IMPLEMENTATION WORKSHOP Success, It Is Oniy a Beginning

TUGboat, Valwne 2, No. 2

floating point). The input text wee very ~impleand Site Reports the output was only a set of numbers, the cantents of '&?Pa DM me. We have no device driver programs and our next UNDER THE NORTH STAR major taak is to write one or more. Potential devices at our site are a Variaa 200 dot/inch plotter, a Part I Linotron 202 typesetter, plus a Xerox 9700 page printer at a local service bureau. (The 9700 would As coordinator for the CDC Cyber implementa- be intended for proof copy of text; full math output tion of 'I@C, 1 have agreed to answer questions from doesn't seem practical on it at present.) interested Cyber sites about our progress with m- F'urther tasks are to finish the QjlC conversion by in-Pascal and with our implementation of drivers for finding out how to remove the temporary changes. the output devices to which we have access. When We believe these changes are due to incomplete im- we have succeeded in our implementation, I will plementation of the system dependent routines. We make available for distribution, for a charge, a tape will also have to convert the remaining font infor- with the following information: mation files to Cyber format. 1. m-in-Pascal source, as modified to com- Our system dependent changes produce a NOS pile successfully on a Cyber under the ver- operating system version using 6- and labit ASCII sion of Pascal maintained at the University of character codes for input and output. If we can find Minnesota. a way to test it, we can build an 8-bitein-12 ASCII 2. A file containing the Computer Modern font set. version for NOS/BE sites. 3. Our device driver(s). 'I)@is a very large program, around 98K words. 4. Relocatable binary of these files. Our site only allows time-sharing for 32K so we run I will also send a copy on microfiche of the TEX 8s a batch program. While we realiae that l)iJC Pascal compilation and cross reference listing. I is intended to be interactive, our CDC Cyber usem wilI provide sympathy, understanding, and hand- will have to live with this problem until (or if) we holding as you put 'I@C up on your system. can find ways to make our version of '&jX smaller. I have also agreed to act as g+bet.ween for other If all goes well, we plan to announce matters which may come up from time to time. availability for CDC Cyber usera at the VIM users However. I don't intend to act as a roadblock! Bob group meeting in October. We will need to write Welland, editor of TUGboat, would be delighted installation documentation and we would like to try to receive articles and letters-to-the-editor directly out our installation procedure at a nearby CDC site. from you. If there is enough time before October, we will My address is in the front of TUGboat; my phone do a conversion of a more recent version of 'I)$ number is (612) 373-4599. This number is always (rather than the December 1980 version). There are answered and I irn reasonably good about returning a number of attractive features in recent versions calls. which would be helpful. Several recently written In Part n of this report, M. J. Frisch, our im- Stanford programs are also valuable, particularly plementor, describes our current progress. However, one by David F'uchs that prints the contenta of a please note that calls should be directed to me, not DVI Rle. to hini. In summary, we really can see some light at the Thea D. Hodge end of the tunnel. Michael J. Ffiseh

We recently got a proper device independent (DM) tile output from our Pascal version of This is a version we received from Stanford in DEC&0hn-10/20 December, 1980. While we are pleased with this IMPLEMENTATION WORKSHOP success, it is onIy a beginning. Phii Shed Several temporary changes .had to be made to get l3iJC to work because our conversion is incom- An implementation workshop for persons inter- plete. Our font information file had only one font ested in installing l)jX on a DECSystem 10 or 20 which was converted to 60-bit Cyber words from will be held at 'Vanderbilt University in Nashville, 36-bit DEW0 words (containing mixed integer and Tennessee, on September 10 and 11. The course will TUGboat, Volume 2, No. 2

discuss the details of connecting T@C output devices Reijuests for Alphatype jobs will be considered, on a to DEC 10s and 20s as well as software issues in- time-available basis; for price and other details, call volved in installing w. Conference attendees will or write to have access to a system running m. For details, Ftaymond Goucher contact American Mathematical Society Phil Sherrod P.O.Box 6248 Vanderbilt University Providence, RI 02940 Box 1577, Station B 401-272-9500 Nashville, TN 37235 615-322-2951 l)jX AT THE 1981 SPRING DECUS U. S. SYMPOSIUM AMS SITE REPORT Patrick Milligan BNR Inc. Barbara Beeton In keeping with a long standing tradition, there Since our last report (TI'Gboat Ibl. 2, So. 1, was a T)$ Birds-of-a-Feather session at the recent page 48), an integrated spooling system has been DECL% symposium held May 18-21 in Miami. (In developed and installed to control output to both case you didn't know, DECI!S stands for "Digital the Varian and the Florida Data. This spooler will Equipment Computer Users Society.") The panel print stra&ht text .(ASCII) files, using the rather members were: primitive Fnts supplied with the output devices, as well as output. Output defaults to the "local" Patrick Milligan Coordinator for DEC-20 device, using device-appropriate fonts (i'arian in Phil Sherrod Coordinator for DEC-10 Providence, Florida Data in Ann Arbor), but may Barry Doherty .herican Mathematical Society be overridden; output page positioning may be ad- Hob Phillips Oregon Software justed either interactively or from a predefined op It comes as no great surprise that DEC users tion (.OPT) file, allowing each column of a multi- are interested in 'w: Don Knuth's book, column page to be treated by l)j.Y as a separate page and MET;IFONT: New Directions in Typesetting to oonserve memory space, then overlaid on output: is jointly published by the American Mathematical multiple copies, output of selected pages, deletion Society an$ Digital Press. In addition, versions of (by submitter) of a job from the queue. and other '&X exist. for the DEC-10, DEC-20, and V4X/VMS similar features are also supported. computers. '&e have acquired the capability of generating Since the DECUS Symposium was held im- fonts for the Alphatype via METAFONT (see ar- mediately after the 'l&X Implementors' Workshop ticles by W. J. LeVeque, page 39, and Ron Whitney, at Stanford, emphasis was placed on the installa- page 40). The Alphatype is being used for limited tion of TE)L. Output samples from various devices internal production. (Another composition system (including Versatec, XGP, Canon LBP-10, and is still being used for our major journals, pending Alphatype) were passed around. A free sample completion of necessary work on symbol and math (the TUG membership forin) was distributed. The fonts.) The entire two-column portion of this issue content and utility of recent TUGboat issues was sf TUGboat and the separate address list have been described. In short, a sales pitch was made for T@C, generated on the Alphatype. A number of articles TUG, and TUGboat. were sent in on mag tape, from VAX and Univac Another session on '&X is planned for the next 1100 systems as well as from DEC 10s and 20s; this DECUS Symposium to be held December 7-11 in experiment has been most successful, and we hope Los Angeles. to receive many articles for future issues on tape. Some files have also been received via a phone file- transfer mechanism (not a network). We have prepared camera copy on the Alphatype for several other organizations that do not have their Rachel Schwab own high-quality output device; for these jobs, '&jX input was provided to us on tape, and the out- exists on the DECsystem-10 at the National put proved to be virtually identical to the original. Institutes of Health in Bethesda, Maryland. At TUGboat, Volume 2, No. 2 preeent it may be used only by the DECsystem-10 ing =,The difficulties were of several kinds. A st&. Hopefully we will soon have facilities available major difficulty ia in the differing syntax of exter- so that it may be opened up for general usage. nal compilation between various Pascal compilers, a Cumntly there are four steps involved in running feature not defined in the original Pascal definition. at NIH. First one runs rn to translate text Differences in 1/0 on the host machines and in the into a DVI file. Then the DVI file is used as input ways the compilers handle 1/0 were another prob- to a program called DVIPDP. DVIPDP essentially lem. Memory limitation problems were encountered, takes DVI records, extracts all the important infor- and these were made worse by the fact that we were mation, and outputs new records that contain op altering Pascal code rather than the macro language codes along with other information. For example: 'QjX was written in. The typesetter we are using is opcode 3 indicates a rule and the information that not addressable as a raster device, and so we were goes with a rule are height and width. Opcode 2 unable to use standard W/METAFONT fonts. indicates a "delta yn item-the vertical coordinate Instead we had to write a pre-processor to convert is moved by a certain number of pixels, et cetera. descriptions of the typesetter fonts intoFformat. DVIPDP produces a "PDPn file of records like Various errors occurred in using m,some of which the ones described above. This file is transferred were traced to errors in the '@X code, others of to a PDP 11/70 by means of a high-speed com- which related to improperly formatted fonts. The munications link. A Benson/Varian printer plotter 1100 operating system dlows a user to end terminal is hooked up to this PDP 11. A driver program, interaction with a program by typing an end-of-file ~/11,was written for the 11 that takes "PDP" signal-m doesn't consider the existence of this records and translates them into scan lines. concept. Because we have no inexpensive proofing T@l at NIH has the ability to take graphs and device we had to write a line printer proofing pro- figures and place them inside a file. itself gram which has severe limitations due to the limited has not been modified to do this, but the DVlPDP range of actions and characters possible on a line and m/11 driver have the ability. Graphs are printer. We have found the overfull box messages created by using MLAB or OMNIGRAPH on the to be unhelpful in correcting justification problems DECsystem-10. A program has been written that and have replaced them with a more informative will convert a plot file into a file of alphanu- message. We have found it very difficult to set type meric records which in turn may be translated by in narrow columns with m. Details of all these DVIPDP. problems a= discussed below. As one can see by the above description, us- We sbouM all hope that the project of transport ing 'Q$ at NIH is complex. The four steps (m, ing 'QjX makes future language designers careful to DVIPDP, CLINK (transfer to the ll), and QX/11) be complete in defining language standards. Hours will need to be compressed before 'QjX can come up and days of work could have been saved if the Pascal in a production environment. This necessarily in- definition included external compilation, the default volves getting our own small computer and printer case in the CASE statement, etc. to hook up to the DECsystem-10 for usage with W. External compilation in UW-Pascal uses a ayn- NIH is a community of scientists and researchers tax very different from that used at Stanford. Any who produce technical papers in abundance. The shared types, variables, procedures, and function8 pomtial for use of '&jX at NIH could be very high. must be declared in a separate environment module. However, 'Q$ is not an easy system to learn. We Only the head of procedures and functions is in- are hoping that the &$'I$$macros will make us- cluded there: the name, parameters, and result type ing '&?la little easier, and we are looking into the comprise the head. Each code module includes a list possibility of writing some type of 'I)$ of procedures and functions which correspond to the heads in the environment. To convert into this format required a lot of hand coding. AN IMPLEMENTATION REPORT Differences in 1/0 handling required a lot of re FOR TEE UMVAC 1100 coding in the system-dependent module, SYSDEP. Bill Kelly Of course, it is to be expected that one would have University of Wisconsin-Madison to recode the system-dependent code, but it could be made clearer what part of the Stanford code We have W running on the Univac 1100 uaing refers to PDP-10 peculiarities and what is required University of Wisconsin-Pascal. This is a report on by T@C. For instance, 'QjK expects lines to be ended some of the problems we encountered in implement- by carriage returns or line feeds, etc., which is the TUGboat, Volume 2, No. 2

internal file format of mod PDP computers, In problem in trying to reduce the aim of the 'mem" the Univac file architecture, carriage returns and so array to meet a memory restriction in our compiler. forth are not preserved in the file.. This must be The ultimate solution is to distribute UMH)C with the case with many computer systems, and espe- T)jX. But admittedly this work should take sewnd cially in Pascal compilers where such system-specific priority to work on l)jX itself. A difficulty which things are supposed to be transparent. What we may be impossible to surmount with this approach did was to read a line from the Univac file, and is that the output from UNDOC may still be un- when Pascal returned true as the value of the end-of- suitable for compilation, depending on the features line function EOLN, SYSDEP appended a carriage of the compiler. For us this came about because of return. This would have been much easier had the the environment feature of UW-Pascal; other sitm SYSDEP code not also dealt with ignoring line feeds have had to hand code the default case of the Pascal and nulls, or rather had it documented and localized CASE statement. If this situation occurs, it would these local quirks. To make it easier for future sites be impractical to correct errors in the macro source to implement 'QjX, procedures like SkipTrailingStuff and all fixes will end up being done in the Pascal to ignore nulls after a line end, and functions like code anyway. It is unclear whether it is realistic to EndOfLine which would check for PDP-10 line end expect UNDOC to be able to produce compilable characters, but note that on non-PDP machines a code for all compilers automatically. function like EOLN might be used. On the Univac 1100, a user at a terminal can The fact that 'I)$ expects input files to be page type Oeof to force an end-of-file condition in the oriented is a minor nuisance. Apparently files on the terminal input to the current program. The program PDP-10 are made up of "pagesn delimited by a form must terminate without further intervention. This feed charace. This is not the case on our machine possibility is not allowed for in m, most likely be- and on maniothers. This causes minor problems in cause this concept of a terminal end-of-file does not twoVrespects:one is that if a file containing a form- exist on many computers. We have circumvented feed character on our system were read by 'QjX, this probl'em by having the InLnTer procedure in it would be interpreted in a manner inconsistent SYSDEP pretend that the user typed \end if he with our operating system. The other is that "page indeed types an Qeof. This is not the ultimate numbersn appear along with line numbers in the solution, since there are circumstances where l)jX error messages. This tends to be very confusing since will not accept a \end command, as for instance in there is no page number in the file structure, but the middle of a paragraph. To help out with this there are page numbers in the formatted output-so problem (and to make 'Q?C generally more friendly), a Univw T)jX user would assume the page number our basic macro file redefines the \end command referred to the output Ages. We circumvented this to be \par\vfill\end. This avoids the problem by changing the error message from "p.0, 1.50" to of an unfinished parkgraph at the end, but not of "Line 50". This could be built into 'I)$ by providing an unfinished \halip, etc. The suggested solution a Pascal constant "pqeOrientedn which the 'I)jX to this is to have a function TerEof in SYSDEP to implementor could set to true or false. detect a terminal end-of-file condition. On detect- Implementing w is difficult because the pro- ing this, a clean-up procedure would terminate any gram is written in a macro language which is current levels of processing, possibly issuing warn- converted by the UNDOC program into Pascal. ing messages. Without having examined the code Unfortunately the UNDOC program is not presently closely, we don't know if the TerEof idea would be available in Pascal insofar as we know. The macro the best approach to modifying the existing code serves well as structured documentation, but code, but the general idea would be the same no it is difficult to refer back and forth between the matter what the implementation. This added code macro code and the Pascal code. This is because the would not impact sites that do not have a terminal macro code is divided into sub-sections in a struc- eof concept, because the installation documen- tured heirarchy, which expand into linear Pascal tation would say to have TerEof always return false code. Also the use of macros to provide alternate at these sites. names for variables in the macro language means Ueing 2)jX with a non-meter device that a single Pascal statement may look entirely Our primary output device is a Compugraphic different in the two listings. Another drawback is 8600 typesetter. The 8600 does use digitized data that expressions involving constah are folded into to create its characters, but it will not accept user- a single constant. This makes compilation efficient defined character shapes. Therefore our T&jX font but changing array sizes difficult. We ran into this files reflect the shapes of font characters available TUGboat, Volume 2, No. 2

from Compugraphic, and our device driver outputs mand in l@C, which could dm be expressed in terme characters and commands to the 8600 rather than of more primitive commands like HORZRULE and bit rasters. We get an excellent quality of out- VEW'RULE. A user with an intelligent device could put this way, but the approach is not without its configure T@C to output the 'box' command while difficulties. other configurations could output 4 separate line One problem was in converting character shapes commands. To maintain transportability, a etan- into m-readable format. The data had to be dard skeleton for a device driver would have to be manually typed, and a program was written to trans- distributed which could break a box command dawn form the raw data into a pair of files, one for into 4 line commands at that stage. The user with and one for the 8600 driver. Until all the bugs were the intelligent machine would bypase this procedure. worked out of this program. mysterious '&X errors Alternatively, all configurations of a program Jilce resulted from improperly formatted font files. For '&JC could output the box command and distrisute example, parentheses (and other delimiters) of vary- the rest of the work to be divided between the device ing size form a linked list in the font information. At driver program and the device as the programmer first, our font preprocessor did not correctly indicate sees fit. the end of the linked list. This problem was difficult A Limitafion of l)ijX to diagnose, because it. caused a Pascal error; We hawe found one apparent limitation to TjijX: assumed the font information to be correct and so it. seems impossible to set text in narrow columna did not issue any warnings. This is as it should be without extensive manual hyphenation. One cus- for maximum efficiency of but it points up the tomer wanted to set 9 point type in 2 inch columns. usefulness of a program such as the one described When we ran QjX, we got many overfull box mes- in TUGboat Vol. 2. No. 1 which ar~tomatically(and sages. This leaves you to figure out the cause of the correctly!) converts font files into readable data and error. It may well be that there is a word that vice versa. does not know ow to hyphenate. We have devel- Using a device like the 8600 limits the portabi- oped a non-stan1 ard message for overfull horizontal lity of our 'QjX-produced documents because the boxes that prints the box as a single line on the ter- fonts do not correspond exactly to METAFONT minal, representing characters as their correspond- fonts. We have similar fonts, for instance our ing ascii character (or a question mark if the charac- English Times resembles Computer 14odern Roman. ter is non-printing), glue as a space, and hyphens but slightly different widths can lead to different tion nodes as dashes. This gives you an idea of paragraphing and pagination, overfull entries in whether 'I)$knows how to hyphenate a word at the tables, etr. We cannot use METAFONT with beginning or end of the line, and you can insert dis- the 8600 because the manufacturer provides no cretionary hyphens,as needed. However, if the word method of addressing the typesetter as a graphics that is giving trouble occurs on the previous or device, and the cost of having fonts custom-made next line, you must consult W's hyphenation des by Compugraphic is prohibitive. to find whether it knows how to hyphenate the word. An int.eresting problem encountered in using a Thus I have found myself hyphenating words at ran- micro-computer driven typesetter like the 8600 in dom, which$ a distinct inconvenience in itself, and conjuction with 'Z)E.X is the lack of cooperation be often find that there are still overfull boxes. tween the two devices. The typesetter itself has a In addition to poor diagnostics, we had trouble fair degree of intelligence, but it seems impossible to in getting to set the narrow columns at all use any of it with 'QX because of the need to make without overfull boxes. We tried adjusting the the DVI file deviceindependent: 'I&X must assume \jpar parameter to have look at more poe it is dealing with an entirely stupid device! The sibilities, and it still gave overfull boxes. Manual 8600 does justification, tabulation, automatic accent hyphenation did not work. We eventually had to in- placement, box-drawing, etc. The commands we crease the \spaceskip (or word spacing) parameter paas to the typesetter bypass all these features! The to allow a very large space between worda, and create question of cooperation between smart devices is a typographically bad output. In some cases even this difficultone, and it seems a shame that the "smarts" did not work, and a \linebreak command had to of the typesetter can't be better utilized. This is be used. a general question, not one addressed specifically At MACC we charge customera as closely as pop at that of portability vs. efficiency for a par- sibie for the resources they use, and we have found ticular device. One approach would be to have a T@C to be very 'expensive in both memory and CPU heirarchy of commands, for instance a 'box' com- time. This is not redly a surprise to anyone, of TUGboat, Volume 2, No. 2

course. But it means that. most users cannot afford (Jhrther progress on s UNIX/~at Brown awaits to run a document through '&jX four or five times to the arrival of the U. of Washington compiler.) The find the right discretionary hyphens needed to avoid accompanying article by Bany Smith explains how an overfull box. Nor is this user-friendly. you can obtain a copy of the VMS version. You There may be no easy way to solve this should understand that, in spite of the efforts made justification problem. It may be an unavoidable by Oregon Software, this is not the final version side effect of the justification algorithm used. One (in fact, Stanford has just released a new version possibility would be having a "paragraph debugger'' of m). Barry has overcome numerous bugs in that would help to diagnose the problem in a given DEC Pascal to get us to our present state, but more paragraph: e.g., point out a large word that can- remains to be done. You can help in this effort by not be hyphenated that is at the root of the prob- carefully documenting any bugs that you encounter lem. This would be a difficult program to write. after making sure (insofar as possible) that they are An siternate hyphenation routine would be another bugs and not errors you are making during your pos~ibility. Allowing Tjj?C to violate its hyphena- learning process. We have encountered our share of tion rules if necessary to justify a paragraph, mean- both in the short while the system has been up and time issuing a warning, would at. least focus on a running at Sandia. If you fix a bug, please send your specific problem, i.e. a questionable hyphenation. fix ta Barry! rather than leaving the user guessing. The distribution tape is 9 track at 1600 bpi Ueefulnees of macro sharing and consists of about 6500 blocks. The tape 'QX in itself seems to provide a standardized lan- was made using the command MCR BCK MTAO: guage for typesetting. However, a non-compu ter TEX.BCK=*. *; *. The data can be recovered programmer will find this language difficult to use. by creating a directory "CTEXIn, setting your and will not have a ready facility for writing his own default to that directory, mounting the tape (using macros. '&X users need to do much more macro- MOU/OVER=ID MTAO:), and then recovering the files sharing as was done in TUGboat Vol. 2, No. 1. Good using macros put the typesetting capability in the reach of MCR RST *.*;* = YTAO:TEX.bck most users. There are still problems when an error It is suggested that you put all the files on the tape occurs through improper use of a macro. The diag- in the ITEX1 directory and initially use the program nostics are geared toward one with a knowledge of there. programming. One aid in this would be to provide The special Versatec driver (LVDRIVER . EXE) must 8 way for a macro to send a message directly to the be copied from ITMI to ISYSEXEI and installed. terminal. This was shown in TUGboat Vol. 2, No. 1 This is done for Versatecs utilizing a DMA inter- at! an extension added b$&e site; it seems to be a face by inserting two lines in [SYSMEISTARTUP. COM useful enough feature to incorporate into the stan- after 8 RUN SYS$SYSTEb(:SYSGFA but before dard 'Q$. AUTOCONFIGURE ALL. The two lines to add are LOAD LVDRIVER/DRIYER=mS$flsTEY:LVDRfVER.EXE and AVAILABILITY OF CONNECT LPAO/ADAP=3/HUYVEC=2/VEC=%Ol74 OREGON SOFTWARE IMPLEMENTATION /CSR=m7775lO/DRIYE-LVDRIVER OF 'J)$K FOR THE VAX/VMS Reboot the system aftkr these changes have been Monte C. Nichols made. The programs LVSPOOL and LHSPOOL spool output This VAX/T' site report consists of an ab from th8 DVI file to the Versatec printer, placing breviated version of a memo recently sent to all pages of '&X output vertically or horiaontaHy on V.l@C users. The information contained here the printer. To function properly both programs should allow anyone who obtains the .VMSversion need the privileges AUSPOOL and PW- 10. When from Oregon Software to implement same on their running from the TM account the user must have VAX/VMS system. The only output device sup these privileges; if LVSPOOL and LHSPOOL are run ported at this time is a Versatec 1200 A or V80 in CSYSEXEI, the system manager can instal1 them printer having a DMA interface to the VAX. with these privileges so they can be used by any A preliminary VAX/lMS implementation of T)ijX, unprivileged user. with auxiliary programs to support a Versatec Although you could dart your first wperjence printer, is now available from Oregon Software. by using I)$ interactively, or with a file built us- TUGboat, Volume 2, No. 2 ing your system editor (using Knuth'e sod But, when it's good, it's very, very good-we're METAFONT as a guide), it is suggeuted that you converting our skeptics to 'QJCnicians. might want your first attempt to be with the file Details: T)iJC-in-Ppal for the VAX (11/750 and TYPO.TW. To run this example from the f'IPO 11/780) running WbIS, with about 50 Computer directory (without using the startup command file Modem fonts in assorted sizes. Comes with two provided) enter: spoolers (horizontal and vertical) for a Versatec R TM (Wait for the * prompt-- 1200 printerlplotter, using the standard Versatec this may take up to 20 seconds) interface. Uses about 7000 blocks of disk space \input typo (Note lowercase.) (mostly for fonts) and about a megabyte (whew!) Watch the l)ijC comments-wait for system $ ) of virtual memory while running. Comes in source R LVSPOOL or R LHSPOOL and binaryjexecutable forms, sources for spoolers, (the system should handle the rest) utilities, etc. (i.e., we'll send you everything we have). If you want to play with the Pascal pro- grams, you'll want the fancy listings available from The mailing list for future informational Stanford. A copy of the Pascal manual mentioned memos regarding VAX/'ZE)( will be the list of above will be included on request. DUES PAYING TUG members who indicate a VAX Supplied only on magnetic tape (600 foot) in interest, so make sure you join. "BCK/RSTn format (that's an MCR utility for backup/restore, which works well for binmy files). We can write 800 or 1600 b/in tapes-hearing noth- ing, we'll send 1600. Fees: fiky dollars will get you a tape and ship 'QX FOR VAX/VMS ment via UPS. To minimize our overhead, p+ Barry Smith don't send a tape, and do send a check or negotiable Oregon SoFtware securities 80 we don't have to deal with purchase or- ders and billing (I'd prefer small unmarked bills). If Well, it works-7)ijC for the VAX running VMS is you (or your fFjend) rrre in the "truly needy", just alive, available, and in production use. (Production let me know. use is defined by example-we've just finished a Maintenance: We'll be working on bugs and con- 192 page manual for our optimiring PDP-11Pascal venience improvements for the perpetual future, compiler that is entirely typeset by "&K, including and will be pleased to hear comments, suggestions, charts and diagrams.) gripes, bugs-no prorniees of any response except It's not y& perfect, nor something a "naive user" through the TUGboat and Monte's newsletters. should be expected to enjoy. To list the major deficiencies: ******* There are still annoying bugs, such as the Fonts darning crash that occurs when one tries to in- sert a footnote. Most of these seem to be dut to ******* bugs in Digital's V'Paecal, which just makes them harder to trace. (There's a new version FONT COMMiTTEE due out soon, always.) as Barry C. W. Doherty 0 It's the "old* ?jEX-in-Pasdacquired from Stanford in late December. (Some of the bugs Getting 'IjiJC up and running is only the first are real '&X bugs, too.) We've just received the step in producing output. One needs aleo an recent official release, and as soon as I get a free output device with an approprh driver and weekend, .. . fonts. llansportable 'QjX output requires compa- 0 There's absolutely no documentation what tible fonts. Those sites which have aim* output aoever that relatea to the VMS version. devik (Varians etc.) and are willing to use the Stan- (Gwon, another weekend, please!) The am- ford Computer Modern family have this problem iag thing ie that thie doesn't seem to matter- solved. Knuth's book describes the input format ex- There are many, however, who feel limited by actly, down to subtleties like tracing, and the the CM family adablefrom SWord, either for hatallation directions are rather concise: "put reasons of design, completeness or merely preference eveTything in mount [TEXJ ". for more traditional fonts (Helvetica, for instance).