mega- byLles; one Uline serial mulliplexor for Lerminals; and @ compute network interface. It is expected lo be running by the end of the summer: disk interfaces and memory management have nol been duspned yet. White nothing can be promised as yet, it is expected thal BN will be marketing lhe MHI under UNIX by the summer of 1960, and that it will give Lhe performance of an 11/70 for Lhe cosl of an 11/34. -20. UNIX Summer Conference June 20-23, 1979 Speaker 43, at 4:10 - UNIX on a UNIVAC ¥77-600 Harold Pierson RLG Associates, Inc. 11250 Roger Bacon Drive, Suite 16 Reston, Virginia 22090 (703) 471-1106 Harold Pierson described UNIX v77 on the Univac as “an experience in portabilily". RLG has been doing the project for Sperry Univac, who took over Varian and are producing the “Varian V77" as Lhe “Univac V77". The writing of the C compiler was subcontracted to Cassandra Inc. (Bob NcClure). UNIX is now up and running moat of PWB on the V77, Harotd discussed some of the problems resulting from the fact that the V77 is a word- addressed machine. Assignment conversions, parameter passing and word/byte reloca- tion in the foeder al! presented a problern. The solution implemented was to pass al- ways by byte pointer when calling subroutines. Within routines, word pointers are used. The loader was modified to cope with this. The were other assorted incompatibilities. The final result was code that is larger by 20-30% on a machine which is much more powerful and faster than the PDP-11. Addi- tionally, because of 512-word paging, there are 16 address spaces in core; inslead of just the kernel, the supervisor and the user you can actually have 15 users running in core. At present UNIX (the real thing, not a look-alike) is running multi-user on the V77-800. RLG is handling the program development and administrative software: Sperry Univac will be marketing it in the future. Other developments to follow are a version for the v77?-800 (with separate I and D space), multiple address spaces, scaller page swapping and micro-coded sasists. Speaker 44, at 4:35 An Accounting System for UNIX Robert N. Jesse Johns Hopkins University Electrical Engineering Dept. Barton Hail Baltimore, Maryland 21218 The 11/45 at JHU ts being used by students, which aulomatically creates a hostile en- vironment. All the Uhings Lhat one can normally get away with on UNIX must be moni- tored and controlled. The /ete/passwd Sle has been left intact, but a parailet file, fete/uif, contains more in- formation for each user; the encrypted password; Lime and tty of last login; cpu and connect Lime used and quotas; block-1/0 and linc-printer-page count and quotas; togin terminal restrictions; and special permissions and options. Monitoring of system usage is done by a new prograrn, inil2, which ceceives the termmi- nation slaluses of all prograins run by the used on a pipe fro inté. The information is passed with a new system call, waitinfo(}, which provides the proce and user slruclures of Lhe deceased prucuss. When the shell exits, init calls quota, which will nol let Lhe user tog out until he has reduced his disk consumption below his quota. (if the user ts on a phone line and hangs up, be will nol be permitted Lo log in again by phone until he -ts0 - UNIX Summer Conference June 20-23, 1979 has discussed his problem with the syslLem administrators.) Another system call writ- ten for inil2 is getppid(i]}, which returns the PID of the parent. The program qcheck is run periodically Lo check for people owning files in others’ directorics. When the distribution of this system is ready, it will be available to educational UNIX licensees only. Those wishing to obtain it must sign 3 non-dislribution agreement (the JI[U people want all copies to come Trom them), Spesker 45, at 4:55 A High Performance UNIX System Mike Muuss Johns Hopkins University Hleetrical Engineering Department Barton Hall Haltimore, Md. 21218 Recause it waa late. Mike spoke very briefly about the changes made at JHU to improve performance. “Performance” in this context is not just good response Lime; it includes: (a) excellent response Lo small interactive processes, while keeping the CPU as fuit util- ized as possible: (b) good security in a hostile user environment; (ce) detection and avoidance of certain hardware failures; {J) good data inlegrily, especially ina hostile hardware environment; (c) a more pleasant user interface: and (f) accountabilily and restrictions on resource consumption. The JHU people have divided Lhe solution up into four areas: (1) resource measurement and accounling; (2) teletype driver; (3) security and integrity; and (4) performance mods. Resource measurement. See Robert Jesse's talk (Speaker #44) for a description of the accounting system. The file /dev/data contains dynamically modified information about processes running, system load and performance, and resources. Teletype driver. Lots of enhancements increasing performance; others adding a better user interface: ctrl-r to retype Lhe buffer; page mode; stati mode; character rubout via BS-SP-BS; ctri-t Lo print status of process being wailed for. Security and integrity. Special files and SET[UG}ID operate only on Lhe root file sys- ten: front panel interference protection (); /dev/error and /etc/logerr for logging and recording history of non-falal errors, atong with interpretative software; memory test and lockout at boot time and dynamic memory tesling; a slack limit regisler; use of IDB - kernel runs a lrae split | and ) space, with the I-space write protected; fixing of assorted errors and Bugs in the system source. Performance mods, Process dixpalehing has been fixed up by implementing full pro- cess queues. The resull has been much faster reponse Lime for editors and other small CPU usage programs. (f-second lypicat reponse. with 30-t5 users on the 11/45), Swap- ping increases, however, so i is advisable lo have an RKOS dedicated Lo awapping. In the future, 1/0 will be inlegraled into the dispatching priority. The priority evaluation doue now keeps teack of how long Lhe process was in core before and takes into ac- coounl Unie size of Lhe proccss, -31- UNIX Summer Conference : June 20-23, 1979 Some fatal bugs in the system source were also found, but Mike could not discuss {hem publicly because of the presence of “aliens” (people not bound by 6 Western Blectric non-disclosure agreement), Speaker 46, at 5:15 UNIX on the IHM 370 Steve Rellovin Universily of North Caroline New West Hall 095 A Chapel Hill, North Carolina 27514 Steve spoke about the Triangle Universities Compuling Center (TUCC) group. which has been running TSO for six years and would now tike to move to a Lime-sharing systern. fhe plan is to move the commend level of UNIX to an IBM $/370-166/1), and thence to the Amdaht 470. The reason these machines have been chosen is thal their hardware and hardware service are feit to be far more reliable than DEC. The 370 UNIX will be CRT-terminal oriented. It wilt be an FBCDIC system (agh!). Assort- ed suff will be imported from other systems. It will be implernented using VM-370; the goat ia to have 100 simultaneous users, gach with a virtual machine. The UNIX file sys- tem, pipes and shell will be kept intact, It is hoped that in a year’s time the system will be operational and capable of handling 20-30 simultaneous users. By two yeors from now it should be on the market. Steve has no idea yet as to what the price will be, or whether it will be available cheaply to educa- tional inatitutions. Does anyone have a DX-11 driver for Steve? SATURDAY MORNING Session 7: Educational uses, and what's happening at ?.... Chair; Sandra Wright, DCJEM Attendance was low at Saturday‘s session (about 35 inslead of 350), These notes were taken by Sandy Wright end expanded by Dave Sherman. Speaker 47, at 9:05 UNIX in the Undergraduate Lab Environment Don Schertz Bradiey University Dept. of Electrical Engineering Peoria, Illinois 61625 UNIX at Bradley is runnning on an 11/40, where haif of the connect time is spent on software development for a rnicroprocessor. RJE into the university’s CYBbKR is planned in the near fulure. They are running Mini-UNIX from the Johns Hopkins Univer- sity, which is a good package. Communications lo micros is being worked an. Curreatly RS-292 loops are used, but there will be @ move to the M086 on parallel ports. Wilh a variety of cross assemblers available, the micros are downloaded with binaries. - is UNIX Summer Conference dune 20-23, 1979 Speakor 48, at 9:15 UNIX in a large educational onvironmest Nike O’Deti University of Okiahoma Engineering Computer Network Engineering Center Norman, Oklahoma 73019 The University of Oklohoma’s 11/70 has 384K of core with a fioating-point processor and RMOQ's on separate controllers; four DH-11's (6 dialup lines), and a user population of 500 to 700. The maximum number of users supported simultaneously |s 37. They use Lhe Purdue tty driver with some moda for weird terminals. Work is being done on RIE to the IBM 970/156; they are trying to get the PWE RJE software working. Speaker 49, at @:50 QED, or The Little Bd That Grow Robert Pike David Tu brook Caltech 220-47 Bell-Northern Systeme Research Pasadena California 247 Brunswick Avenus USA 01125 Toronto, Canada M6S 2M6 CSRG at the University of Toronto bas bad ed hecks for a long time. Years ago, Tom Duff and Rob Pike added assorted useful features; these were expanded and rewritten recently by Hugh Redelmeier. Fd at U of T now baa the following features: o singte er- ror charncler following the ‘7’ indicating what type of error waa found; '* ae an easier way of typing 1.8 (or N® for N,$}; ‘&’ as e@ simple form of page-oriented addressing: file saving upon a hangup signal (in “@lename:sav’}; extensions to regular expressions for easier matching: s2/xxx/yyy/ for the second occurrence in a line (and so on); ‘ed file’ for changing the working directory; *«' command for character editing; ‘j' to join lines; *u' to undo Che last substilution; and others. : Qed was originally written by tom Duff by taking the code for ed and adding multiple buffers, number registers and assorted other goodies. Improvements by Hugh Redel- meier, David Tilbrook and Rob Pike have taken place over the intervening four years. eoek more enhancements are still in progress, qed is already a powerful editing ool, The principal features of qed are multiple file/mulliple buffer editing, macro capability and an enhanced intertace to the shell. For experienced programmers, these festures can greatly simplify dificult or repetitive text editing tasks, and make iL easy to develop special-purpose editors that have knowledge about the structure of the text they are working with, As much as poasible, ged is a rigorous superset of the version of ed running at U of T. The couple of minor differences are considered to be improvements over the original ed, bul it was decided to leave ed‘s behaviour as it was to avoid incompatibilities with eds running elsewhere (specifically the PWB {Bench) version). The main extension in ged is the presence of 58 buffers, labeled by upper and lower -33- UNIX Summer Conference June 20-23, 1979 case alphabetics and four other characters which are reserved for several purposes. All normal ed commands work within a buffer, so thal internal lo o particular buffer the user sees what appears to be a regular editor. The only exceptions to this are the move and copy commands, which allow inter-buffer transfers, and, of course, the com- mand te change the currant buffer, At any time when input is expected, be it command or append input, the contents of a buffer may be inserted using a simple escape sequence. As well, ged provides registers which are accessed in a similar manner, but which have some commands to deal with them directly rather than by changing the current buffer. The user, at startup, can specify the name of a file which is to be read into a reserved buffer and executed be- fore reading commands from the terminal, so thal he can preset registers and buffers to contain useful commands and text. For programming purposes, qed also has sirmple contro! structures and message print- ing capabilities, so that it can be used as a rather simple programming Innguage. The implementers of qed claim that their version is considerably better human engineered than previous versiona of qed on which it is based. A tape with source and documentation for qed. U of T's ed, and a few other useful pro- grams, may be obtained from Robert Pike. Please send « blank tape, address label and return postage. coffee Spoaker 50, at 10:50 News from the U.K. RP.A. Collinson University of Kent Canterbury CT2 7NF Kent, England The are 15-20 UNIX sites in the United Kingdom, all with educational licenses. Most of them are 11/40's, although there are Lhree 45's and one 70. They are in touch with Eu- ropean UNIX sites in Holland and France. Only Glasgow bas tape drives ~ RKOS‘s are used for the most part, which haa the side effect of making distributions more selective and carefully thought out, As there are only perhaps five system backers in the whole of Groat Britain, most loca- liens stick to the conventional sluff, using "Standard C" (whatever that is), and keeping any changes to system source code well documented. Languages being developed in Britain include: BCPL (two different versions of thts port- able high-level predecessor to C); Modula (at the University of York: Jan Cottam); PDP11 (Sussex: Steve Hardy); and micro assemblers in various places. An implementa- tion of Aigol 66 ~ « major project - is underway at St. Andrews, The UK UNIX User Group has been in operation for two years. It is a DEC SIG; it pub- lishes a {theoretically} quarterly newsletter. Glasgow acts as a software distribution centre. The users’ meetings draw about 50 people. UK/UUG is prepared to exchange tapes with USENIX. -34- UNIX Summer Conference dune 20-23, 1979 Speaker 51, at 11:05 . What's happening at BBN Carl D. Howe Bolt, Reranek & Newman, Inc. 10 Moulton Street Cambridge. Mass. 02139 (817) 491-1850 x3642 Carl briefly mentioned a number of the projects underway at BBN: the new C machine {see speaker #42); a C compiter for the 28060 and the DEC-10; manuat revisions; ex- tended documentation; non-blocking !/G; ports, nelworking; a new Rand editor — for CRIs with cursor addressing and shared text; word operations user configurable; cross-nel debugging; screen managers under investigation; ARPAnet work: work on bigh bandwidth local networking is starting. The BBN UNIX mods distribution tape costs $300. Speaker 52, at 11:15 Tuning PWH UNIX George Pajari © GTE Automatic Electric (Canada) Itd. 100 Strowger Bivd. Brockville, Ontario, Canada K6V SW8 The RIE modification under PWB needs rewriting. Kernel modifications have also been made at CTE: a query feature which gives utilization of tables; and gmon, which keeps track of high- and tow-water marks in tables, Knowledge about these watermarka al- lows sensibte settings of tabie sizes, and also prevents some crashes by detecting table overflows. ‘ This software might be available. Contact Paul Hart at GTE (address above). Speaker 53, at 11:26 Sereen editors Mark Pearson Yourdon Inc. 1153 Avenue of the Americas New York, NY 10038 Yourdon has a screen editor which works ass front end to ed. It supports dumb termi- nals. The backslash character is not speciat, The cursor can be moved a word at a time. Jt should be able to Iront onto Lhe Toronto ged, and it was agreed by Dave Til- brook and Mark Pearson to try and do so. -~35-