haberman@s1.msi.umn.edu (Joe Habermann) (05/24/91)
Here is the summary of the replies to my request for opinions on the SPARCprinter and NeWSPrint software. Thanks to all who responded: hr@sirius.astro.uiuc.edu (Harold Ravlin) brsmith@cs.umn.edu (Brian R. Smith) ekrell@ulysses.att.com (Eduardo Krell) TOMP@YALEVM.BITNET (Thomas Plunkett) iapsd!hopi!glenn@uunet.UU.NET (Glenn Herteg) ehrlich@psuvax1.cs.psu.edu (Dan Ehrlich) hr@sirius.astro.uiuc.edu (Harold Ravlin) intran!clam!dale@uunet.UU.NET (Dale Mensch) doug@perry.Berkeley.EDU (Doug Neuhauser) wes@thor.srl.caltech.edu (Wes Boudville) Catch-all disclaimer: To the best of my knowledge, each of these people speak for themselves and not for their respective organizations. Joe Habermann haberman@msi.umn.edu >o Do the printer+workstation really do 12 ppm consistently even on > complicated PostScript? (I saw this claim in an article.) X-From: intran!clam!dale@uunet.UU.NET (Dale Mensch): Nope. I've never seen that claimed, either. The print engine will run 12 ppm. NOTHING can print "even complicated PostScript" at 12 ppm. X-From: hr@sirius.astro.uiuc.edu (Harold Ravlin): Once the printer starts cranking, a print job usually comes out at the rate of one page/5-6 seconds. I suppose more complicated pages could take longer, but most things I've seen are that fast. X-From: brsmith@cs.umn.edu (Brian R. Smith): Yes, on a SPARCstation 1+ with 16meg of RAM. (With or without a window system running.) No, on a SPARCstation 1+ with only 8meg of RAM. (And one user running sunview and FrameMaker.) X-From: ekrell@ulysses.att.com (Eduardo Krell): Yes, the printer does keep printing 12 ppm even with complicated graphics. X-From: TOMP@YALEVM.BITNET (Thomas Plunkett): It does print near the 12ppm mark, even with complicated postscript. The images being generated are Siesmological plots (FYI). Relatively complicated PS, and the SPARCprinter handles the work adequately. X-From: iapsd!hopi!glenn@uunet.UU.NET (Glenn Herteg): [I can't speak for the SPARCprinter, but I can talk about NeWSprint:] (1) Seems fast enough for general use. Very slow, though, to handle raster images (e.g., Sun rasterfiles), even when asked to do bit-for-bit printing. On the other hand, a PostScript printer connected over a serial line would have a different kind of bottleneck. (...) (3) Startup time for NeWSprint seems to be quite long, though the printing speed once it's rolling is okay. (...) (6) At this price, why do you care about constant 12 ppm output? I'd say it works very well now, and with a round of software bug fixes it will be essentially unbeatable. It beats most 8 ppm printers, and the real competition will be the next round of Adobe-PostScript Level 2 printers, which aren't out yet. X-From: ehrlich@psuvax1.cs.psu.edu (Dan Ehrlich): I bought NeWSprint so I could drive a Versatec color electrostatic plotter so I can not speak to the SPARCprinter's performance. ME: Incendentally, the article that I refer to in my question may be found in SunExpert, March 1991 Vol. 2 Num. 3 page 54 in an article about printer options for Suns. Not a terrible article. >o Have you found that the system is able to handle most PostScript > that you've thrown at it? X-From: brsmith@cs.umn.edu (Brian R. Smith): Pretty much. We haven't had any repeatable problems with FrameMaker output or text files. I have found a core file, but I couldn't say what job caused it. There was a mac document, though, that contained a blank page (lord knows why). The blank page put the postscript interpreter into an (apparent) infinite loop. I managed to print all the rest of the document. X-From: ekrell@ulysses.att.com (Eduardo Krell): I have had some problems with PostScript files which print ok on our PS printers but fail on the SPARCprinter. I think Sun has patches available for the NeWSprint software, but I haven't received them yet. X-From: TOMP@YALEVM.BITNET (Thomas Plunkett): We have found that SUN is a little "behind" in their support of the printer. We are printing to the SPARCprinter from a MAC, the directions for NEWSprint say this is possible, but it took time. I won't go into the problems, but I will tell you the solution, which is a simple one. You must use the LaserPrep 6.1 to print MAC PostScript succesfully. I ended up setting up CAP 6.0 on the IPC to allow AppleTalk service to the MACs, and by adding the LaserPrep 6.1 to the CAP LWSRV filter, it works great! Everything is printing to the server, and all users are happy. X-From: iapsd!hopi!glenn@uunet.UU.NET (Glenn Herteg): NeWSprint inherits all NeWS bugs. Perhaps not too serious, but I've run into some myself. OpenWindows 3 is supposed to be a bug-fix release, so that may fix most of such problems. X-From: ehrlich@psuvax1.cs.psu.edu (Dan Ehrlich): We have come across one or two examples that cause XNeWS to core dump. These have been sent off to Sun for analysis. XNeWS is very particular in regards to a postscript program following the structured comments that define page boundries and such. I am not really a postscript wizard so I just sent the offending program to my Sun SE. X-From: hr@sirius.astro.uiuc.edu (Harold Ravlin): Yes and no. Most Postscript we use is either lpr, the enscript command from Transcript or dvitps from the contrib directory of the TeX distribution. Most things seem to work. However, dvitps is using the Adobe .afm files as reference, but the SPARCprinter uses its own. Persons requesting Postcript fonts from TeX find that the text drifts left or right by about a 1/2 character per line. I don't know how to convert Sun's fonts to the data needed to dvitps. (The programs are there, but the formats are currently incompatable.) A number of programs that send images to the printer get faint lines running across the page. I don't know if it is a newsprint bug or just a really bad halftoning scheme. X-From: intran!clam!dale@uunet.UU.NET (Dale Mensch): Yep. Since it's not Adobe, there are some differences. The most noticable is in fountains: NeWS puts out a sort of dithered effect at the pale end of a fountain, where Adobe kind of gives up and puts out exponentially-spaced dotted lines. The only noticeable flaw I've seen is some bezier line ending connections within characters that are huge, graduated tints, eg: /iterate {25} def (abcdefg) true charpath stroke iterate -2 0 { /loop_var exch def gsave 1 loop_var iterate div sub setgray loop_var setlinewidth stroke grestore X-From: wes@thor.srl.caltech.edu (Wes Boudville): I recently installed Newsprint on a sun4, writing to an HP 3 printer. I encountered similiar problems. One of our users made a graph using Excel on a PC to generate a Postscript file. He sent this to the sun4 & thence to the printer. The top halves of the labels were omitted, and spurious vertical gaps appeared in the histogram. Axes numbers were also incorrectly placed. The same file was sent to another sun that had an Apple laserwriter and wasn't using Newsprint. It gave correct output. In another case, a user ran a graphical program called mongo on a sun. It made a Postscript file. This was sent to the HP via Newsprint. The errors here were subtler, but still apparent. The ends of the axes ticks were misaligned by ~ 1 pixel. And the labels looked smudged - almost every character had misplaced pixels. Sending this file to the Apple & not using Newsprint produced correct results. I informed Dan Quinn of Sun about this. He was very helpful. I emailed the Postscript files to him. He found that sending these via Newsprint to a Sparcprinter also gave the same problems. This isolated the problem to the Newsprint software. Because now the printers had been varied, and the input software making Postscript was from 2 different sources. He filed 2 bug reports with Sun Engineering. Sometime in the future, they'll fix it. >o Is the SparcStation printer host overly weighed down by performing > the raster-image processing? Is it very noticeable to a user > working on the workstation? X-From: brsmith@cs.umn.edu (Brian R. Smith): Not noticable to OUR users, but I can tell when a print job hits. You suddenly have a cpu-bound job running flat out, with the usual crunch on window drawing, etc. They don't last long, though. X-From: ekrell@ulysses.att.com (Eduardo Krell): I have it on a SPARCstation 2 and I don't feel any performance degradation when I use it (but I do hear the local disk seeking all over the place). We are using an IPC as the print server, and there is no noticable degradation of performance on the system when output is sent from another machine on our network to the printer. X-From: TOMP@YALEVM.BITNET (Thomas Plunkett): We are using an IPC as the print server, and there is no noticable degradation of performance on the system when output is sent from another machine on our network to the printer. X-From: ehrlich@psuvax1.cs.psu.edu (Dan Ehrlich): It does not seem to bother my 4/490, :-), too much. The Versatec is connected to a VME controller on the 4/490. X-From: hr@sirius.astro.uiuc.edu (Harold Ravlin): Sun claims that the printer will take up to 25% of the workstation. Since we don't print continuously (although I think the students try), I don't see it as being much of a problem, but then its not my workstation. (The printer itself doesn't care, everything is a bitstream). X-From: intran!clam!dale@uunet.UU.NET (Dale Mensch): Better have at LEAST 16MB of real memory >o Have you had any h/w problems with the SPARCprinter itself? Is it > noisy? X-From: ekrell@ulysses.att.com (Eduardo Krell): The NeWSprint software doesn't work when the spool directory is NFS mounted (rather, when the directory where the software keeps the temporary files for doing the rendering is NFS mounted). I've been told to install the NFS jumbo patch. I did that on the SS2, but it didn't make any difference. Maybe the patch needs to be installed on the server, but I haven't done that yet. X-From: iapsd!hopi!glenn@uunet.UU.NET (Glenn Herteg): (5) I wouldn't rcommend installing the SPARCprinter on a machine with only a local 105-Megabyte disk. There are various intermediate files generated during raster image conversion that may overflow the disk, if it's reasonably full to start with, even though the printer installation script supposedly tries to insure there is enough space. X-From: hr@sirius.astro.uiuc.edu (Harold Ravlin): Try listening to an Apple LaserWriter about 20,000 sheets passed the need for a major overhaul! I would rate the SPARCprinter as not a noisy laser printer. The cable is several feet long so the printer doesn't have to be right next to the SS. X-From: intran!clam!dale@uunet.UU.NET (Dale Mensch): No. It's largely soundless until it prints. The paper advance noise seems a little louder than 8ppm Canon engines, but it might just be the perception that the paper is moving faster... > o In general, would you recommend the system? Why or why not? X-From: brsmith@cs.umn.edu (Brian R. Smith): Er.... no. The software is still a bit shaky. If you have the imaging directory on an NFS partition it will either: Not print the entire image (Sun's claim; haven't seen it happen) or Hang your machine (what happened to us - Sun was clueless.) So: You NEED a local disk on that machine, in addition to at least 12meg. X-From: doug@perry.Berkeley.EDU (Doug Neuhauser): 1. The idea of using the SPARC engine for PostScript->Raster is a good idea. 2. I think that their current implementation certainly needs work. (...) Lines did not come out with uniform thickness. "T" intersections did not line up correctly (perhaps due to #1 above). There were "spurious dots" on the the page, which could be also be seen using pageview. 3. IMHO, a write-black engine would be better than a write-white engine. 4. Presumably the rendering portion (xnews) is given any information about the output device resolution, but I don't see anything in the .param files for the printer that indicate the engine type (write black vs write white). It DOES make a difference, especially in the rendering of small thin fonts. X-From: TOMP@YALEVM.BITNET (Thomas Plunkett): Essentially, I am recommending the printer, its only major drawback is the lack of people at SUN supporting it. I am sure they will correct this problem soon. X-From: ehrlich@psuvax1.cs.psu.edu (Dan Ehrlich): The concept of NeWSprint is appealing if you have a number of raster marking engines that you want to quickly (i.e. code development time) be able to get postscript output to. X-From: hr@sirius.astro.uiuc.edu (Harold Ravlin): In general, yes I would if you can live with the quirks. Newsprint was released with with lots of bugs. Hopefully Sun will spend more effort fixing them than they have on other unbundled software.