Charles@uw-beaver (Charles) (12/01/83)
The following message was initially written in response to someone who asked about our experience with the 9700. He suggested I should send it to this mailing list, so I am doing so. I have might slight changes to make the message more appropriate for general distribution. Overall evaluation ================== This list has been used mostly for discussion of the $20K type of printers. The 9700 is obviously quite a different kind of beast. It is built for constant production use in large-volume applications. It produces good quality output (when properly adjusted), very quickly. I do not know any real competition. From all that I have heard the other big laser printers don't have quite the wide range of capabilities as far as handling multiple fonts. The smaller printers are not built for a production environment. So if you really need what the 9700 does, and if you understand what you are getting into, it is a good thing. The main purpose of the rest of this memo is to suggest things that you should look into if you are considering getting such a system. Reliability and hardware service ================================ Our 9700 fails a lot. We call for service every day or so. Some people have suggested that this is because it is built to be used flat out, and we don't use it enough. It probably also has something to do with the people who operate it. We use students who are not properly trained and are effectively not supervised at all. (We found an operator on duty on time who didn't know that you had to load paper into it.) However we have some evidence that the failures may simply be inherent in the task of pushing 2 pieces of paper per second through all of this machinery. It does not cause the bad impact that one might think it would. Xerox is well set up to handle this failure rate: They have a good field service organization, and it responds quickly. So it does not actually cause us a problem, except when users get themselves into situations where they are doing final draft 5 minutes before a deadline. The only real problem we have had with service is scheduling PM. We strongly suggest that anyone who gets a 9700 negotiate this before signing on the dotted line. We had a number of problems in getting PM to happen at an acceptable time. In this forum their is no point discussing whether this is our problem or Xerox's. Knowing some of the staff involved with it, there could well be problems on our end. But the point of this note is not to place blame, just to suggest things that prospective users should know about. So the suggestion stands: make sure you have an agreement in writing as to when they will do PM. Doing odd things ================ Xerox views the 9700 as a line printer replacement, with the ability to define special forms. We want to use it as a multifont output device for Scribe. The most positive thing one can say is that Xerox provides no help in this endeavor. In order to use Scribe or any other similar system, one must feed the 9700 what are called "metacodes". There are the commands that allow you to place arbitrary characters at arbitrary positions. Xerox is unwilling to release any of the following information: - the formats in which fonts are stored - what the metacodes are - what command to give to the 9700 to allow you to print files involving metacodes. Some additional explanation is needed for the last point. Because of its high-volume production orientation, the 9700 has a command language of its own. This allows you to define characteristics for each job, such as the fonts to be used for it, the orientation (portrait or landscape), the format of the tape being read (if you use tape input), etc. There are a number of very useful facilities available through the use of this command language, far more than you might imagine if you think of the 9700 as being a lineprinter replacement. However in order to print Scribe output you need one additional option that Xerox will not tell you. Furthermore, Unilogic (the supplier of Scribe) is not allowed to tell you either. So you must describe to them the commands that you are going to need to print your Scribe output. They will then add the magic keyword, compile the file, and send you a binary version. If you are doing anything nonstandard, this could lead to quite a dialog. (In fairness, it should be said that in most installations Unilogic can supply you with an off the shelf file that will work fine. We are routing Scribe output through an extremely unusual path, so things were more complex for us.) If you only need to do off the shelf things, you may find this sufficient. We did not. This is of course a marketing issue. There is not question that Xerox has the right to decide the applications that it is interested in supporting, and to adopt policies consistent with that. They never claimed that the 9700 was intended as a phototypesetter. However a number of people find it frustrating to know that the capabilities are there and we can't get at them. Fortunately, the situation is now better than it was when we first started this. The 9700 user community does now know what the metacodes are, how to enable them, and what the format of the font files is. I believe we are free to give out at least the following information: - a program (written in Interlisp) that will display on a bitmapped screen files produced by Scribe with its @DEVICE(X9700) option. - a program that can be used to dump and modify Xerox fonts. Someone suggested that there might be a legal problem with this. However we debugged our font editor using a font provided by a third party. The font was digitized from copy that I drew by hand. Let me make some comments on the resources needed to use Scribe (and presumably other similar systems): - you pay for font memory. All fonts that are going to be used on a given page must be in memory at once. The whole font must be there. We have either 512 or 768 Kbits of font memory. We put 10 fonts in at once, including boldface and italics, some large fonts, Greek letters, and math symbols. In practice we find that this is all that is needed. I strongly recommend having at least 512Kbits. - the file then contains commands to change fonts, and to move to any position on the page. So you can put any character in any font at any pixel. Note however that the commands for doing this are the metacodes referred to above. - fonts must be stored on the system disk. You cannot put new ones on while the system is in normal operation. So this makes it hard for users to play with fonts. You more or less have to schedule standalone time to try a new one. It is not clear to us that having users design their own fonts is all that useful anyway. About Operations: ================ In order to use special font capabilities, most software switches into a special mode. (In principle it is probably possible to avoid doing this, but as far as I know, all fancy typesetting programs do in fact use metacode mode.) This mode is not compatible with normal line printer simulation. If you are using magtape input that is no problem, since each tape will require an operator command anyway. But if you are running online, realize that you will have to drain the queue on the IBM system, change modes on the 9700, change forms types on the IBM system, and restart output. Similarly when you are finished. In general, the 9700 seems to assume a full-time operations staff. If nothing else, somebody has to take paper out of it, and keeps its input filled. The bins are fairly large, but you wouldn't just leave it running unattended during the day. The printer jams now and then, just as copiers do. So you will need an operator capable of doing the following: - clearing printer jams - feeding paper in - taking printout out - starting and stopping the printer when the users need special modes - if offline, mounting tapes and knowing what command to type for each format of tape (this assumes you have several kinds of machine feeding it) I strongly recommend using fulltime people, or at least having careful supervision of students. You would have to be crazy to try running a 9700 as a self-service printer. About Fonts =========== I agree with the comments made on this list about fonts. It is difficult to get a font that looks nice on a 300 by 300 bit matrix. Perhaps there are a few people in the world who can make fonts that really look good. So far I have not seen any. When I was at CMU, they commissioned one of these supposed experts to do something really exciting. It was indeed exciting, but I sure wouldn't want to read a book written in it. Mostly the practical solution seems to be to stick with simple things. The fonts supplied with the 9700 seem to bear this out. Xerox supplies two font families with the printer. They call them Univers and Times Roman. Univers is a simple sans-serif font. It is certainly not a work of art. But it looks even, and is readable. It is not as readable as a well-typeset book. But it is better than output produced by typewriters. Times Roman is your typical Roman font. I.e. it has serifs, and the widths of the lines vary in what is supposed to be an esthetic manner. It is supposed to look like the fonts used in most newspapers and books. This font is much less successful than the Univers. When you run your eye over a page, individual letters jump out at you as being heavier than the letters next to them. I find this very distracting. Some of the narrower parts of letters tend to disappear. I suggest that you stick with Univers. It isn't as ambitious, but does what it is supposed to do. There are also fixed-width fonts. You use them when you are running the 9700 as a lineprinter, and for certain pieces of text that where you need to duplicate the effect of a terminal or printer. The fonts that are taken from the Xerox 1200 are readable, and are reasonably compatible in style with the Univers fonts. There are some problems with details. The most serious is the fact that Xerox does not supply any full ASCII fonts. They are missing accent grave, and tend to have less used characters in randomly chosen positions. (I.e. different size fonts will have slightly different character sets, with no obvious reasons.) We have defined a standard character set, and made up fonts that implement it for the most useful sizes. (No, we didn't try to create our own fonts. We just moved characters around.) The next most serious is the fact that none of the big fonts include lower case letters. We sometimes do titles by using 14 point upper case letters for the capitals and 10 point upper case letters for the lower case. This works well enough to be useable, but I would love to get away from it. (No, we haven't trained ourselves to think of it as looking right.) There is a good reason for this, by the way: A large font with a full character set would take a lot of font memory. I think we would be willing to buy the extra font memory to hold them. We had some problem getting mathematical symbols. Xerox does not supply fonts that have the mathematical symbols that our theorists want to use. We have created a font for that. The process we went through in doing so is illuminating. You can get Xerox to create fonts for you. The price isn't even too outrageous. The problem is that (at least the last time I checked) they did it by mechanical digitizing. So you had to supply them with the fonts in the correct size and style. We were unable to find any one source that had all of the characters we wanted in a compatible style, the correct size. We tried to find a typesetting house that we could pay to do it, and found that no local typesetter will touch mathematical copy. We went to Texas Language Research. They will create the font for you from hand drawings if necessary. However the person who did it was new, and botched it fairly badly. We went back and forth about 3 times before they came close. By that time I had given up and written a font editor, and did it myself. [Let me say that I think our experience with Texas Language Research is probably a fluke. They are very good to deal with, and were willing to stick with it until we were satisfied. It is just that I had run out of time.] I can testify to the comments about the difficulty of doing good fonts. Fortunately I didn't have to do letters. It is harder to botch a left arrow than a letter. My abstract symbols look fairly good. My Greek letters weren't as successful. Fortunately we are now using the Xerox proprietary Greek letter font, which is quite reasonable. So when you combine that with my symbols, we have a symbol font that is almost up to the quality of the Univers text. How to get data to your X9700 ============================= The 9700 comes in two versions: online and offline. Online machines get their data from an IBM system. They look like line printers to the IBM system. Offline machines get data from tapes. They can read an astonishingly wide variety of tape formats. You have to analyze your shops operations. If you have enough operators to be writing tapes on all your machines and constantly shuffling them to the 9700, then that may be the way to go. It requires minimal modification to software on your systems, since most spoolers can write tapes. The problem is that our users have become addicted to 9700 output. They want their output immediately. We find that turnaround time for printout causes an order of magnitude more problems with users than anything else. Even when it is fairly good, as it is here. We use online operation. Then we use IBM RJE interfaces to submit printouts from all other systems to the IBM system. This generally requires special hardware and modifications to the spoolers. We have probably spent a man-year working on the DEC-20 to get this to work smoothly. I am not sure whether it is worth it. I think we might have been better off with a smaller laser printer attached to each machine. As I am sure you know, there are now a number of printers available for $20K. They would have the advantage of being much easier to play with. I.e. to generate your own fonts, and generally write special software to take advantage of their capabilities. But the one big advantage of the 9700 is that it is *fast*. We keep all of our manuals on line. A class can each print a 300-page manual without causing serious problems. The small printers can't be used for that kind of thing. If you go for online operation, you should strongly consider having a tape drive also. Strictly it is not needed. We don't have one. But all software distribution and backup is via tape. So Xerox software support people have to back up our disk by taking it to a nearby site. And similarly when a new release comes out they take away a disk pack and build it elsewhere. It works, but I can imagine situations where it would not. The Organization ================ From my perspective, there is some room for improvement in the Xerox organization. Some of it is just orientation. We all know what we would like to use the 9700 for: Scribe and Troff. They obviously don't agree, since they have not made it easy on those of us who have tried to do this. But as I said above, every vendor has to decide how their product is to be oriented, and there are bound to be people who will disagree with whatever they do. What is harder to explain are couple of other problems we have had. The most serious may have been due to an accident. We leased a 9700 for a year or two. Finally we decided to buy it. We made it clear to them that our intent was to buy the machine we had been leasing. The day we signed the contract, they came out and removed 256Kbits of font memory. It turns out this had been present by mistake, and they just discovered it at that time. This led to about a year of somewhat heated discussions. It was finally resolved, but during the meantime our users were unable to print their old Scribe output, because the fonts it used wouldn't fit. The second episode was something that has given me pause about sending this communication at all. I answered a request from an Arpanet user about our experience. My note ended up in Xerox's hands. As it happens, I saw the various covering letters where Xerox people forwarded it to other Xerox people. The final covering letter directed someone to take the appropriate action. We will hope that the manager who wrote this did it in good faith, and some subordinate had simply seen The Godfather too recently. Anyway, Xerox sent lawyers to my boss's office, demanding a retraction. They came out about 3 times, with escalating threats. I finally got them off my back when I pointed out that my note had been a private communication, and had been circulated on the Arpanet by someone who was employeed by Xerox. (By the way, at no time did they claim I had said anything incorrect.) It will be interesting to see whether I get a similar reaction to this note. (I would feel a lot better about Xerox if someone in PSD who reads this list would arrange to have a written apology sent to me. I know enough about Xerox as a company to know that the upper level management wants to deal with customers fairly.) In both cases, my tone was the same: The 9700 is basically a good product. The design takes seriously the challenges inherent in designing a machine for use in a large production shop. (These require a bit more than just speeding up an Imagen.) I think we were right to get it. But like any other product and any other organization, there are limitations and problems. Potential users should know what those are and be prepared to deal with them. If this is slander, then the computing community needs to have a long talk with the legal community about the purpose of laws. -------
laser-lovers@uw-beaver (12/08/84)
From: decvax!esquire!newman@uw-beaver.arpa Troff can also be made to work with the 9700. However, one must get a META-CODE agreement from Xerox to do so. We have been using the 9700's to do typesetting for four years here by first writing our own code and then purchasing a more sophisticated package from a consultant. The limitation with the 9700's in typesetting is the extremely limited font memory. Right now you have room for about 7 of the high quality Merganthaler Fonts. However, it seems that with the next release of software, one will be able to replace fonts in the font memory AFTER the job starts and thus increase the number of fonts available in each job. The new software also will support new font memory which will increase storage from 1024 K-bits to either 8 or 16 Mega-bits. Note: this is font memory to be used in each job, not the total font storage of the machine. Finally, I haven't heard of a higher resolution Xerox machine and I would doubt an early release of one. Xerox has standardized so far on 300 DPI and has made a heavy investment in good fonts for these machines. However, they have released the 9700 MODEL 5, which upgrades the Xerographic engine to the 9500 copier engine. We just had ours upgraded to this last week and it is bea-yoo-ti-full. Edmund Newman Davis Polk & Wardwell 1-212-530-4420
laser-lovers@uw-beaver (12/17/84)
From: ihnp4!utzoo!henry@uw-beaver.arpa
Various people have flamed me for characterizing the 9700 as a big dumb
line printer. (I note that nobody has argued with this description of
the IBM laser printer, though.) Several of them have pointed out the
existence of graphics and typesetting packages for it. So I'll modify
my previous observations...
If you want to run the 9700 as anything but a big dumb line printer,
make sure that the exact software you need, with the exact fonts you
need, is available; insist on seeing it demonstrated. Writing it
yourself is not very practical, since getting the necessary information
out of Xerox is quite a chore. As far as I know, it is not possible
to do your own fonts at all; you have to buy them from Xerox, which has
its own odd ideas of what should (and shouldn't) be in a font. Note in
particular that if you need full ASCII in your fonts, insist that Xerox
show you; many of their fonts are not full ASCII. Do not, repeat, not,
assume that something which looks "close enough" can be convinced to do
exactly what you want; insist on a 100% hit.
Yes, I have worked on a 9700. We never succeeded in running it as
anything but a big dumb line printer, although other people have.
The company which bought U of T's 9700 (sold for complicated reasons)
did so on faith that it would prove possible to exploit the considerable
potential of the device. They eventually sold it in disgust, because
the means to do so had not materialized from Xerox. Caveat emptor.
"At heart, Xerox is still a photocopier company; they're trying to change,
but it's difficult and slow." -- Butler Lampson
Henry Spencer @ U of Toronto Zoology
{allegra,ihnp4,linus,decvax}!utzoo!henrylaser-lovers@uw-beaver (12/17/84)
From: Brian Reid <reid@Glacier> > "At heart, Xerox is still a photocopier company; they're trying to change, > but it's difficult and slow." -- Butler Lampson > > Henry Spencer @ U of Toronto Zoology > {allegra,ihnp4,linus,decvax}!utzoo!henry Allow me to point out that Butler Lampson now works for DEC, which is a VMS company instead of a copier company. As nearly as I can tell from my vantage point at Stanford, Xerox is no longer trying to change, and in fact they are regressing back into being a copier company..... [go ahead, Xerox folks on this mailing list, flame me out.....]